Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Diverging fertility patterns? Racial and educational differences in fertility behaviors and their implications for socioeconomic mobility
(USC Thesis Other)
Diverging fertility patterns? Racial and educational differences in fertility behaviors and their implications for socioeconomic mobility
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
DIVERGING FERTILITY PATTERNS? RACIAL AND EDUCATIONAL DIFFERENCES IN FERTILITY BEHAVIORS AND THEIR IMPLICATIONS FOR SOCIOECONOMIC MOBILITY by Sandra M. Florian A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (SOCIOLOGY) May 2016 Copyright 2016 Sandra M. Florian ii Dedication To my dad, whose value in education and hardworking ethics have inspired me throughout my life. To life, because being alive is precious and should not be taken for granted. To knowledge, because it makes life stimulating, challenging, and enjoyable. iii Acknowledgements I am very thankful to the members of my dissertation committee for all their support and guidance that helped me accomplish this work. First, I would like to express my gratitude to my adviser and committee chair, Dr. Lynne M. Casper, who influenced my areas of study in demography and family sociology, challenged me, and supported me throughout my Ph.D. program. Her mentorship helped me developed my research questions, put them into a broader sociological framework, and made me think about the importance and implications of my research. Thank you Lynne for your understanding and patience. I am grateful to Dr. Tim Biblarz, a man with a keen analytical mind and one of the best instructors in statistics. Tim inspired me to continue my specialization in quantitative research methods, always stressing the importance of telling the story behind the numbers. Special thanks to Dr. Jennifer Hook, who read my work several times and provided me with valuable feedback. I am also grateful to Dr. Eileen Crimmins for her constructive feedback from the inception of this project and her constant support during this time. I would also like to thank Dr. Dowel Myers for introducing me to the field of Demography and for guiding my research as I started my academic life. I am grateful to Dr. Robert Mare, whose hardworking ethics, sharp mind, and research have been great sources of knowledge and inspiration. Thanks to Dr. Jody Agius Vallejo for her thoughtful advice and guidance as I began developing my research interests. A special thank you goes to Dr. Edson Rodriguez, Demetri Psihopaidas, and Yu-Kan Fan, whose iv friendship, advice, and support gave me strength, and made my time as a graduate student more enjoyable. I would also like to thank Dr. Radheeka Jayasundera for her great mentorship and support as I started the Ph.D. program. Finally, I gratefully acknowledge financial support from the University of Southern California throughout these years, including funding from the USC Graduate School College Doctoral Fellowship, the Russell Endowed Fellowship, and the Dissertation Completion Fellowship, which made possible my research by generously funding my training and this dissertation project. v Abstract Sociologists have long recognized that family structure interacts with individuals’ characteristics, most notably their class and racial/ethnic identity, shaping socioeconomic outcomes, and broader patterns of social stratification. A vast literature indicates that social inequality is exacerbated by patterns of family formation, particularly by childbearing behaviors. In this dissertation I analyze the trends and patterns in fertility behaviors in the U.S. from 1973 until 2013, including age at first birth, fertility rates, and levels of childlessness. I identify the socioeconomic and demographic factors associated with these trends and evaluate how they vary by race/ethnicity and social class. Then, I investigate the effects of fertility on women’s opportunities for social mobility. The analyses were performed using two nationally representative datasets: (i) the National Survey of Family Growth (NSFG), from 1973 until 2010–2013, and (ii) the National Longitudinal Survey of Youth, (NLSY79) 1979–2012. Chapter 3 begins with an analysis of the trends in age at first birth and fertility rates using data from the NSFG from 1973 until 2011–2013. Then, using event history models, Poisson regression, and a variation of Hurdle models, I conduct multivariate analyses explaining racial differences using the 2006–2010 and 2011–2013 waves. The results indicate that racial minorities start childbearing at younger ages and have more children than Whites. However, the racial disparities in age at first birth have stopped diverging since the early 2000s, while the differences in fertility rates are relatively small. My analyses reveal that most of the racial differences are concentrated at low levels of education, whereas no significant racial vi differences exist among college graduates. The findings suggest that racial differences are superseded by attaining a college education. The results support the racial stratification perspective, which places structural factors at the base of racial/ethnic disparities, but also considers the importance of cultural norms and the legacy of historical discrimination constraining the fertility behaviors of disadvantaged racial minorities. Chapter 4 analyzes the trends and patterns of childlessness, an important fertility outcome that has received less attention in demographic research. Despite the increased prevalence of childlessness from the 1980s until the late 2000s, little research has studied its socioeconomic and demographic correlates or compared them to patterns of childbearing behaviors. Using data from the National Survey of Family Growth (NSFG), I analyze changes in the associations between childlessness and key socio-demographic factors, namely, partnership and marital status, race/ethnicity, education, and income, from the 1980s to the late 2000s. Contrary to the analyses presented in Chapter 3 that revealed a bifurcating pattern in childbearing behaviors, I found a convergence in the prevalence of childlessness among women ending their reproductive life by marital status, education, and, race/ethnicity; however, I found a growing divergence by economic well-being. I claim that even though education, race, and marital status have remained strong predictors of childbearing behaviors, these factors have become weak predictors of childlessness over time. Based on the results in Chapter 3 and Chapter 4, I argue that the mechanisms that shape the patterns of childbearing are different from those that shape childlessness. vii In Chapter 5, I explore the effect of fertility on women’s opportunities for social mobility. Because for most individuals, employment is an important means for social mobility, I assess the effect of fertility on women’s employment over the life cycle, using fixed effects models and data from the National Longitudinal Survey of Youth, NLSY 1979–2012. I use two different measures of employment: (1) labor force participation, a temporal measure, and (2) cumulative years of work experience, a long-term measure, further distinguishing between full-time and part-time employment, and paying special attention to variations by race/ethnicity. The analyses reveal declining effects of children on mother’s labor force participation over time, but enduring effects on cumulative years of work experience. Children discourage full-time employment among mothers during their 20s and 30s, but encourage full-time employment by their early 50s. These effects are stronger among Whites, and smaller among Hispanics, but, remarkably, motherhood does not deter employment among Black women. However, having children reduces cumulative years of work experience for all mothers regardless of race; these effects become evident during the 30s, and amplify during the 40s, being more pronounced for Whites, Hispanics, and women with two or more children. By the early 50s, White mothers accumulate more years of work experience, however, the results reveal that White mothers’ advantage results from gains in part-time rather than full-time work. This study shows that aggregating the effects of fertility over time and across racial groups obscures these significant variations. In sum, the results of this dissertation indicate a decreasing predictive power of race over time as a determinant factor of childbearing behaviors, and an increased viii importance of education. I argue that college education constitutes a strong homogenizing force of fertility behaviors across racial/ethnic groups, emerging as an important tool for policy implementation. In addition, despite gains towards gender equality in the labor market, the findings show that motherhood still hinders women’s employment, limiting their opportunities for upward social mobility. These findings indicate the need for more policies supporting working mothers to mitigate the effect of motherhood as a source of social inequality. ix Table of Contents Dedication .......................................................................................................................... ii Acknowlegments ............................................................................................................... iii Abstract ............................................................................................................................... v Table of Contents ............................................................................................................... ix Chapter 1: Fertility Patterns and Social Inequality: An Introduction ........................ 1 1.1. Introduction ............................................................................................................. 1 1.2. Importance of the Project ........................................................................................ 3 1.3. Research Plan and Methodology ............................................................................. 4 1.3.1. Childbearing Patterns: The Timing of First Births and Fertility Rates ................ 5 1.3.2. Childlessness over Time ....................................................................................... 7 1.3.3. Evaluating the Effects of Fertility on Women’s Employment ............................. 8 Chapter 2: Data ............................................................................................................... 10 2.1. Introduction ........................................................................................................... 10 2.2. National Survey of Family Growth (NSFG) ......................................................... 10 2.3. National Longitudinal Survey of Youth (NLSY79) .............................................. 13 Chapter 3: Beyond Race? The Homogenizing Effect of College Education on Fertility Behaviors ..................................................................................................... 16 3.1. Introduction ........................................................................................................... 16 3.2. Background ........................................................................................................... 18 3.3. Theoretical Perspectives: Explaining Racial Inequality ........................................ 22 3.4. Competing Hypotheses .......................................................................................... 27 3.5. Factors Affecting Fertility Behaviors .................................................................... 28 3.6. Contribution to the Literature ................................................................................ 30 3.7. Data, Measures, and Methods ............................................................................... 32 3.7.1. Data .................................................................................................................... 32 3.7.2. Measures ............................................................................................................. 33 3.7.3. Methods .............................................................................................................. 35 3.8. Results ................................................................................................................... 38 3.8.1. Descriptive Results ............................................................................................. 38 3.8.2. Multivariate Results ........................................................................................... 40 x 3.8.2.1. Racial/Ethnic Differences in the Timing of First Birth ................................... 40 3.8.2.2. Racial/Ethnic Differences in Number of Children .......................................... 44 3.8.3. Summary of findings .......................................................................................... 48 3.9. Conclusions: Beyond Race? .................................................................................. 50 Chapter 4: No Nest? The Converging Growth of Childlessness in the United States, 1980s-2000s ............................................................................................. 56 4.1. Introduction ........................................................................................................... 56 4.2. Background ........................................................................................................... 58 4.2.1. Childlessness across Marital and Partnership Status .......................................... 59 4.2.2. Trends in Childlessness by Educational Attainment .......................................... 60 4.2.3. Childlessness by Race/Ethnicity ........................................................................ 63 4.2.4. Childlessness and Income .................................................................................. 64 4.3. Data, Measures, and Methods ............................................................................... 65 4.3.1. Data .................................................................................................................... 65 4.3.2. Measures ............................................................................................................. 67 4.3.3. Methods .............................................................................................................. 69 4.4. Results ................................................................................................................... 70 4.4.1. Descriptive Results ............................................................................................. 70 4.4.2. Multivariate Results ........................................................................................... 77 4.5. Conclusions ........................................................................................................... 86 Chapter 5: Racial Variations in the Effect of Fertility on Women’s Employment: Declining or Enduring Effects? ................................................................................... 93 5.1. Introduction ........................................................................................................... 93 5.2. Background ........................................................................................................... 95 5.2.1. Gender and Race in the Labor Market ............................................................... 96 5.2.2. Intersectionality across Stages of the Life Course ............................................. 98 5.3. Measuring Employment: Temporal and Long-Term Measures .......................... 100 5.4. Contribution to the Literature .............................................................................. 101 5.5. Data, Measures, and Methods ............................................................................. 102 5.5.1. Data .................................................................................................................. 102 5.5.2. Measures ........................................................................................................... 103 5.5.3. Methods ............................................................................................................ 105 5.6. Results ................................................................................................................. 106 xi 5.6.1. Descriptive Results ........................................................................................... 106 5.6.2. Multivariate Results ......................................................................................... 108 5.6.2.1. Motherhood and Women’s Labor Force Participatio .................................... 108 5.6.2.2. Race/Ethnicity and Women’s Labor Force Participation over the Life Course ...................................................................................... 112 5.6.2.3. Motherhood and Cumulative Work Experience ............................................ 114 5.6.2.4. Cumulative Work Experience and the Persistent Salience of Intersectionality ............................................................................................ 118 5.6.2.5. A Note on Marriage and Race ....................................................................... 120 5.7. Conclusions ......................................................................................................... 121 Chapter 6: Conclusion and Discussion ................................................................... 126 6.1. Summary ............................................................................................................. 126 6.2. Contributions to Current Research ...................................................................... 132 6.3. Limitations ........................................................................................................... 134 6.4. Discussion and Implications ................................................................................ 137 Bibliography .............................................................................................................. 140 Appendix ................................................................................................................... 151 Appendix A: Additional Findings for Chapter 3 ........................................................ 151 Appendix B: Additional Tables for Chapter 4 ........................................................... 153 Appendix C: Additional Tables for Chapter 5 ........................................................... 155 1 Chapter 1 Fertility Patterns and Social Inequality: An Introduction 1.1. Introduction Since the 1970s, the U.S. experienced significant changes in family formation and fertility patterns. Until the 1960s, childbearing followed marriage, and this was the standard life course trajectory for most Americans. Marriage used to be the only legal institution for childbearing, and although, marriage is no longer a requirement for children’s legal rights or for parental responsibilities, significantly better outcomes for children occur within marriage. Nonetheless, marriage rates have declined as individuals are waiting longer to settle with a partner (Cherlin 2004; Goldstein and Kenney 2001; Popenoe 1993). But, while most individuals are postponing marriage, not all demographic groups are postponing childbearing. Nonmarital births have become an important element of fertility in the U.S. In 1960, 5% of all births occurred out-of- wedlock; this percentage increased to 41% by 2013 (Martin et al. 2015). These trends evidence a decoupling of marriage and childbearing (Edin and Kefalas 2005; Ellwood and Jencks 2004; Goldstein and Kenney 2001). Having a child has now become a more common pathway to start a family. Fertility is a demographic behavior that has direct socioeconomic implications. The timing and level of fertility affect family structure and the resources available to 2 family members. Although maternal age has steadily increased as more women are delaying childbearing, this increase has been modest among women of lower socioeconomic status and disadvantaged racial minorities (Bongaarts 2001; Ellwood and Jencks 2004; Morgan 2003), resulting in a growing divergence in age at first birth and nonmarital births along social class and ethnic lines. Early births are associated with higher subsequent fertility, and having more children imply that parents need to spread family resources among more people. Scholars have argued that the diverging demographic trend in family formation, particularly in childbearing behaviors, has exacerbated the recent growth in social inequality (McLanahan and Percheski 2008; McLanahan 2004, 2009; Western, Bloome, and Percheski 2008). The main goals of this dissertation are to analyze the patterns of childbearing behaviors by race and social class, and evaluate their links to socioeconomic inequality. More specifically, in this dissertation I: (1) analyze the trends in age at first birth, total fertility, and rates of childlessness in the U.S. from the 1970s until the early 2010s, (2) identify the socioeconomic and demographic factors associated with the patterns of childbearing and childlessness, investigating the role of race and education, an important indicator of social class, (3) evaluate the predictive power of key socioeconomic and demographic factors to explain fertility outcomes, and investigate how their predictive power has changed over time, (4) investigate the extent to which racial/ethnic differentials in fertility behaviors persist, and evaluate whether the data support the various theories that have been proposed to explain these variations, and (5) explore the implications of fertility trajectories for women’s opportunities for social mobility. 3 1.2. Importance of the Project This research elucidates the debate surrounding unequal socioeconomic origins and the significance of race on individuals’ life chances. Using data for recent cohorts of women, this dissertation links lines of research that have previously been investigated separately, or that have focused on unidirectional causality among childbearing, childlessness, and social inequality. Moreover, most of the research on socioeconomic inequality and stratification has focused on men, with little attention paid to the experiences of women and families. Analyzing inequality among women is essential to provide a more complete account of overall patterns of social stratification. This dissertation aims at assessing the extent to which fertility behaviors shape and are shaped by race and socioeconomic characteristics. Evaluating whether childbearing patterns are still diverging is important because early childbearing and higher fertility can prevent women from achieving educational and occupational goals, hindering their economic opportunities (Brand and Davis 2011; Martinez, Daniels, and Chandra 2012; Rumbaut 2005). Investigating jointly the patterns of childbearing and childlessness is important to understand the extent to which opportunities for social mobility may be blocked by early and higher fertility, or facilitated by delaying or forgoing childbearing. An important goal of this dissertation is to evaluate whether socioeconomic factors, such as education and income, have become more consequential for fertility outcomes in recent decades than ascribed characteristics, such as race and ethnicity. This study demonstrates the growing importance of education, and also reveals the persistence salience of race shaping childbearing outcomes among the lower class. By identifying vulnerable populations and evaluating the mechanisms that account for social and racial 4 disparities, the results of this dissertation can help policy makers devise effective policies to better the lives of women and children with limited opportunities for social mobility. 1.3. Research Plan and Methodology This dissertation is divided in five chapters. In the rest of Chapter 1, I introduce each of the topics discussed in the following chapters. Chapter 2 provides a description of each of the datasets used in this study. All the analyses were performed using data from two U.S. nationally representative surveys: (i) the National Survey of Family Growth (NSFG), 1982 to 2013, a cross-sectional dataset (i.e., data collected at different periods), suitable for the analysis of fertility trends; and (ii) the National Longitudinal Survey of Youth, (NLSY79) 1979-2012, a longitudinal dataset (i.e., data collected over time from the same individuals), appropriate for the study of causal effects. In Chapters 3 and 4, I analyze the trends in childbearing and childlessness, paying attention to the socioeconomic factors associated with the bifurcation of fertility behaviors, including marital status, education, race/ethnicity, and income. Chapter 5 aims at assessing the consequences of different fertility trajectories for women’s socioeconomic outcomes. The study of the socioeconomic determinants and consequences of fertility behaviors has been challenged in the past by data and methodological limitations. However, recent releases of harmonized cross-sectional and longitudinal data, and advances in quantitative methodologies for data analysis make a more thorough study possible. I use several advanced statistical tools, as described in each chapter, including event history analysis, a method suitable for the study of events occurring over time, such as childbearing, and Poisson models, a method that predicts small counts, such as the 5 number of children. In Chapter 3, I innovate by using a variation of Hurdle models to evaluate the extent to which racial differences in total fertility result from variations in the proportion of women entering motherhood. In Chapter 5, I use multi-level fixed- effects models to assess how changes in parity, i.e. the number of children, affect mother’s employment over the life course, analyzing how the salience of race evolves over time. Whereas prior studies have often restricted this type of analysis to mothers only, these statistical tools allow me to jointly model the fertility trajectories of mothers and childless women. 1.3.1. Childbearing Patterns: The Timing of First Births and Fertility Rates Chapter 3 focuses on the analyses of the trends and patterns of age at first birth and the number of children. Until the 1970s, the onset of fertility did not differ much by socioeconomic characteristics (Dye 2010). Since the 1970s the mean of maternal age at first birth began steadily increasing, however, at a faster pace among the highly educated and Whites, resulting in a growing divergence in the timing of entrance into motherhood. Maternal age is significantly associated with women’s and children’s outcomes. Prior research indicates that older mothers are more likely to plan and want their pregnancies, have more stable relationships, and hold steady jobs (McLanahan 2004; Musick 2002). Women who delay childbearing also tend to have fewer children, which implies more resources for each offspring (Bianchi 2000; Goldstein and Kenney 2001; McLanahan 2004, 2009). By contrast, younger parents tend to be more immature, and form less stable relationships, which implies that their children are at a greater risk of experiencing family 6 disruption (McLanahan 2009; Wu and Martinson 1993). Some scholars have argued that having children at early ages limits young women’s ability to complete their education and thwarts their occupational trajectories (Landale, Schoen, and Daniels 2010; Rumbaut 2005). Other scholars have argued that women who become teen mothers often face poor socioeconomic prospects even before pregnancy (Edin and Kefalas 2005). Although the mechanisms are hard to disentangle, early childbearing is associated with lower maternal education and parental income. In addition, children of teenage mothers are at a higher risk of dropping out of high school, exhibiting behavioral problems, becoming teen parents, and being poor when they become adults (McLanahan 2009; Wu and Martinson 1993). Given the important implications of maternal age, Chapter 3 begins with an analysis of the racial and educational differences in age at first birth. Chapter 3 also analyzes the trends in fertility rates. Although fertility rates in the U.S. have stabilized slightly below replacement levels, around two children per woman, significant variation is observed across demographic groups. More children implies that families need to spread thin their economic resources. Early entrance into motherhood is also strongly associated with higher achieved fertility. Although some scholars have argued that the spread of contraceptive use since the mid-1960s has weakened the link between early births and subsequent higher fertility (Morgan and Rindfuss 1999), some studies have found that contraceptive use in low-income communities is still deficient (Edin and Kefalas 2005; Frost, Singh, and Finer 2007; Musick et al. 2009). I show that age at first birth is still an important determinant of achieved fertility among disadvantaged groups. After presenting over all trends, Chapter 3 provides a review of some of the major theories that attempt to explain racial inequality as they relate to 7 fertility behaviors, and evaluates whether the results of the multivariate analyses support the predictions of these theories. 1.3.2. Childlessness over Time Most research in fertility has focused on childbearing, and less attention has been paid to childlessness, which is also an important fertility outcome. I dedicate Chapter 4 to the study of childlessness, an alternative fertility trajectory that has become a more popular choice among the younger generations. In addition, as more women delay childbearing, an increasing proportion of them are approaching the end of their reproductive lives without having met their fertility intentions (Lundquist, Budig, and Curtis 2009; Quesnel- Vallée and Morgan 2003). Because infertility increases with age, further postponements of motherhood imply that a growing number of women will experience age-related infertility problems (McQuillan et al. 2003). For many women, childlessness is an involuntary outcome resulting from infecundity issues experienced by themselves or their partners, while for others, it is a voluntary choice (Abma and Martinez 2006; Heaton, Jacobson, and Holland 1999; McQuillan et al. 2003). Some scholars have argued that the growing emphasis in individualism and self-development, as well as the incompatibility of combining work and family, has reduced individual’s willingness to undertake long- term commitments, such as marriage and childbearing, making childlessness a more appealing option (Cherlin 2004; Hagestad and Call 2007; Lesthaeghe 1995). Although childbearing and childlessness are highly related, higher fertility rates at the macro-level do not always result in lower rates of childlessness. In Chapter 4, I explore how the associations between key socioeconomic and demographic factors and 8 the prevalence of childlessness have change from the 1980s to the 2000s. Comparing these results to those presented in Chapter 3, I demonstrate that the mechanisms that shape childbearing patterns differ from those driving childlessness. 1.3.3. Evaluating the Effects of Fertility on Women’s Employment Since the 1970s, women’s employment substantially increased, reducing the gender gap in labor force participation. Much of this growth was driven by the entrance of mothers to the labor market (Casper and Bianchi 2002; Sayer, Cohen, and Casper 2005). By 2013, 70% of women with children under the age of 18 participated in the labor force (U.S. Department of Labor 2014). Despite significant progress towards gender equality in the labor market, little progress has been made in reducing gender inequality within families. Women still perform a higher share of unpaid domestic labor, usually assuming the majority of childrearing and household chores, which limits their time for paid labor (Bianchi 2000; Sayer, Cohen, and Casper 2005). A vast literature demonstrates that having children is associated with significant penalties in labor market outcomes among women. These penalties are evidenced in reduced employment rates, fewer hours worked, lower earnings, and blocked career opportunities (Abendroth, Huffman, and Treas 2014; Budig and England 2001; Budig 2003). In Chapter 5, I evaluate the extent to which fertility reduces women’s employment, how its effects change over the life course, and how they vary across racial/ethnic groups using longitudinal data for women who have recently ended their reproductive lives. This research seeks to shed light on how fertility patterns affect the economic well-being of women and their children. 9 Chapter 6 summarizes the main findings of this dissertation, noting that although significant racial and class differences remain, the trends in age at first birth and fertility rates have stopped diverging, and the gaps in the prevalence of childlessness have been reduced. I explain the extent to which the findings provide support for the theories evaluated in each chapter. In Chapter 6, I highlight that although sociodemographic factors have become important determinants of childbearing behaviors, race still continues to shape women’s fertility and employment outcomes, particularly among low- educated groups. Next, I outline the contributions of this project to current research and recognize some of its limitations. I end Chapter 6 with a discussion of the major findings and the policy implications suggested by the results of this dissertation. 10 Chapter 2 Data 2.1. Introduction In this dissertation, I analyze data from two U.S. nationally representative designed surveys: (i) the National Survey of Family Growth (NSFG), a cross-sectional survey (i.e., data collected at different period of time), suitable for the analysis of fertility trends; and (ii) the National Longitudinal Survey of Youth, (NLSY79) 1979–2012, a longitudinal dataset (i.e., data collected over time from the same individuals), appropriate for the study of causal effects. I now describe both datasets, including the population samples, periods of data collection, and characteristics of the surveys. 2.2. National Survey of Family Growth (NSFG) The National Survey of Family Growth (NSFG) is a probability sample design conducted by the National Center for Health Statistics that collects data on fertility, including pregnancies, births, number of children ever born for women ages 15 to 44 living in the U.S. (National Center for Health Statistics 2015). The NSFG also collects data on transitions in union formation, socioeconomic background, demographic characteristics, and health related issues. The NSFG has been conducted repeatedly since 1973, using 11 similar questions on relevant variables related to fertility behaviors, thus, it is suitable for the analysis of fertility trends (Martinez, Daniels, and Chandra 2012). All survey waves until 2002 were collected during the year indicated in the survey, and were obtained from the Integrated Fertility Survey Series (IFSS), which contains harmonized data from all cycles of the NSFG from inception until 2002 (Smock et al. 2013). The IFSS is available online at http://www.icpsr.umich.edu/icpsrweb/IFSS/. The last two waves of the NSFG are continuous surveys conducted during consecutive years, the first one covering 2006 through 2010, and the last one from 2011 until 2015. As of the writing of this dissertation, the National Center for Health Statistics has only released preliminary data from 2011 to 2013 for the last survey, which I include in the analyses in this dissertation. The 2006–2010 and 2011–2013 waves were obtained directly from the National Center for Health Statistics website, and are available online at http://www.cdc.gov/nchs/nsfg.htm. Table 2.1 provides a brief overview of the population covered and the unweighted sample sizes for each of the NSFG survey waves. Table 2.1. NSFG: Unweighted sample size for the women aged 15–44. Year a Population Covered Sample Size 1973 Ever-married and single women age 15–44 with children 9,797 1976 Ever-married and single women age 15–44 with children 8,611 1982 All women age 15–44 7,969 1988 All women age 15–44 8,450 1995 All women age 15–44 10,847 2002 All women age 15–44 7,643 2006–2010 All women age 15–44 12,279 2011–2013 b All women age 15–44 5,601 began to be oversampled. b Preliminary data release for the 2011–2015 wave. a All these years include an oversample of Black women. In addition, since 1995 Hispanic women also 12 Since 1973, this survey oversampled Black women, and since 1995 it did the same for Hispanic women. Individuals were randomly selected using a screening process in which respondents reported their gender, age, race, and ethnicity, and then they were randomly chosen by a computerized process taking into consideration their representation, probability of selection, and probability of non response. In this study, race is measured by respondent’s self-identification in any of the racial categories including White, Black, and other. Ethnicity was measured by whether the respondent identified as being Hispanic or not. I used the answers to both questions to create mutually-exclusive racial/ ethnic categories including: non-Hispanic White, non-Hispanic Black, and Hispanic. In Chapter 3, I use data from all the survey waves from 1973 through 2011–2013 to analyze overall trends in maternal age at first birth and the number of children ever born, then I conduct multivariate analyses using data from the last two surveys, 2006– 2010 and 2011–2013. As Table 2.1 shows, never married women were included for the first time in the NSFG in cycle 3 conducted in 1982. Because a great proportion of women who have not had children by the end of their reproductive lives have never been married, I take 1982 as the baseline year for comparisons in Chapter 4, which focuses on the study of childlessness. The sample in Chapter 4 consists of mothers and childless women who are at the end of their reproductive span, between ages 40–44. Given this age restriction and to increase the sample size, I combine the 1982 and 1988 waves to assess the population of childless women during the 1980s, and compare the results with those using the 2006–2010 wave, which represents women who ended their reproductive years during the late 2000s. I also present results conducted using the preliminary data release for 2011–2013. 13 2.3. National Longitudinal Survey of Youth (NLSY79) The National Longitudinal Survey of Youth (NLSY79) is a nationally representative longitudinal survey that began in 1979 and collected data of 12,686 men and women living in the U.S. who were between the ages of 14 and 22. The NLSY79 has repeatedly interviewed the same individuals, following them over time since 1979. Data collection for the NLSY79 still continues. Respondents were interviewed annually through 1994, and biannually thereafter. The last available survey data, as the writing of this dissertation, was 2012. The NLSY has collected respondent’s information as they made educational decisions, moved out of their parents’ homes, entered the labor market, changed occupations, married, and started families. This sample represents the cohort born during the second half of the baby boom period. Because this dissertation focuses on fertility, I restrict the analyses to female respondents. By 2012, respondents were between the ages of 47 to 55, an age when most women have ended their reproductive lives. The NLSY79 oversamples economically disadvantaged Black and Hispanic populations. A military subsample and an oversample of a non-Black, non-Hispanic disadvantaged group were excluded from the study because they either became ineligible or were dropped from the survey after 1990. Excluding these groups, 4,926 women remain eligible for the study. One advantage of longitudinal datasets is that it provides detailed event history data on the life course experiences of the individuals surveyed. Particularly, the NLSY79 includes data on the timing of marital transitions, cohabitation, education, occupation and income for the respondent and his/her spouse, childbearing intentions, fertility history, rural/urban residence, region, among other variables. However, as with most longitudinal 14 datasets, one limitation is the problem of attrition, as individuals stop participating in the survey and, thus, the possibility that the remaining sample could be affected by selectivity. Nonetheless, the NLSY79 presents a relatively high retention rate. Until 2000, the retention rate exceeded 80%, and from 2002 until 2012, the retention rate was above 70% (Bureau of Labor Statistics 2015). In Chapter 5, I use fixed-effects models to analyze these data. Fixed-effects models only require that respondents have valid data for two of the periods of observation, thus, minimizing the loss of data due to non-interview or missing values. As in the NSFG, race/ethnicity in the NLSY79 is measured during the screening process. An important limitation of the NLSY79 is that the racial/ethnic categorization is based on the interviewer’s identification, which can be subjective. Another limitation is that this variable is coded as “Hispanic,” “Black,” and “non-Hispanic, non-Black.” Although the last category is composed primarily of Whites, it includes respondents with other racial/ethnic backgrounds. In Chapter 5, I use this disclosure and follow most published research using the NLSY79, taking the “non-Hispanic, non-Black” category as a proxy for Whites. An advantage of the NLSY79 is that the survey was designed to oversample Hispanics, African Americans, and disadvantaged population; thus, the sample sizes are large enough to conduct racial/ethnic comparative analysis. Table 2.2 provides a brief description of the unweighted sample of female respondents and their retention over time since the first year they were interviewed. As Table 2.2 shows, excluding the military sample and the discontinued disadvantaged non- Black/non-Hispanic oversample, the total sample size of female respondents was 4,926 in 1979 of which 3,767 were re-interviewed in 2012. The analytical sample used in Chapter 15 5 consist of 4,880 respondents, who satisfy the fixed-effects requirement of having at least two periods of observation with valid data. Table 2.2. NLSY97: Unweighted sample size for female respondents a Sample 1979 1984 1990 1994 2000 2002 2004 2006 2008 2010 2012 Total Females 4,926 4,722 4,490 4,466 4,102 3,945 3,973 3,905 3,965 3,885 3,767 Race/Ethnicity Nonblack/non-Hispanic 2,477 2,365 2,271 2,243 2,065 1,999 1,982 1,950 1,963 1,929 1,851 Black 405 393 365 363 343 328 326 320 323 323 317 Hispanic 226 217 198 203 185 178 177 174 172 171 163 Economically disadvantaged Black 1067 1034 984 987 912 879 901 883 912 886 868 Hispanic 751 713 672 670 597 561 587 578 595 576 568 a The analytic sample excludes military sample and the economically disadvantaged non-Black/non-Hispanic supplemental sample. The latter one was dropped after 1990. 16 Chapter 3 Beyond Race? The Homogenizing Effect of College Education on Fertility Behaviors 3.1. Introduction Fertility trends in the U.S. show persistent racial/ethnic differences in childbearing behaviors, disadvantaged ethnic minorities exhibit higher rates of early childbearing and achieved fertility than Whites (J. A. Martin et al. 2015; Musick et al. 2009). Demographers and family sociologists have noted that a bifurcating trend in the timing and marital context of childbearing by class and race/ethnicity has emerged, with higher proportions of early and nonmarital childbearing among the lower class and racial minorities (Ellwood and Jencks 2004; McLanahan 2009; Rindfuss, Morgan, and Offutt 1996). Given the increasing diversity in the racial/ethnic population composition and continuous changes in fertility behaviors, it is crucial to evaluate whether racial differences in the timing and level of fertility persist among younger cohorts analyzing recent data. In this chapter, I use data from the National Survey of Family Growth (NSFG) to evaluate trends in age at first birth and fertility rates from the 1970s until 2013. I conduct multivariate analysis using data from the two most recent waves, 2006–2010 and 2011– 2013, to investigate educational and racial disparities in fertility behaviors. Because childbearing influences women’s life chances, the analysis of racial differences in fertility outcomes is important because it can shed light on the mechanisms that contribute to 17 social inequality. Early and higher fertility can disrupt young mothers’ educational and occupational pathways (Rumbaut 2005). Fertility patterns also have important implications for family size and the resources available to adults and children within families, also affecting social inequality among the younger generation (McLanahan and Percheski 2008; McLanahan 2009; Western, Bloome, and Percheski 2008). The divergence in childbearing behaviors has been linked to the persistence of racial/ethnic disparities in socioeconomic outcomes (McLanahan and Percheski 2008; Western, Bloome, and Percheski 2008). In this chapter, I show that by the turn of the twenty first century significant differences still remain in age at first birth and fertility rates by major sociodemographic characteristics, but the differences have narrowed over time. I begin by exploring the trends in age at first birth and fertility rates from 1970s until 2013, then I review the different theories that have been used to account for racial inequality, next, I conduct multivariate analyses using data for the two most recent waves, from 2006 until 2013, evaluating whether the results support the predictions of the different theories. Finally, I conclude by summarizing the most important findings and discussing the implications of current racial/ethnic disparities in fertility outcomes for socioeconomic inequality. The results indicate that the postponements of motherhood among Whites and college educated women have come to a halt in recent years, nonetheless, significant racial and educational disparities remain. I find that most of the racial differences in age at first birth and total number of children are concentrated at low levels of education, with substantive homogeneity in fertility behaviors among college educated women. The results are consistent with the racial stratification theory that poses that racial inequality 18 can be explained by variations in socioeconomic and demographic characteristics, and implies that racial inequality would be more salient among the lower class because their limited resources make it hard to overcome the cumulative disadvantage of historical discrimination (Frank and Heuveline 2005; Telles and Ortiz 2008). This theory suggests that disadvantaged racial minorities are exposed to social contexts conductive to early and higher fertility. The findings imply that if cultural norms and values shape differently fertility behaviors across racial/ethnic groups, they seem to be overridden by attaining a college education. Based on these results, I argue that college education homogenizes fertility behaviors across racial/ethnic groups, emerging as an important tool for policy implementation. 3.2. Background Racial differences in fertility behaviors have captured the attention of demographers, sociologists, and policy makers. Official reports and prior studies consistently show significant differences in fertility outcomes by education and race/ethnicity, with high rates of early and nonmarital childbearing, and higher achieved fertility among low- educated women and ethnic minorities (J. A. Martin et al. 2015; Martinez, Daniels, and Chandra 2012; Musick et al. 2009; Yang and Morgan 2003). Understanding these differences is important because childbearing behaviors affect family composition and the resources available to family members, both of which can exacerbate racial inequality (McLanahan and Percheski 2008; McLanahan 2009; Western, Bloome, and Percheski 2008). Although the racial differences in the timing of entrance into motherhood and fertility rates have decreased in recent years, significant differences remain. 19 Figure 3.1 illustrates the trends in age at first birth and total number of children by race/ethnicity and education from 1973 until 2011–2013 for women aged 32–44 1 . As mentioned in Chapter 2, age, race, and ethnicity in the NSFG are self-reported by the respondents. I constructed mutually exclusive racial/ethnic categories based on the race and Ethnicity variables (see Measures section 3.7.2 for more details). For simplicity I will refer to non-Hispanic Whites as Whites, and non-Hispanic Blacks, as Blacks. Panel A in Figure 3.1 shows increasing disparities by race/ethnicity until 2002, followed by smaller, but persistent gaps. The greatest difference in age at first birth is observed in 2002, with an average of 25.4 for Whites and 21.7 for Blacks, resulting in a 3.7 years gap; while the mean age at first birth for Hispanics 2 fell in between at 23.0. In subsequent years, the racial gap slightly decreased to 3.3 years. The growing disparities until 2002 1 Ideally, total fertility is more accurately assessed among women who have ended their reproductive lives. The National Survey of Family Growth (NSFG) provides data for women aged 15–44, and although women aged 40-44 are the closest to have completed their fertility, using data for women aged 32–44 produce similar results with the added benefit of increased sample size. The decision to include in the analytical sample women aged 32–44 was based on the maximization of sample size, without substantially altering the main results were women aged 40–44 selected instead. 2 Before 1995, averages for native-born Hispanics include those of foreign-born Hispanics because small sample sizes to distinguish between the two. Beginning 1995 the NSFG began oversampling Hispanic women, resulting in larger sample sizes that allowed separating native-born from foreign-born Hispanics. 20 were driven by greater delays in childbearing among Whites, however, as Panel A shows, in recent years this postponement has come to a halt. Figure 3.1. Trends in age at first birth and number of children: U.S. 1973-2013, women aged 32-44. Notes: National Survey of Family Growth (NSFG), 1973 through 2011-2013. Author’s calculations. a Before 1995, averages for native-born Hispanics include those of foreign- born Hispanics given that small sample sizes do not allow to distinguish between the two groups. Panel B in Figure 3.1 shows even larger differences in age at first birth by education. The differences by education peaked in 1995 when the mean age at first birth reached 28.1 for college graduates compared with 20.0 for women who did not graduate from high school, a gap of 8.1 years. The educational gaps were largely driven by an increasing A. Age at first birth by race/ethnicity C. Number of children by race/ethnicity B. Age at first birth by education D. Number of children by education 18 20 22 24 26 28 30 1973 1976 1982 1988 1995 2002 2006-2010 2011-2013 Less than HS High school Some college College 1 2 3 4 1973 1976 1982 1988 1995 2002 2006-2010 2011-2013 Less than HS High school Some college College 1 2 3 4 1973 1976 1982 1988 1995 2002 2006-2010 2011-2013 White Black Native-born Hispanic Foreign-born Hispanic a 18 20 22 24 26 28 30 1973 1976 1982 1988 1995 2002 2006-2010 2011-2013 White Black Native-born Hispanic Foreign-born Hispanic a 21 postponement of motherhood among college graduates; nonetheless, the mean age at first birth for college graduates has stopped increasing. Among women who did not graduate from high school the mean age at first birth has remained relatively stable only slightly increasing since the early 2000, reaching 21.4 in 2011-2013. Panel C shows trends in the number of children ever born by race/ethnicity. The largest racial/ethnic differences are observed in the 1970s. In 1973, White women had on average 3.0 children, while Blacks and Hispanics had, respectively, 3.8 and 3.5 children on average. This is the end of the baby-boom period, which was followed by large decreases in fertility rates for all groups and a substantial convergence by race from the 1970s to until the mid-1990s. Since 1995, fertility rates have remained pretty much stable. It is worth noting that foreign-born Hispanic women’s fertility remains an outlier, however, many of these women began childbearing in their countries of origin. By contrast, native-born Hispanics show similar fertility behaviors to African Americans. In the 2011–2013 period, the most recent period with available data, racial/ethnic minority groups still averaged more children than Whites. The average number of children among Whites was 1.8, while it was 2.3 among Blacks and native-born Hispanics (differences with Whites significant at p < .01), and 2.7 children among foreign-born Hispanics (difference significant at p < .05). Panel D shows the trends in the number of children by education. In all periods, women with less than high school averaged more children than all other groups, showing declining fertility rates until 2002, when they average 2.6 children, followed by a slight increase. In 2011–2013 women who did not graduate from high school had on average 2.9 children; while college graduates averaged 1.5 children, a difference of 1.4 children. 22 Although we did not observe large differences in achieved fertility by race/ethnicity in Panel C, significant racial differences emerge when we compare racial groups by level of education, as shown later in section 3.8. In sum, racial differences in age at first birth increased from the 1980s through the 2000s, although the differences have slightly declined in recent years, they remain sizable. Conversely, the racial/ethnic differences in number of children are small, but still significant, and are concentrated at low levels of education. Scholars have proposed various theories to explain racial differences in fertility outcomes, as I review in the next section. 3.3. Theoretical Perspectives: Explaining Racial Inequality The structural perspective poses that racial/ethnic differences in fertility outcomes are explained by socioeconomic conditions, such as educational attainment, income, and family background characteristics, all of which shape young adults’ life chances, and consequently their reproductive behaviors (Musick et al. 2009; Schoen et al. 2009; Wilson and Neckerman 1987). This argument predicts that after controlling for demographic and socioeconomic factors, racial/ethnic variations in fertility behavior should disappear. Supporting this claim, some scholars have found that poverty and family instability result in high levels of stress and lack of parental supervision that explain early initiation of sexual activity and high rates of early childbearing among Blacks (Wilson and Neckerman 1987; Wu and Martinson 1993). Other scholars have argued that early childbearing and high fertility is common in low-income communities regardless of racial and ethnic background (Edin and Kefalas 2005). Nonetheless, most 23 empirical evidence indicates that socioeconomic factors are important determinants of fertility behaviors, however, these factors cannot fully explain all racial/ethnic differences, norms and values also play an important role. The cultural perspective attributes racial/ethnic fertility differences to variation in attitudes, norms, and pronatalist values that encourage family formation and prize larger families (Forste and Tienda 1996; Landale and Oropesa 2007; Parrado and Morgan 2008). Among Hispanics, familism – a set of cultural norms and beliefs that place high value on marriage and children – has been used to explain early childbearing and higher fertility (Choi 2014; Forste and Tienda 1996; Hartnett and Parrado 2012; Landale, Schoen, and Daniels 2010). Similarly, religiosity also shapes fertility behaviors, and racial minority groups tend to endorse pronatalist religious views. Religious groups tend to hold more traditional family attitudes that are associated with higher fertility (Hayford and Morgan 2008). The cultural perspective predicts that the racial differences in fertility behaviors would remain significant even after controlling for socioeconomic and demographic characteristics. It is worth noting that although racial/ethnic groups possess particular sets of norms and values, culture is a multi-dimensional concept that extends beyond racial differences in practices and norms (Bourdieu 1986; Lareau 2003). However, given limited data and difficulties in measuring culture, most demographic research uses race and ethnicity as proxies for culture (Choi 2014; Frank and Heuveline 2005; Lloyd 2006; Schoen et al. 2009; Yang and Morgan 2003), and when available, 24 measures of religiosity as well. Fortunately, the NSFG provides both measures, and this study takes advantage of these rich data 3 . Classical assimilation theory qualifies the cultural perspective stating that racial cultural norms become less salient over time and across the generations, as immigrants and their children adopt the practices and behaviors dominant in the host society (Alba and Nee 1997; Choi 2014; Parrado and Morgan 2008; Parrado 2011). According to this perspective, native-born immigrant groups, i.e. the second generation and above, would exhibit fertility behaviors that more closely resemble those in the social mainstream, which in the U.S. would be the White middle-class, whereas foreign-born groups will exhibit significantly higher rates of early childbearing and achieved fertility. Supporting the assimilation theory, some scholars have claimed that familism values decrease with time in the U.S. as new cohorts adopt the more individualistic ideals of Western societies, and have found evidence of assimilation over the generations among Hispanic groups (Choi 2014; Hartnett and Parrado 2012; Lloyd 2006; Parrado 2011; Telles and Ortiz 2008). However, the theory of segmented assimilation, proposed by Alejandro Portes and Min Zhou, challenges the unidirectional assimilation perspective, claiming that assimilation depends on immigrants’ human and social capital, context of reception, and place of settlement upon arrival (Portes and Zhou 1993). Accordingly, disadvantaged 3 The NSFG also includes attitudinal data, however, the large proportion of missing values poses significant challenges to statistical power. For this reason, attitudinal variables were not included in this study. 25 immigrants, such as Mexicans, Puerto Ricans, and Salvadorians, tend to settle in poor neighborhoods with higher proportions of people of color, and thus, may adopt the fertility habits of native disadvantaged minority groups rather than those of the dominant White middle-class. By contrast, immigrants with more resources and human capital, such as Cubans, who also experienced a favorable context of reception and often settled in middle-class neighborhoods, are expected to assimilate to the norms and practices pertaining to the middle-class, who are primarily White. Unfortunately, given the small sample size, I was unable to distinguish immigrant groups by country of origin. However, Mexicans and Puerto Ricans compose the majority of the Hispanic population in the U.S., therefore, the segmented assimilation perspective would predict that on average native- born Hispanics would exhibit similar fertility behaviors to those found among disadvantaged native groups, such as African Americans. Structural and cultural explanations are not mutually exclusive, in fact, the racial stratification perspective combines both arguments. The racial stratification perspective underscores exclusionary structural factors at the base of racial disparities, tracing the current socioeconomic situation of racial minority groups to their historical and cultural legacy. The U.S. is a racially stratified society. Historically, privileges and benefits, such as educational, residential, and occupational opportunities, have been distributed based on a racial hierarchy, creating shared racial/ethnic social contexts that shape reproductive behaviors (Forste and Tienda 1996). This perspective emphasizes the effects of historical discrimination, which is still perceptible in current times (Frank and Heuveline 2005; Parrado and Morgan 2008; Telles and Ortiz 2008). According to this argument, socioeconomic factors may not fully capture these historical legacies, and thus, it predicts 26 racial/ethnic inequality among low-income groups, which are more likely affected by the cumulative effect of historical disadvantage. Wilson and Neckerman (1987) have argued that in low-income African American communities, early and nonmarital fertility have become the norm, partly as a result of historical segregation and neighborhood socioeconomic conditions, such as high unemployment rates. Given few opportunities for meaningful adult roles, Black women in low-income communities bestow high value on children. Scholars have found that, in these communities, motherhood represents a source of meaning, a symbol of adulthood and an attestation of womanhood (Edin and Kefalas 2005; Forste and Tienda 1996; Wilson and Neckerman 1987). Prior studies have found support for the racial stratification perspective, indicating the importance of social contexts and patterns of racial exclusion in shaping fertility outcomes (Choi 2014; Frank and Heuveline 2005). A final explanation proposed by Goldscheider and Uhlenberg (1969), the minority group status hypothesis, predicts higher delays in entrance into motherhood and lower fertility among upwardly mobile minorities, claiming that individuals from racial/ethnic minorities may limit their fertility more than Whites in an effort to overcome barriers for upward social mobility (Forste and Tienda 1996; Goldscheider and Uhlenberg 1969; Yang and Morgan 2003). This argument implies that racial differences in childbearing outcomes will be small and may even reverse among the college educated. 27 3.4. Competing Hypotheses In this Chapter, I do not test directly these explanations, however, I evaluate whether the results are consistent with the predictions of each of these theories. Table 3.1 summarizes the competing hypotheses derived from these theories as they relate to age at first birth and fertility rates across the four racial/ethnic groups analyzed in this study, namely, Whites, Blacks, native-born Hispanics, and foreign-born Hispanics. Table 3.1. Competing hypotheses: Theoretical predictions Theoretical Perspective Hypotheses Structural perspective Predicts that no significant racial/ethnic differences in fertility behaviors would remain after socioeconomic and demographic factors are taken into account. Cultural perspective African Americans and Hispanics, in particular foreign- born Hispanics, will exhibit earlier entrance into motherhood and higher fertility levels than Whites, even after controlling for socioeconomic and demographic characteristics. Classic assimilation theory Predicts earlier entrance into motherhood and higher fertility among foreign-born Hispanics; however, the fertility behaviors of native-born Hispanics will be more similar to those of Whites. Segmented assimilation theory Similar to the classic assimilation theory, with the difference that it predicts that the fertility behaviors of native-born Hispanics, most of whom settle in disadvan- taged communities, will be closer to those of Blacks. Racial stratification theory Socioeconomic factors should reduce some of racial/ethnic disparities in fertility outcomes, but significant differences would remain, especially among disadvantaged populations, who may have fewer tools to overcome the legacy of historical discrimination. This theory predicts more pronounced racial/ethnic differences in fertility behaviors among the lower class. Minority group status hypothesis Low educated Hispanics and Blacks are predicted to enter motherhood at younger ages and have more children than Whites. However, it predicts that college educated (a proxy for upward social mobility) native-born Hispanics and Blacks would exhibit higher age at first birth and lower fertility than college educated Whites. 28 3.5. Factors Affecting Fertility Behaviors In addition to race/ethnicity, socioeconomic status also shapes reproductive behavior. Socioeconomic status is typically assessed by education, income, and occupation. In this study, I use education and income as proxies for socioeconomic status. I do not use occupation because I restrict my sample to women, and oftentimes women withdraw from the labor market when they have children, thus their occupation may not adequately capture their socioeconomic status. Education is one of the most important indicators of socioeconomic status, and a major predictor of fertility. As the trends presented in section 3.2 illustrates, education is negatively associated with fertility. It is worth noting that education not only signifies socioeconomic status, but also social class. Social class is a broader concept that includes socioeconomic status, and also the taste, norms, and values that drive behaviors (Bourdieu 1984). In this chapter, I pay special attention to the role of college education. In the U.S, high school institutions are highly segregated by race and socioeconomic status, offering few opportunities for inter-ethnic and inter-class interaction, thus reinforcing racial norms and practices (Hall 1992; Kalmijn and Kraaykamp 1996; Rumbaut 2005; Wilson and Neckerman 1987). However, young adults often cross geographical boundaries to attend college; thus, college institutions open opportunities for shared contexts and inter-racial and inter-class interaction, reducing ethnic differences in behavioral norms. Montgomery and Casterline (1996) argued that educational institutions are important sources of social influence that spread norms a values regarding reproductive behaviors. College goers tend to limit their fertility until education completion, and later, while they get established in the labor market. Consequently, 29 among college graduates, delayed and lower fertility is normative (Musick et al. 2009). Individuals from different ethnic backgrounds who attend college are exposed to social contexts in which low fertility is normative, thus, in addition to the knowledge acquired during years of education, peer pressure and social expectations among college goers can also homogenize reproductive behaviors (Montgomery and Casterline 1996). Because college education occurs during early adulthood, which coincides with the prime years of childbearing, I argue that college institutions exert a strong influence in fertility outcomes. In addition, because colleges are important socialization milieux for historically disadvantaged racial minorities, whose family backgrounds and often segregated neighborhoods would otherwise preclude them from the acquiring reproductive behaviors characteristic of the middle-class, I claim that college education has the potential to reduce disparities in fertility behaviors across racial groups. Income is another important socioeconomic indicator that influences fertility outcomes. Income measures the level of economic resources available to individuals. Since the 1970s economic inequality increased greatly driven by the unequal returns to education. While wages for the highly educated substantially increased, wages at the bottom remained stagnant (Autor, Katz, and Kearney 2008). This growing earnings inequality has increased the economic opportunity cost of childbearing for women with greater earnings potential. However, at the same time, higher income implies that couples are more able to afford childcare services and to outsource domestic labor, both of which may reduce the opportunity cost of having children for better-off women. Moreover, the growth in educational homogamy suggests that women with higher earnings potential are also more likely to marry high earning partners. These offsetting forces, the increase in 30 opportunity costs coupled with greater resources, make the relationship between income and fertility unpredictable. In this Chapter, I evaluate the extent to which income is associated with the timing into entrance into motherhood and achieved fertility. Marital and partnership status is another important sociodemographic determinant of childbearing behaviors. Americans regard marriage as an ideal institution for reproduction, and childbearing is still expected to occur within marriage (Cherlin 2004, 2010; Furstenberg 1996). However, in recent decades, marriage has become decoupled from childbearing as nonmarital births increased from 5% in 1960 to 41% in 2013 (Ellwood and Jencks 2004; Hayford, Guzzo, and Smock 2014; J. A. Martin et al. 2015). Nonmarital cohabitation has gained acceptance as a family setting for having children, although it has not achieved the status of marriage (Guzzo and Hayford 2014). Because marriage rates vary greatly by race/ethnicity (Hayford, Guzzo, and Smock 2014; Musick 2007), the multivariate analyses in this study pays special attention to marital and partnership status, distinguishing between married, cohabiting, and never-married unpartnered women. 3.6. Contribution to the Literature I contribute to the literature in several ways. First, I analyze racial differences in two important fertility behaviors, age at first birth, and number of children ever born, using nationally representative data that extend until 2013. Using recent data is important given evolving fertility behaviors and increasing diversity in the racial/ethnic population composition. Second, I expand the traditional White/Black comparison by including a sample of Hispanics, disaggregated by nativity. Most prior studies have limited their 31 samples to Whites and Blacks (England, Wu, and Shafer 2013; S. P. Martin 2000; Musick et al. 2009; Schoen et al. 2009; Yang and Morgan 2003), or have analyzed the fertility behaviors of young adult women who are in their mid-20s, many of whom have not yet begun childbearing (Schoen et al. 2009). Hispanics have become the largest minority group in the U.S., thus analyzing their childbearing behaviors provides a more holistic view of fertility patterns at a national level. Acknowledging this need, some studies and fertility reports include statistics for Hispanics; however, most of these studies do not distinguish between native-born and immigrant Hispanics (J. A. Martin et al. 2013; Martinez, Daniels, and Chandra 2012). Because the fertility behaviors of immigrant women are influenced by the social contexts in their countries of origin and the immigration process (Choi 2014; Frank and Heuveline 2005; Parrado 2011), it is crucial to distinguish between native-born and foreign-born Hispanics. Third, this study analyzes differentials by race fully interacted with education. In a prior study, Yang and Morgan (2003) used this technique comparing Blacks and Whites with less than 13 years of formal education with those with 13 years or more; however, as most women now obtain higher education, heterogeneity among educated women has increased (Musick, Brand, and Davis 2012); thus, it becomes crucial to differentiate women who obtain some college education from women who complete college degrees. Musick et al. (2009) included four educational categories, however, their sample was limited to Whites and Blacks. In this study, I use four educational groups – those with less than high school, high school, some college, and college education. In addition, while most prior studies control for maternal age, little attention has been given to the extent to which racial differences in fertility levels are due to variations in age at first birth. I compare the 32 results before and after controlling for age at first birth, and I show that in some cases the timing of entrance into motherhood accounts for part of the racial differences in achieved fertility. Finally, whereas most prior studies adjudicate only among a limited number of theories, I evaluate whether the evidence supports each of the six theories outlined in sections 3.3 and 3.4. 3.7. Data, Measures, and Methods 3.7.1. Data I use data from the National Survey of Family Growth (NSFG) from1973 through 2011– 2013 to analyze overall trends in fertility. This dataset was described in Chapter 2. Then, I conduct a multivariate analysis using data from the last two surveys, 2006–2010 and the preliminary data release for 2011–2013, available online at http://www.cdc.gov/nchs/nsfg.htm. For this analysis, I restrict my sample to women ages 32–44, most of whom, 83%, have already entered motherhood. Although women aged 40–44 are the closest to have completed their fertility, the decision to include women aged 32–44 in the analytical sample was based on preliminary analyses to maximize the sample size and use a sample of women that more closely resemble the fertility of women with completed fertility, without substantially altering the main results. I conducted preliminary analyses using samples of women aged 25–44, 30–44, 32–44, 35–44, and 40–44, and the group 32–44 was the one that more closely met the criteria. I include in the sample Whites, African Americans, and Hispanics. Since 1995 the NSFG began over sampling Hispanic women; the increase in sample size of Hispanic allowed me to separate native-born from foreign-born Hispanics. However, the trends 33 shown before 1995 present the results of native-born and foreign-born Hispanic women combined. A small number of respondents in the other race category or with missing race information were excluded from the sample. I also excluded mothers with no information on the timing of their first birth, those who reported having had their first child before age 15, and cases with missing values in some of the main covariates. The final sample size is 4,356 women for the 2006–2010 survey and 1,976 women for the preliminary 2011–2013 data release, resulting in a combined sample size of 6,332 women aged 32–44, of whom 3,525 are Whites, 1,310 are Blacks, 611 are native-born Hispanics, and 886 are foreign- born Hispanics. 3.7.2. Measures The two dependent variables are age at first birth, a continuous variable measured in years and months, i.e. 24.5 equals 24 years and 6 months; and number of children ever born, a discrete variable. The main independent variable is race/ethnicity, measured with a set of mutually exclusive dummy variables for non-Hispanic Whites, whom for simplicity I refer to as Whites, non-Hispanic African Americans, or Blacks, native-born Hispanics, and foreign-born or immigrant Hispanics. For all the analyses Whites are used as the referent category for three main reasons, first because Whites compose the group with the largest sample size, facilitating statistical comparisons; second, because making Whites the referent category allows the test of the assimilation and racial stratification theories, both of which consider Whites as the dominant group, the one that can more easily prescribe childbearing behaviors as desirable and normative; and third, because 34 prior studies have customarily used Whites as the referent group, thus this strategy facilities comparisons with prior research. Education is measured in four categories for those with less than high school (<12 years of education), high school (12 years), some college (13–15 years), and college education (16 or more years of education). Because childbearing behaviors depend on marital and partnership status (Hayford, Guzzo, and Smock 2014; Musick 2007), all models control for this variable. When the dependent variable is age at first birth, marital status is measured at the time of the birth, or at the time of the survey if the first birth is not observed (censored cases), and it is coded in four categories: married (referent category), single never married, cohabiting, and previously married. The latter includes divorced, separated, and widowed women. When the dependent variable is the number of children ever born, marital and union status is a cumulative measure coded in three categories, ever married (referent); never-married never-cohabited women, referred to as single for simplicity; and ever-cohabited but never-married women. Other structural measures, include family income, recoded using interval mid- points, adjusted for inflation using official Consumer Price Index from the U.S. Bureau of Labor Statistics taking 2013 as the base year, and divided in quartiles using weights for the entire original sample. The first quartile is the lowest income category (referent), and the fourth is the highest. I also control for family of origin characteristics including a dummy variable that takes the value of 1 if the respondent’s mother worked full-time when the respondent was aged 5–15, and 0 otherwise. Family structure is measured with a dummy variable that takes the value of 1 if the respondent’s father was absent when she was age 14, and 0 otherwise. 35 To measure parental education, I use the respondent’s mother’s education, which is also coded in four categories, less than high school or no mother figure (referent), high school, some college, and college education. Unfortunately, 2011–2013 data does not include a measure of the respondent’s father’s education. However, given the relatively high levels of educational homogamy among couples, father’s education tends to be highly correlated with mother’s education (Schwartz and Mare 2012). The respondent’s level of religiosity is measured with two dummy variables, the first one takes the value of 1 for those reporting that religion is very important in their lives, and 0 otherwise; and the second one measures attendance at religious services, taking the value of 1 for those who never attend religious services, and 0 otherwise. I also control for other demographic factors including rural residence, measured with a variable that takes the value of 1 if the respondent lived in a rural area, and 0 otherwise. In the analyses modeling number of children, a dummy variable is introduced to control for intendedness of the first birth, taking the value of 1 if the first birth was unintended, which includes unwanted births and births mistimed by two or more years before desired. Women’s age and age squared are introduced as a continuous variables to adjust for non-linearities in the effect of age. Because I pooled together data from two surveys, all models control for period effects with a dummy variable indicating whether the observation comes from the latest 2011-2013 period. 3.7.3. Methods First, I present bivariate descriptive statistics for the age at first birth and number of children by education and race/ethnicity for the NSFG 2006–2010 and 2011–2013 pooled 36 sample. Then, I conduct multivariate analyses using models fully interacted with education. The analysis of age at first births uses a semi-parametric discrete time Cox model. Cox regression is a duration model that estimates the average risk, also called hazard, of having a child conditional on not having had a child before, taking into account the length of exposure to the risk. Risk time is expressed in months, and all women are considered at risk of becoming mothers from age 15 (no births are observed before age 15 in this sample) until they either have a child or the age at interview if they remain childless (censored cases). For this analysis, I present two models; Model 1 adjusts for marital status and survey period, and Model 2 adds controls for family income, the respondent’s mother’s education and work status, family structure, religiosity, and rural residency. The results are presented in hazard ratios or exponentiated coefficients, which are analogous to odds ratios from logistic models. I use Poisson regression for the analysis predicting the number of children ever born, using the exposure option in STATA to adjust for the length of exposure to reproductive years since age 15. The results are provided in incidence rate ratios (IRR), which is the rate at which women of one group bear children relative to the referent group, holding constant other variables in the model. Poisson regression is suitable for modeling small counts, such as number of children; however, it assumes that the distribution of zeros (i.e. women having no children) and positive counts (women having one or more children) are generated by the same process, which often is not the case (Long and Freese 2006). Although most of the women in the sample have already entered motherhood (83%), the proportion remaining childless vary by race/ethnicity. To overcome this obstacle, I replicate the Poisson model presented in Model 1 using a zero- 37 truncated or conditional Poisson model (ZTP) in Model 2, and then adding on control variables in Models 3 and 4. The ZTP model predicts the number of children conditional on having entered motherhood. Thus, whereas Model 1 includes all women in the sample, Models 2 through 4 include mothers only, given that having had a child is a condition for the ZTP model. This strategy that overcomes the potential influence of different rates of childlessness across ethnic groups on achieved fertility. Although Poisson models have been widely used in fertility analysis (Brand and Davis 2011; Choi 2014; Telles and Ortiz 2008), ZTP models have not been previously applied. However, comparing the results from the Poisson to the ZTP model can reveal cases in which racial differences in achieved fertility are due to different probabilities in overcoming the hurdle of being childless rather than to having more children. In this analysis Models 1 and 2 control for marital status, women’s age, and survey period. Model 3 introduces a control for women’s age at first birth; this model adjust for exposure of time from the age at first birth until the age at the time of the survey. Finally, Model 4 adds controls for intendedness of the first birth, family income, family background characteristics, religiosity, and rural residency. All analyses are weighted to account for differences in the probability of being selected in the sample. I use transformed weights that average one to avoid artificially inflating the standard errors, thus, reducing the probability of committing type II error (the failure to identify a significant difference). 38 3.8. Results 3.8.1. Descriptive Results Table 3.2 presents weighted descriptive statistics for the pooled 2006–2010 and 2011– 2013 NSFG analytic sample by educational group. The average age at first birth was 24.7 for the entire sample, varying from 21.1 for women who did not graduate from high school to 28.8 for college graduates, a range of 7.7 years. By race, the average age at first birth ranged only over 3.3 years, with a mean of 25.7 for Whites compared with 22.4 and 22.5 for African Americans and native-born Hispanics, respectively. Disaggregating the mean age at first birth of racial groups by education reveals a smaller difference, indicating that part of the racial differences is due to variations in education. Among women who did not graduate from high school the mean age at first birth ranged only over 2 years, from 19.9 for native-born Hispanics to 21.9 for immigrant Hispanics, with Blacks and Whites falling in between with means of 20.0 and 20.9, respectively. Among high school graduates, Blacks and native-born Hispanics exhibited the earliest entrance into motherhood with a mean age at first birth of 21.2; whereas Whites presented the latest entrance with a mean age of 23.4. Age at first birth increases with education across all racial/ethnic groups. Among college graduates, the range by race/ethnicity was only 1.5 years, with a mean of 29.0 for Whites, and approximately 27.6 for all other racial minorities. The average number of children for the pooled sample was 2.0 per women, which is close to the official reports of the total fertility rate in the U.S. (J. A. Martin et al. 2015). However, as Table 3.2 shows, the average ranged from 2.9 for women who did not graduate from high school to 1.6 for college graduates, with larger racial/ethnic 39 differences among the low educated. Figure 3.2 illustrates the differences in the average number of children by race/ethnicity and education for the preliminary data release corresponding to the recent 2011–2013 period. Among women who did not complete high school, Whites had 2.4 children on average, while Blacks had 3.0, native-born Hispanics 3.4, and foreign-born Hispanics 3.2. It is worth noting that these differences are not very large, and the differences get smaller as education increases. Among high school graduates the average number of children ranged from 2.0 for Whites to 2.7 for foreign-born Hispanics. Among women with some college education no significant racial differences are observed. Surprisingly, among college graduates, Blacks and native-born Hispanics in this sample exhibited instead lower fertility than Whites, having only 1.3 Table 3.2. Descriptive weighted statistics by education: Women aged 32 to 44. National Survey of Family Growth (NSFG) 2006–2010 and 2011–2013. Unweighted N 6,332 1,042 1,626 1,831 1,833 Weighted % 100% 13.97 25.89 28.05 32.09 % Mothers 83.2 93.0 88.4 84.9 72.9 Mean age at first birth All women 24.7 21.1 22.7 24.4 28.8 Non-Hispanic white 25.7 (ref.) 20.9 (ref.) 23.4 (ref.) 25.1 (ref.) 29.0 (ref.) Non-Hispanic black 22.4 ** 20.0 * 21.2 * 22.4 * 27.6 * Native-born Hispanic 22.5 ** 19.9 21.2 * 23.6 ** 27.7 Foreign-born Hispanic 22.9 ** 21.9 * 22.8 23.6 † 27.6 ** Mean number of children ever born All women 2.0 2.9 2.2 2.0 1.6 Non-Hispanic white 1.8 (ref.) 2.4 (ref.) 2.0 (ref.) 1.9 (ref.) 1.6 (ref.) Non-Hispanic black 2.2 ** 3.0 2.6 ** 2.1 * 1.3 * Native-born Hispanic 2.4 * 3.4 ** 2.5 † 2.2 1.3 ** Foreign-born Hispanic 2.8 *** 3.2 * 2.7 * 2.1 2.0 Married or previously 2.2 (ref.) 2.9 (ref.) 2.3 (ref.) 2.1 (ref.) 1.8 (ref.) Ever cohabited, never married 1.8 * 3.0 2.1 1.3 ** 0.6 ** Single, never married 0.7 *** 1.8 1.2 ** 0.6 ** 0.1 *** †p < .10, *p < .05, **p < .01, ***p < .001 All women Less than high school High school Some college College 40 children on average compared with 1.6 observed for Whites, which at first sight, is consistent with the minority group status hypothesis predicting lower fertility for upwardly mobile disadvantaged minorities (Goldscheider and Uhlenberg 1969). Moreover, among college graduates, immigrant Hispanics stand out as an outlier, with a mean of 2.0 children per women, in line with the assimilation theory, which predicts higher fertility for Hispanic immigrants (Alba and Nee 1997). However, these observed racial differences can be the result of differences in characteristics favoring higher fertility, thus in the next section, I adjust for these characteristics using multivariate analysis. Figure 3.2. Number of children ever born: U.S. 2011–2013, women aged 32–44. Source: National Survey of Family Growth (NSFG), 2006–2010 and 2011–2013, author’s calculations. 3.8.2. Multivariate Results 3.8.2.1. Racial/Ethnic Differences in the Timing of First Birth In this section, I present the results for analyses of racial/ethnic differences in age at first birth. Table 3.3 shows the hazard ratios from discrete-time Cox regressions modeling the 2.4 2.0 1.9 1.6 3.0 2.6 2.1 1.3 3.4 2.5 2.2 1.3 3.2 2.7 2.1 2.0 1.0 1.5 2.0 2.5 3.0 3.5 <High school High school Some college College White Black Native-born Hispanic Foreign-born Hispanic 41 Table 3.3. Discrete-time Cox regressions on the risk of becoming a mother versus remanining childless by education. Hazard ratios and standard errors (in parenthesis). NSFG 2006–2010 and 2011–1013: Women aged 32–44. Race/Ethnicity Non-Hispanic white (ref.) --- --- --- --- --- --- --- --- Non-Hispanic black 1.21 1.26 1.26 1.11 1.26 * 1.16 † 1.15 0.96 (0.17) (0.18) (0.14) (0.14) (0.07) (0.07) (0.18) (0.15) Native-born Hispanic 1.45 1.48 1.69 * 1.59 * 1.25 ** 1.19 *** 0.83 * 0.75 † (0.26) (0.28) (0.20) (0.19) (0.04) (0.01) (0.04) (0.09) Foreign-born Hispanic 1.15 1.25 1.30 1.12 1.32 1.27 1.52 † 1.32 (0.15) (0.18) (0.16) (0.13) (0.17) (0.23) (0.26) (0.30) Marital status a Married (ref.) --- --- --- --- --- --- --- --- Single, never married 1.30 ** 1.25 * 1.51 * 1.39 † 1.01 1.04 0.32 ** 0.32 ** (0.03) (0.06) (0.15) (0.17) (0.09) (0.07) (0.05) (0.06) Cohabiting 1.18 1.20 1.18 † 1.17 1.40 1.45 † 1.03 1.06 (0.11) (0.15) (0.07) (0.11) (0.22) (0.20) (0.22) (0.24) Previously married 0.34 † 0.31 † 0.22 ** 0.20 ** 0.37 ** 0.37 * 0.12 ** 0.12 ** (0.15) (0.14) (0.05) (0.04) (0.06) (0.07) (0.02) (0.02) Family income 1st Quartile (lowest, ref.) --- --- --- --- 2nd Quartile 0.89 1.07 1.10 0.67 (0.09) (0.13) (0.10) (0.13) 3rd Quartile 0.90 1.05 0.97 0.60 (0.13) (0.04) (0.10) (0.14) 4th Quartile (highest) 0.73 * 0.79 1.06 0.63 (0.07) (0.10) (0.10) (0.16) Respondant mother's education Less than high school (ref.) --- --- --- --- High school 1.04 0.77 *** 0.94 0.96 (0.10) (0.01) (0.04) (0.20) Some college 1.41 0.90 0.98 1.02 (0.36) (0.06) (0.07) (0.21) College 1.10 0.75 0.79 0.77 (0.11) (0.15) (0.12) (0.13) Family of origin characteristics R's mother worked when R was 5–15 1.14 1.09 1.07 1.02 (0.11) (0.05) (0.08) (0.04) R's father absent at age 14 0.97 1.23 * 1.12 1.03 (0.12) (0.06) (0.08) (0.09) Religiosity Religion very important 1.05 1.13 1.27 * 1.47 * (0.14) (0.06) (0.07) (0.11) Never attends relig. serv. 1.10 1.01 0.97 0.89 † (0.07) (0.15) (0.11) (0.04) Rural residency 1.24 * 1.19 1.31 * 1.00 (0.07) (0.17) (0.10) (0.12) Period 2011–2013 0.99 0.98 1.16 1.14 1.13 1.14 † 0.97 0.99 (0.13) (0.12) (0.10) (0.09) (0.10) (0.06) (0.03) (0.04) N 1042 1042 1626 1626 1831 1831 1833 1833 †p < .10, *p < .05, **p < .01, ***p < .001 a Marital status at first birth or at time of survey if censored. Model 1 Model 2 Less than high school High school Some college College Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 42 timing of first births, the standard errors are shown in parentheses below the hazard ratios. Hazard ratios equal to one indicate no difference between the group observed and the reference group; the farther the hazard ratio value falls away from one, the greater the difference in age at first birth between the groups being compared. Model 1 controls for marital status and survey year, and Model 2 adds controls for structural and cultural factors. In preliminary analyses, I introduced control variables by groups at a time, however, the final results did not substantially differ from the full models presented in Model 2. The results in Model 1 of Table 3.3 show that most of the racial differences in age at first birth among native-born racial/ethnic groups are concentrated at low levels of education, net of union status and period effects. In line with pronatalist cultural arguments and racial disparities in cultural capital, these hazard ratios indicate that less educated Blacks and native-born Hispanics enter motherhood at younger ages than Whites; however, due to large variances, some of those differences do not reach statistical significance despite the large coefficients. For example, the differences between Whites and Blacks are significant only for women with some college education; in Model 1 Black women exhibit a 26% ([1.26-1] x 100% = 26%) greater risk or hazard of entering motherhood than Whites, net of other factors. However, controlling for structural and cultural factors in Model 2, 4 including, family income, respondent’s 4 In preliminary analysis, I introduced structural and cultural factors in the model separately, however, this strategy did not substantially changed the racial differences, thus for simplicity, I only show the full model containing all the control variables. 43 mother’s work status, having grown up without a father, respondent’s mother’s education, religiosity, and rural residency, reduces the hazard ratio to 16%, which is marginally significant at p < .10. Similarly, native-born Hispanics with high school degrees and some college education exhibit respectively 69% and 25% greater hazards of having a child compared with Whites in Model 1, but these differences are only slightly reduced in Model 2. By contrast, among college graduates, the difference between Whites and Blacks is small and not significant; remarkably native-born Hispanics exhibit instead a 25% ([0.75-1] x 100% = -25%) lower risk than Whites of having entered motherhood according to Model 2. This difference is also significant at p <.10, providing some evidence for the minority group status hypothesis, predicting greater delays in fertility among upwardly mobile minorities. Immigrant Hispanics behave differently, as predicted by the assimilation theory. The difference between Whites and immigrant Hispanics in the risk of entering motherhood increases, rather than decreases, with education, although the difference is significant only among those with college education (p <.10). According to Model 1, college educated immigrant Hispanics exhibit a 52% greater risk of entering motherhood than Whites, a difference that is reduced to 32% in Model 2, but no longer reaches statistical significance due to a large variance. Nonetheless, the results suggest that the childbearing patterns of immigrant Hispanic women are different from those of native- born groups. This finding could reflect the fact that many immigrant women complete their education and begin childbearing in their countries of origin (Calvo and Sarkisian 2014), and thus may not be exposed to the same childbearing norms that college education offer to native-born groups. Consistent with prior research, the reproductive 44 behaviors of immigrant women seem to be affected by different mechanisms than those shaping childbearing among native-born groups (Choi 2014; Forste and Tienda 1996; Frank and Heuveline 2005; Landale, Schoen, and Daniels 2010). 3.8.2.2. Racial/Ethnic Differences in Number of Children Table 3.4 shows the multivariate results predicting number of children ever born. The results are shown in incidence-rate ratios (IRR), the standard errors are shown in parentheses. All models adjust for time of exposure to reproductive years. Overall, the results show an educational gradient in racial/ethnic differences, with the greatest differences in the number of children found among low educated women, decreasing at higher levels of education. I start by discussing the first part of Table 3.4, which presents the results for those without high school education and high school graduates. As the results from the Poisson regression in Model 1 shows, among women who did not graduate from high school, Blacks had 25% more children than Whites, native-born Hispanics 32% more, and immigrant Hispanics 26% more children, net of marital status, women’s age, and survey period. A similar pattern is observed for high school graduates. As mentioned in the methods section, Model 2 reproduces Model 1, using instead a zero- truncated Poisson (ZTP) or conditional model, that is, it predicts the number of children conditional on having entered motherhood, and thus, restricts the analysis to mothers only. Interestingly, restricting the sample to mothers in Model 2 reveals slightly higher racial differences (and this is also true for women with some college education, shown in the second panel of Table 3.4), indicating that racial minorities who become mothers end up having slightly more children than Whites than previously indicated by Model 1. 45 Table 3.4. Poisson and conditional Zero Truncated Poisson (ZTP) models on children ever born for women aged 32–44. Incidence rate ratios (IRR) and standard errors (in parenthesis). NSFG 2006–2010 and 2011–1013. Poisson Model ZTP Model 2 ZTP Model 3 ZTP Model 4 Poisson Model 1 ZTP Model 2 ZTP Model 3 ZTP Model 4 IRR y≥0 IRR y>0 IRR y>0 IRR y>0 IRR y≥0 IRR y>0 IRR y>0 IRR y>0 Race/Ethnicity Non-Hispanic white (ref.) --- --- --- --- --- --- --- --- Non-Hispanic black 1.25 ** 1.30 ** 1.23 * 1.20 * 1.33 *** 1.44 *** 1.28 ** 1.21 * (0.09) (0.11) (0.10) (0.10) (0.08) (0.10) (0.10) (0.09) Native-born Hispanic 1.32 ** 1.41 *** 1.33 ** 1.32 ** 1.18 ** 1.20 * 1.10 1.08 (0.12) (0.14) (0.13) (0.13) (0.07) (0.10) (0.09) (0.09) Foreign-born Hispanic 1.26 *** 1.31 *** 1.38 *** 1.34 *** 1.31 *** 1.37 *** 1.34 *** 1.31 ** (0.07) (0.09) (0.10) (0.11) (0.08) (0.11) (0.09) (0.11) Marital status a Ever married (ref.) --- --- --- --- --- --- --- --- Ever cohabited, never married 1.04 1.05 1.08 1.09 0.88 * 0.85 † 0.87 † 0.84 * (0.07) (0.08) (0.07) (0.07) (0.06) (0.07) (0.07) (0.07) Single, never married 0.78 † 0.79 0.91 0.93 0.60 *** 0.64 * 0.66 * 0.65 * (0.11) (0.14) (0.15) (0.15) (0.08) (0.13) (0.13) (0.14) Age at first birth 1.02 † 1.02 1.01 1.01 (0.01) (0.01) (0.01) (0.01) First birth unintended 1.03 1.01 (0.08) (0.08) Respondant mother's education Less than high school (ref.) --- --- High school 1.05 0.95 (0.08) (0.06) Some college 1.00 1.12 (0.09) (0.09) College 0.99 1.17 (0.13) (0.12) Family of origin characteristics R's mother worked when R was 5–15 0.98 1.00 (0.05) (0.05) R's father absent at age 14 0.91 1.11 (0.06) (0.07) Religiosity Religion very important 1.08 1.02 (0.06) (0.06) Never attends relig. services 0.95 0.95 (0.06) (0.06) Family income 1st Quartile (lowest, ref.) --- --- 2nd Quartile 0.98 0.95 (0.06) (0.07) 3rd Quartile 0.88 † 0.87 * (0.07) (0.06) 4th Quartile (highest) 1.02 0.91 (0.10) (0.08) Rural residency 0.98 0.96 (0.07) (0.06) Survey year 2011–2013 0.91 * 0.92 0.91 † 0.90 † 0.96 1.07 0.97 0.94 (0.05) (0.05) (0.05) (0.05) (0.05) (0.06) (0.05) (0.05) Age 1.08 1.14 1.03 1.03 0.94 0.95 0.85 0.86 (0.14) (0.18) (0.16) (0.16) (0.13) (0.18) (0.15) (0.14) Age squared 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Constant 0.05 0.02 0.14 0.14 0.73 0.52 6.63 5.18 (0.13) (0.05) (0.40) (0.38) (1.91) (1.82) (21.44) (16.00) N 968 955 955 955 1451 1378 1378 1378 Log likelihood -1695.7 -1621.4 -1579.6 -1572.6 -2781.0 -2487.7 -2410.1 -2396.9 Less than high school High school 46 Table 3.4. Continued… Poisson Model 1 ZTP Model 2 ZTP Model 3 ZTP Model 4 Poisson Model 1 ZTP Model 2 ZTP Model 3 ZTP Model 4 IRR y≥0 IRR y>0 IRR y>0 IRR y>0 IRR y≥0 IRR y>0 IRR y>0 IRR y>0 Race/Ethnicity White (ref.) --- --- --- --- --- --- --- --- Black 1.24 *** 1.31 *** 1.14 * 1.18 ** 1.00 0.93 0.91 0.92 (0.06) (0.08) (0.06) (0.07) (0.07) (0.09) (0.08) (0.08) Native-born Hispanic 1.17 * 1.21 * 1.11 1.10 0.98 1.05 1.04 1.04 (0.08) (0.11) (0.10) (0.09) (0.09) (0.13) (0.09) (0.08) Foreign-born Hispanic 1.09 1.12 1.03 1.01 1.26 * 1.27 * 1.21 † 1.23 † (0.07) (0.10) (0.08) (0.09) (0.12) (0.15) (0.14) (0.14) Marital status a Ever married (ref.) --- --- --- --- --- --- --- --- Ever cohabited, never married 0.72 *** 0.67 *** 0.64 *** 0.65 *** 0.49 *** 0.51 * 0.48 * 0.53 * (0.05) (0.07) (0.06) (0.06) (0.07) (0.15) (0.14) (0.14) Single, never married 0.47 *** 0.63 * 0.63 ** 0.63 ** 0.19 *** 0.79 0.68 0.79 (0.09) (0.13) (0.11) (0.10) (0.06) (0.21) (0.17) (0.20) Age at first birth 1.01 ** 1.01 1.05 *** 1.04 *** (0.01) (0.01) (0.01) (0.01) First birth unintended 0.90 * 0.87 † (0.05) (0.07) Respondant mother's education Less than high school (ref.) --- --- High school 1.01 0.93 (0.06) (0.07) Some college 1.01 0.91 (0.06) (0.07) College 0.97 1.01 (0.07) (0.08) Family of origin characteristics R's mother worked 0.87 ** 0.92 † (0.04) (0.04) R's father absent at age 14 1.03 1.05 (0.06) (0.06) Religiosity Religion very important 1.02 1.15 ** (0.05) (0.05) Never attends relig. services 0.93 0.95 (0.07) (0.08) Family income 1st Quartile (lowest, ref.) --- --- 2nd Quartile 1.01 1.22 (0.08) (0.19) 3rd Quartile 0.97 1.19 (0.07) (0.18) 4th Quartile (highest) 1.09 1.34 * (0.08) (0.19) Rural residency 0.97 0.93 (0.06) (0.08) Survey year 2011–2013 0.93 † 1.08 1.01 1.02 0.76 *** 1.00 0.92 † 0.95 (0.04) (0.06) (0.05) (0.05) (0.04) (0.05) (0.04) (0.04) Age 1.19 1.25 1.10 1.07 0.79 † 0.89 0.80 0.82 (0.12) (0.18) (0.15) (0.14) (0.10) (0.15) (0.11) (0.11) Age squared 1.00 † 1.00 † 1.00 1.00 1.00 † 1.00 1.00 1.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Constant 0.01 * 0.00 * 0.05 0.11 12.27 1.11 13.65 10.40 (0.01) (0.01) (0.12) (0.27) (28.85) (3.50) (36.80) (26.43) N 1560 1456 1456 1456 1415 1229 1229 1229 Log likelihood -2854.2 -2456.5 -2303.0 -2289.2 -2806.1 -2203.9 -2092.6 -2075.2 †p < .10, *p < .05, **p < .01, ***p < .001 a Cumulative measure of marital status. Some college College 47 Interestingly, age at first birth, introduced in Model 3, reduces the racial differences among low educated native-born women, but instead increases the difference between White and immigrant Hispanic women. This finding suggests that the timing of first birth differently shapes the fertility outcomes of native-born and foreign-born women, providing again evidence that different forces drive the fertility behaviors of immigrant Hispanic women in line with the assimilation theory. Model 4, which introduces controls for other structural, cultural and demographic variables, reduces only the difference between Whites and Blacks with high school degrees, while the rest of the coefficients remain mostly unchanged. The second part of Table 3.4 shows the results for women with some college education and college graduates. Among women with some college education, the racial differences tend to be small or not statistically significant. Moreover, most of these differences are further reduced adjusting for age at first birth in Model 3, indicating that the higher fertility observed for Blacks and native-born Hispanics is partly explained by their early entrance into motherhood. Model 4, which introduces other control variables, shows virtually no changes. Among college educated women, however, the results indicate no significant racial/ethnic differences in the number of children among native- born ethnic groups, in line with the racial stratification theory. By contrast, college educated immigrant Hispanics have 23% more children than Whites according to Model 4, although this difference is significant at p < .10. The fact that educated native-born Hispanics do not differ from Whites while immigrant Hispanics exhibit higher fertility supports the assimilation theory and structural perspective. However, the lower fertility observed among college educated Blacks and native-born Hispanics in the descriptive 48 results (previously shown in Figure 3.2), are explained by age and marital status, thus, the results of this part of the analysis no longer support the minority group status hypothesis. This finding instead provides evidence indicating that college education homogenizes the fertility behaviors of native-born women of different racial/ethnic backgrounds. The effect of other structural and cultural factors are more limited to explain racial/ethnic differences once education is parceled out. The results indicate that being raised by a mother who completed high school reduces the risk of having a first child relative to those raised by mothers who did not complete high school, but this effect is only significant for women with high school degrees. Interestingly, religiosity is associated with earlier entrance into motherhood and higher fertility, but only for more educated women, and it has no effect among the low educated. Surprisingly, income, an important structural factor, has a small effect. The results provide some evidence that being in the highest income quartile decreases the risk of early childbearing, but only for women who have not completed high school. Moreover, high income seems to be associated with lower fertility for low educated women, but higher fertility for college educated women. Nonetheless, these factors do little to explain racial differences in the two fertility behaviors I analyzed in this study, namely age at first birth and number of children. 3.8.3. Summary of findings Table 3.5 provides a summary of the findings and evidence supporting each of the six theories evaluated. 49 Table 3.5. Summary of findings: Empirical evidence Theoretical perspective Empirical evidence Structural perspective Partially supported: No racial differences among the college educated. Socioeconomic and demographic factors partially accounted for racial/ethnic differences in age at first birth and number of children among low-educated groups. Cultural perspective Vaguely supported: Racial/ethnic differences among college educated were never significant. Socioeconomic and demographic characteristics explained some of the racial/ethnic differences at low levels of education. Classic assimilation theory Partially supported: Foreign-born Hispanics had more children than Whites after controls. Segmented assimilation theory Supported: The fertility behaviors of native-born Hispanics were more similar to those of Blacks than those of Whites. Racial stratification theory Supported: Socioeconomic factors reduced some of racial/ethnic disparities in fertility outcomes, but racial stratification was still evident among low educated groups. No significant differences among native-born college educated racial/ethnic groups, supporting the claim that socioeconomic resources provide tools to overcome historical disadvantage and discrimination. Minority group status hypothesis Partially supported: Net of controls, upwardly mobile minorities, such as college educated native- born Hispanics and Blacks, delayed entrance into motherhood more than college educated Whites. However, once they enter motherhood, they do not differ from Whites in achieved fertility, presumably because fertility rates among the college educated are already low (1.6). 50 3.9. Conclusions: Beyond Race? In the U.S. disadvantaged racial/ethnic minorities enter motherhood at younger ages and have more children on average than Whites, a pattern that has been linked to growing socioeconomic inequality. Early childbearing and higher fertility are associated with family instability, poorer educational and occupational outcomes among young women, parental immaturity, and poor children’s outcomes (McLanahan and Percheski 2008; McLanahan 2004; Western, Bloome, and Percheski 2008). Although prior studies have analyzed racial differences in fertility behaviors, they have relied on data from earlier periods, often times using a Black/White paradigm, excluding Hispanics from the analyses. In this Chapter, I use nationally representative data from the National Survey of Family Growth (NSFG) that expand until 2013, and include in the sample non-Hispanic Whites, non-Hispanic Blacks, native-born Hispanics, and foreign-born Hispanics. Using a series of advanced statistical methods, including a variation of Hurdle models, semi- parametric discrete time Cox models, Poisson regression, and zero truncated or conditional Poisson regression, I evaluate the different theories explaining racial/ethnic differences in fertility outcomes, including the structural and cultural arguments, the classic assimilation and segmented assimilation theories, the racial stratification theory, and the minority group status hypothesis. Prior studies have explained racial differences using mainly structural, cultural, and assimilation theories. Often times, the structural perspective and assimilation theory have prevailed, indicating that most of the racial differences are due to socioeconomic variations and length of residency in the U.S., while alternative theories have rarely been evaluated (Forste and Tienda 1996; Parrado and Morgan 2008; Telles and Ortiz 2008). In 51 this Chapter, I aimed to evaluate a larger set of theories explaining differences in fertility behaviors. Overall, the results provide partial support for the structural perspective, the assimilation theory, the segmented assimilation theory, and the minority group status hypothesis; while the cultural perspective argument did not receive substantial support. The fact that there is little variation in age at first birth and number of children regardless of race among college educated native-born groups, speaks to the power of class and structure. However, the racial/ethnic differences among the less educated indicate an important interaction between race and class, in which race only matters for the disadvantaged. The racial stratification theory, which combines the stratification and cultural perspective, received the strongest support. The racial stratification theory places structural factors at the base of racial/ethnic inequality, but also acknowledges the role of cultural norms and historical discrimination in explaining racial/ethnic differentials. This perspective suggests that racial differences will be greater among low educated groups, who given structural constraints, such as their poor socioeconomic resources, have a harder time overcoming the current and historical legacy of racial discrimination (Frank and Heuveline 2005; Telles and Ortiz 2008). The results from fully-interacted models with education show an overall gradient effect of education on fertility outcomes among native-born groups, revealing higher racial/ethnic differences in the timing of motherhood and fertility rates at low levels of education, and small or no differences at higher levels of education. In line with the racial stratification theory, the findings confirm that most of the racial/ethnic differences are concentrated at low levels of education (Musick et al. 2009; Yang and Morgan 2003). Conversely, the results indicate a different childbearing pattern for immigrant Hispanics. 52 The differences in the timing of motherhood between Whites and immigrant Hispanics increase rather than decrease with education, and, by contrast to native-born Hispanics, foreign-born Hispanics who are college educated exhibit a higher number of children than Whites, a finding consistent with the assimilation theory predicting higher fertility among first generation immigrants (Alba and Nee 1997). I claim that college education constitutes an important mechanism to reduce racial disparities in fertility behaviors, particularly for historically disadvantaged racial minorities whose family backgrounds and often segregated communities would have most likely prevented them from acquiring the fertility behaviors characteristic of the middle-class. I argue that besides opening up occupational opportunities, college institutions represent centers of social contagion and peer influence, homogenizing fertility behaviors among native-born racial/ethnic groups who share similar social educational settings. While norms and values pertaining to racial groups may shape fertility behaviors, attaining a college education in similar settings seems to override these differences. The results support this argument showing no significant racial differences among native-born college graduates, most of whom attend U.S. college institutions; but different behaviors for foreign-born college graduates, many of whom migrated after completing their education in their countries of origin (Calvo and Sarkisian 2014). The striking differences in age at first birth and number of children between college educated and less educated coethnics among African Americans and native-born Hispanics suggest that the social distance between lower and middle-class ethnic minority groups may have increased in recent decades. The results provide partial 53 evidence for the minority group status theory, which poses that in an effort to achieve upward mobility, racial/ethnic minorities may limit their fertility more than Whites, at the same time that they distance themselves from their more disadvantaged coethnics (Goldscheider and Uhlenberg 1969). I found that college educated native-born Hispanics delay entrance into motherhood more than college educated Whites. Moreover, once they enter motherhood, native-born Hispanics do not differ from Whites in achieved fertility, but greatly differ from their low educated coethnics. Among the low educated, earlier entrance into motherhood explain part of the higher fertility for Blacks and native-born Hispanics. Interestingly, age at first birth did not explain the difference in achieved fertility between Whites and immigrant Hispanics, confirming that different mechanisms shape the reproductive behaviors of immigrant women, in line with the assimilation theory (Frank and Heuveline 2005; Oropesa and Landale 2004; Parrado 2011). These findings indicate that not only race/ethnicity, but also immigrant status is an important factor for analysis. This research has some limitation. First, unfortunately, I was not able to distinguish between immigrant Hispanic women who attended college institutions in the U.S. from those who attended college abroad, which could elucidate the findings in this study (Calvo and Sarkisian 2014). In addition, I used years of education at the time of survey to classify respondents in educational categories. Because respondents are aged 32 and older, this is very close to measuring completed education. Nonetheless, I ignore their level of education when they became mothers and whether their education was completed before they finished childbearing. Moreover, although education is an indicator of socioeconomic status, education represents more than years of instruction 54 and potential economic gains. College institutions also have a cultural component acting as influential social milieux that normalize fertility behaviors, an aspect that is greatly overlooked in studies of fertility. Furthermore, structure and culture are intertwined; thus, although education is an important socioeconomic indicator, it is also an indicator of cultural capital, but in this study, I am not able to isolate these components. To overcome this limitation, I included in the analysis other socioeconomic and cultural measures including income, mother’s education, religiosity and urbanicity. Unfortunately, I was unable to include more direct measures of culture given limited variables in the dataset and the large percentages of missing values on alternative measures of culture. In addition, although in this Chapter I expand the traditional Black/White paradigm by including a sample of Hispanics, further distinguishing between native-born and foreign-born Hispanics, due to sample size limitations, I was not able to distinguish between Hispanics of 1.5, second or third generation, nor groups from different countries, thus the categorization Hispanic can be problematic because it obscures the great variability by subgroups (Alba and Nee 1997). However, some scholars have contended that this categorization may not be very problematic because the American society exposes ethnic minority groups to a similar racialization process (Frank and Heuveline 2005; Telles and Ortiz 2008). Finally, this analysis is limited to racial differentials in two main outcome variables, namely age at first birth and achieved fertility, and the results of this study may not apply to other outcomes in which racial differences may persist even after a college education has been attained. The results are in line with the idea that colleges act as mediums of social contagion, spreading norms and practices in childbearing behaviors. In 55 other words, the evidence of this research supports the argument that college education remains the strongest homogenizing force of fertility behaviors across racial/ethnic groups, emerging as an important tool for policy implementation to reduce racial inequality. 56 Chapter 4 No Nest? The Converging Growth of Childlessness in the United States, 1980s-2000s. 4.1. Introduction “None is Enough” read a recent headline in Time magazine, announcing an increase in the number of childless women (Sandler and Witteman 2013). The prevalence of women who end their reproductive lives without having had children grew significantly in the U.S. since the 1970s. However, the growth in childlessness is not a new phenomenon. The rates of childlessness have followed a U-shaped pattern that has mirrored historical economic circumstances (Kirmeyer and Hamilton 2011; Rowland 2007). In the U.S. childlessness reached a record high among women born in the early 1900s who became of reproductive age during the Great Depression. Among this cohort, the percentage of women remaining childless by age 49 reached 20%. Then, childlessness declined to 10% among the cohort of women born during the 1930s, who came of reproductive age during the baby boom period, when fertility increased and childbearing became nearly universal (Dykstra 2009; Kirmeyer and Hamilton 2011; Rowland 2007). Since the 1970s, as women began increasingly postponing or eschewing childbearing, the rates of childlessness increased again, reaching 16% during the early 2000s, meaning that nearly 57 one in six women will end their reproductive span without having had children (National Center for Health Statistics 2015). Despite the growing prevalence of childless women, childlessness has received little attention as research on fertility has conventionally focused on childbearing. Relatively few studies have analyzed changes over time in the key correlates of childlessness (Abma and Martinez 2006; Hayford 2013; Heaton, Jacobson, and Holland 1999; Livingston and Cohn 2010). As Chapter 3 indicated, childbearing behaviors have followed a bifurcating pattern as maternal age at first birth increased faster among White and college educated women than among ethnic minorities and low-educated women. The more privileged are also are more likely to have children within marriage than disadvantaged women (Cherlin 2010; Ellwood and Jencks 2004; McLanahan 2009). Given the recent growth in socioeconomic inequality, which disproportionately increases the opportunity cost of having children among women with better job prospects and higher earnings potential, we would expect that the likelihood of remaining childless would have increased faster for advantaged women; however, as the analysis in this Chapter will show, childlessness instead increased faster among disadvantaged groups. In this Chapter, I show that the associations between childlessness, on the one hand, and partnership status, race, and education, on the other hand, became weaker from the 1980s to the late 2000s. Contrary to the increase in the divergence in the timing and partnership status of childbearing by class and race, the results in this chapter show instead a convergence in the prevalence of childlessness from the 1980s to the early 2000s in key sociodemographic characteristics. 58 Because fertility outcomes affect women’s life chances and their opportunities for social mobility (Kahn, García-Manglano, and Bianchi 2014; Morgan 1996; Rumbaut 2005), understanding the trends in childlessness by sociodemographic variables sheds light on how these patterns are shaped by and reproduce social inequality. The purpose of this Chapter is threefold: (i) to evaluate the associations between the levels of childlessness and key socioeconomic and demographic correlates, including marital status, education, race/ethnicity, and income (ii) to investigate how the associations have changed over time from the 1980s to the late 2000s, identifying the factors that have become more relevant and those that have receded in importance; and (iii) to compare the trends in childlessness to the ones observed for childbearing. 4.2. Background Although historically childlessness has reached relatively high levels in the past, the increase in childlessness since the 1970s has occurred amid a time of drastic social change affecting family life in the U.S., including the rise in women’s education and labor force participation, the diffusion of effective contraception, delays in marriage, high rates of divorce, and an increasing emphasis on individualism and self-development (Casper and Bianchi 2002; Cherlin 2004; Koropeckyj-Cox and Pendell 2007; Lesthaeghe 1995). It has now become more acceptable for individuals to eschew long-term commitments such as marriage and children. Parenthood is increasingly seen as a voluntary choice and a means for self-realization, and childlessness has lost some of its stigmatized status (Kirmeyer and Hamilton 2011; Koropeckyj-Cox and Pendell 2007; Lesthaeghe 1995). I build up on previous research by analyzing key correlates of 59 childlessness over time, including marital status, education, race/ethnicity, and income (Bloom and Trussell 1984; Hayford 2013; Smock and Greenland 2010), using data from the National Survey of Family Growth (NSFG) from 1982 until the early 2010s. In this Chapter, I investigate the extent to which women who remain childless by the end of their reproductive years differ among these dimensions, and compare these results to the ones obtained in Chapter 3. I show that marital status, education, and race/ethnicity have become weak predictors of childlessness over time. I argue that different mechanisms drive the patterns of childbearing and childlessness among American women. 4.2.1. Childlessness across Marital and Partnership Status Marital status has historically been a major predictor of fertility; however, as the social norms mandating that childbearing should occur within marriage waned, the link between partnership status and fertility also weakened (Hayford, Guzzo, and Smock 2014; Musick 2002). Delays in marriage imply a growing proportion of unmarried women, which would increase the overall prevalence of childlessness. Nonetheless, this trend has been counteracted by the increase in births to single and cohabiting women. Since the 1970s, nonmarital childbearing has increased, particularly among disadvantaged groups, as singlehood and cohabitation gained acceptance as alternative family settings for reproduction. However, cohabitation has remained qualitatively different from marriage given that individuals who enter cohabiting unions tend to expect greater personal freedom, which may discourage childbearing and bolster childlessness (Guzzo and Hayford 2014; Hayford 2013). In general, the rates of childlessness among cohabitors lie between those of single and married individuals. 60 As predicted by the second demographic transition (SDT) theory, childlessness declined among the never married, and increased among married couples (Casper and Bianchi 2002; Hayford 2013; Rowland 2007). Yet, in the U.S. most women still consider marriage a prerequisite for childbearing. Although the recent increase in childlessness could be attributed to the increase in the proportion of never married women, the rise in childlessness in the U.S. has been mainly the result of the postponement of marriage, the increase in marital childlessness, and the relatively high divorce rates, which increases the proportion of unpartnered women (Dykstra 2009; Hagestad and Call 2007; Heaton, Jacobson, and Holland 1999; Quesnel-Vallée and Morgan 2003; Rowland 2007). In addition, because couples are marrying later in life, and are waiting longer to have children, the probability that many of these couples will become permanently childless has increased, not only because they may grow accustomed to their child-free lifestyles, but also because the risk of age-infertility related problems increases with age (Hayford, Guzzo, and Smock 2014; Heaton, Jacobson, and Holland 1999). A prior study indicated that nearly all couples who remain childless ten years after getting married become permanently childless (Rowland 2007). Consequently, childlessness, which has traditionally been in the realm of the unmarried, has spread among the married (Abma and Martinez 2006; Rowland 2007). The recent trends in union formation and fertility patterns suggest that marital and partnership status has lost some of its centrality as a predictor of permanent childlessness (Hayford, Guzzo, and Smock 2014). 4.2.2. Trends in Childlessness by Educational Attainment Women’s education has substantially increased since the 1960s. Since 2004, women have been graduating from 4-year college at higher rates than men (Buchmann and DiPrete 61 2006). Education is one of the major dimensions of social stratification affecting not only most socioeconomic outcomes, but also fertility behaviors. Nonetheless, the directionality of the relationship between education and fertility is not always clear. Attaining higher education reduces the risk of early childbearing as well as overall fertility rates; however, childbearing can also affect educational attainment, and both can be affected by exogenous factors and selectivity (Marini 1984; Stange 2011). Yet, higher education is positively correlated with delays in fertility, low achieved fertility, and higher levels of childlessness (Bloom and Trussell 1984; Musick et al. 2009; Stange 2011). Childless women have traditionally been overrepresented among college graduates, partly because women tend to postpone childbearing to attend college, and later on, to get established in a career. Giving the intense time demands of a newborn, having a child during early adulthood threatens women’s chances of completing their expected level of education and obtaining a career. Because sterility rises with age, delaying childbearing past age 30 increases the likelihood of becoming permanently childless (Kirmeyer and Hamilton 2011; McQuillan et al. 2003; Quesnel-Vallée and Morgan 2003). The opportunity cost of childbearing is often cited as one of the main reasons for the high rates of childlessness among college educated women (Abma and Martinez 2006; Bloom and Trussell 1984; Hayford 2013). However, college education not only opens up alternatives to motherhood, but it also reinforces nontraditional values and alternative child-free lifestyles (Lesthaeghe 1995; Rowland 2007). Moreover, higher education is associated with better information on birth control and access to more efficient contraceptive methods, which reduces the risk on unintended pregnancies. 62 The stakes of higher education substantially increased in recent decades. Since the mid-1970s, the earnings for college graduates have substantially risen, while the earnings for those with a high school diploma or less either have decreased or remained stagnant, resulting in a greater disparity in the returns to education (Autor, Katz, and Kearney 2008; Western, Bloome, and Percheski 2008). Occupational status and employment rates have also become more strongly correlated with college education. These changes imply that the opportunity cost to childbearing has grown faster for the highly educated, and thus, we would expect a higher increase in the levels of childlessness among college educated women. In line with this argument, early research has found and increasing positive association between education and childlessness (Bloom and Trussell 1984). However, college educated women are more likely to have not only higher earnings but also husbands with greater earnings power, both of which allow these women to be able to afford childcare services and outsource housework, which reduces the incompatibility of work and childrearing responsibilities. These two mechanisms, opportunity costs and greater resources can result in offsetting effects, weakening the link between college education and levels of childlessness. Moreover, although low-income women may face lower opportunity costs in absolute value, their relative value opportunity cost may be high enough to make childbearing unaffordable. In line with this argument, and as the results of this study will show, childlessness has increased faster among women with low education in recent decades (Hayford 2013; Livingston and Cohn 2010). This trend could further contribute to weaken the relationship between college education and childlessness. In this chapter, I analyze how this association has 63 changed from the 1980s to the late 2000s, net of other socioeconomic and demographic factors. 4.2.3. Childlessness by Race/Ethnicity The levels of childlessness also differ by race/ethnicity. Childlessness has traditionally been seen as a White phenomenon driven by the higher prevalence of voluntary child- free lifestyles. Childlessness among African American women was considered to be the result of sterility and health problems, implying that it was mostly an involuntary outcome. Nonetheless, sterility and fecundity issues among Blacks were greatly reduced after World War II, and the rates of infertility among Whites and Blacks became similar (Boyd 1989; Lundquist, Budig, and Curtis 2009). More recently, childlessness among African Americans has been linked to their low marriage rates, as African American women face a small pool of “marriageable men” (Lundquist, Budig, and Curtis 2009; Wilson and Neckerman 1987). Young Black men not only have low earnings power, but they also exhibit the highest rates of unemployment, incarceration, and mortality. Some scholars have suggested that given the relatively low status of African American men, African American professional women often choose to remain single and childless as a strategy to achieve or retain middle class status (Marsh et al. 2007). By contrast to earlier research, recent studies suggest that the factors driving childlessness, including the intentionality to remain childless, are now very similar between Whites and African Americans (Boyd 1989; Lundquist, Budig, and Curtis 2009; Marsh et al. 2007). Nonetheless, not only opportunity costs and situational barriers vary across racial/ethnic groups, but so do cultural norms. Some scholars have argued that racial 64 minorities, including African American and Hispanics have more familistic cultural norms, placing a higher value in motherhood (Hartnett and Parrado 2012; Landale and Oropesa 2007; Landale, Schoen, and Daniels 2010). Consistent with this argument, some scholars have found evidence indicating that racial minorities are more often exposed to pronatalist social messages that stigmatizes childlessness (McQuillan et al. 2012). Heaton et al. (1999) found that Black women were less likely to desire to remain childless. However, since the 1970s the levels of childlessness by race/ethnicity began converging, in great part driven by an increase in childlessness among ethnic minorities. In the late 2000s, the levels of childlessness by race/ethnicity for women ages 40-44 ranged from 14.9% among Whites, 12.3% among African Americans, but only 7.1% among Hispanics (National Center for Health Statistics 2015). In this study, I analyze the differences in the prevalence of childlessness by race/ethnicity over time, controlling for socioeconomic and demographic factors. 4.2.4. Childlessness and Income Economic factors are also strongly associated with fertility behaviors; however, income is usually left out of most analyses because of methodological challenges. First of all, income is correlated with other socioeconomic variables such as education and family background characteristics. Higher education often leads to better job opportunities and to an increased probability of attracting an educated partner, both of which results in a positive association between education and income (Musick, Brand, and Davis 2012; Schwartz and Mare 2012). The relationship between individual’s income and the characteristics of family of origin is also well-established. Parental resources, such as 65 education and income, provide opportunities to children’s lives that affect children’s reproductive behaviors during early and young adulthood (Landale, Schoen, and Daniels 2010). Because of these correlations, most studies on fertility use individual’s education or family background characteristics as proxies for economic well-being, leaving income out of the equation. I expand current research on childlessness by exploring whether income is independently associated with the levels of childlessness once education and family background characteristics are taken into account, and evaluating whether this association has changed over time. Moreover, I use two different measures of economic well-being: (1) family income quintiles, and (2) family income as a percentage of the poverty threshold. In sum, in this Chapter, I build upon prior research by analyzing the associations between childlessness and four key sociodemographic factors, namely marital status, education, race/ethnicity, and family income using data for two recent cohorts of women, one who approached the end of their reproductive lives during 1982–1988, and the other who did so during 2006–2010. I evaluate how these associations have changed controlling for family background characteristics and other correlates of fertility behaviors. 4.3. Data, Measures, and Methods 4.3.1. Data I use data from the National Survey of Family Growth (NSFG) and the Integrated Fertility Survey Series (IFSS) as described in Chapter 2. In this chapter, I use data from the IFSS cycles 3 (1982) and 4 (1988), and from the continuous NSFG survey 2006–2010 66 and 2011-2013. Because I focus on childlessness, I restrict the sample to women aged 40–44 who have virtually ended their reproductive cycles. Although some women may still become mothers, the probability of having a first birth past age 40 is relatively low, thus, this is the closest we can come to approximating permanent childlessness using these data (Abma and Martinez 2006; J. A. Martin et al. 2015). 5 I use the 1982 data as the baseline for comparison because this is the first year the NSFG included never married women in the survey. To increase the sample size, I combined data from cycle 3 (1982) and cycle 4 (1988). Although women in 1988 were more likely to be childless than in 1982, the difference was relatively small. To test whether the results were affected by pooling together these two cycles, I conducted a sensitivity analysis replicating the models with a set of fully interacted dummy period variables for 1988; only a few of the interactions were significant and the overall main results were not compromised. The two samples represent two cohorts of women, one born between 1938 and 1948, who reached ages 40–44 during 1982–1988, and a more recent cohort, born between 1962 and 1970, who reached ages 40–44 during 2006–2010. The original samples contain 1,982 women aged 40–44 for the combined cycles 1982–1988, and 1,708 for 2006–2010. I restrict the sample to White, African American, and Hispanic women. A small number of respondents in the other race/ethnic category were excluded 5 In 2013 the first birth rate, i.e. the number of first births per 1,000 women, for women 40–44 years old was only 10.4 compared with 80.7 for women aged 20–24, 105.5 for those aged 24–29, 98.0 for those aged 30–34, and 49.3 for women aged 35–39 (J. A. Martin et al. 2015). 67 from the sample – a total of 50 in the 1982–1988 sample, and 103 in 2006–2010. I also excluded mothers with no information on the timing of their first birth, those who reported having had their first child before age 12, and cases with missing values in some of the main covariates. My final sample size is 1,874 for 1982–1988, 1,592 for 2006– 2010, and 710 women for the preliminary release of the 2011–2013 wave. Because as of the writing of this dissertation, complete data for the 2011–2015 cycle has not been released, I use the 1982, 1988, and 2006–2010 waves for the main analyses, and I present separate results for 2011–2013; the reader should use caution when interpreting the preliminary results. 4.3.2. Measures In this chapter the dependent variable is a dichotomous variable that takes the value of 1 for women who have not had any biological children by the time of the survey and she is between the ages 40–44. My main independent variables are marital status, education, race/ethnicity, and income; I also include other relevant control variables. I measure marital status in two categories, ever-married (referent category), and never-married. I was unable to identify cohabiting women because the measures for cohabitation were not available in the 1982 cycle. Education is measured in four categories: less than high school, high school (referent), some college, and college education. Race/ethnicity consists of mutually exclusive dummy variables for non-Hispanic Whites (referent group), from now on simply Whites; non-Hispanic African Americans, from now on African Americans, and Hispanics of any race. 68 Measuring economic well-being at the individual level poses some challenges, given that many women leave the labor market or disinvest in their careers once they get married or have children (Budig and Hodges 2010; Kahn, García-Manglano, and Bianchi 2014; Stone 2007); thus, women’s earnings may not adequately reflect their economic well-being. To overcome this difficulty, I assess economic well-being by family income measured in quintiles, which also facilitates comparisons across periods. I imputed a small fraction of missing cases in the samples, 16% in 1982 and 7% in 1988, using a multivariate equation. Imputations and quintile distributions were performed using the entire dataset to approximate the quintile income distribution in the underlying population, i.e. all respondents aged 15-44. I conducted two sensitivity analyses, one excluding cases with missing family income, and a second one replicating the models using the ratio of income to poverty as an alternative measure of economic well-being; I obtained similar findings using both approaches. I controlled for family background using the following three measures. Mother’s education, measured with a dummy variable coded 1 for respondents whose mothers had some college education or a college degree, and 0 otherwise. Mother’s work status, measured with a dummy variable coded 1 for respondents whose mother worked at least part time while the respondent was age 14, and 0 otherwise. Family structure was indicated by a dummy variable coded 1 for respondents who lived with both biological or adoptive parents at age 14, and 0 otherwise. I also controlled for infertility or impaired fecundity. I used the variables fecund, which measures fecundity for all women, and infert, which assesses infertility for married or cohabiting women. These two variables do not completely overlap given that infert 69 includes the fertility status of women’s partners, which could be the reason why some partnered women remain childless (Chandra, Copen, and Stephen 2013; McQuillan et al. 2003). Following Martinez et al. (2012), I combined these two variables to obtain a better assessment of physical barriers to motherhood, and I defined fecundity issues as having impaired fecundity or being sterile for purposes other than contraception. I controlled for religiosity and urbanicity, both of which have been found to influence fertility outcomes (Boyd 1989; Chandra et al. 2005; Koropeckyj-Cox and Call 2007). I use attendance at religious services as a proxy for religiosity, measured with a dummy variable coded 1 for those who never attended religious services, and 0 otherwise. I tried other measures of religiosity such as religious denomination; however, similar to the findings of a previous study (Abma and Martinez 2006), I found no significant differences between mothers and childless women in religious denomination. Finally, I controlled for urban residency with a dummy variable coded 1 for respondents who lived in a city or metropolitan area, and 0 if they lived in a rural area. 4.3.3. Methods Because childlessness is often a process resulting from successive decisions to postpone parenthood, changing over the life course (Hagestad and Call 2007), I begin by using survival analysis, also called event history analysis, to explore the bivariate associations between childlessness and my four main independent variables; marital status, education, race/ethnicity, and family income. I present these results using survival curves, which summarize bivariate associations. The survival curves depict the monthly risk or conditional probability of remaining childless starting at age 12, conditional upon not 70 having had a child before. Because we cannot model a nonevent (e.g. not having had a child), I model instead the risk of having a first birth, and take the survival function from this model to obtain the conditional probability of remaining childless (Cleves et al. 2010). All analyses account for survey design and sample weights. 6 For the multivariate analyses, I use logistic models predicting the odds of being childless by age 40–44 and evaluate the association between childlessness and the main independent variables, controlling for other factors. I present log-odds coefficients and odds ratios, and summarize the findings by graphing the predicted probability of remaining childless based on the multivariate models for each period. Although the NSFG provides event history data, it is not longitudinal and, thus, the timing of some events, such as educational trajectories, is not available. Given these limitations, I evaluate associations and not causal effects. 4.4. Results 4.4.1. Descriptive Results Table 4.1 shows weighted descriptive statistics for the sample. In 1982–1988, 12.5% of women aged 40–44 were childless, and this percentage increased to 15.8% in 2006–2010. The preliminary data release for 2011–2013 indicates that this percentage may have slightly decreased to 13.1% in recent years. As Table 4.1 shows, childless women’s sociodemographic characteristics differ from those of mothers. I conducted t-tests to identify significant differences between mothers and childless women in each period. Childless women are significantly more likely than mothers to be never-married by age 6 To avoid inflating the standard errors, I use adjusted weights that average to 1. 71 40–44, but the difference decreased across periods; by 2006–2010, 18.0% of childless women were never married vs. 2.1% of mothers. The preliminary results from 2011– 2013 indicates that the gap was further reduced as the proportion of never married increased to 22.2% among childless women and 12.1% among mothers, however given the small sample size, this difference did not reach statistical significance. Table 4.1 also Table 4.1. Characteristics of childless women and mothers aged 40 to 44. Weighted percent distribution: NSFG 1982–1988, 2006–2010, and preliminary release of 2011–2013. All Childless Mothers All Childless Mothers All Childless Mothers Overall distribtuion 100% 12.5% 87.5% 100% 15.8% 84.2% 100% 13.1% 86.9% Marital status Ever-married 95.1% 70.3% 98.7% *** 95.4% 82.0% 97.9% ** 86.5% 77.8% 87.9% Never-married 4.9% 29.7% 1.3% *** 4.6% 18.0% 2.1% ** 13.5% 22.2% 12.1% Education Less than high school 17.9% 9.0% 19.2% *** 14.7% 9.0% 15.7% 10.2% 1.6% 11.5% * High school 38.3% 29.9% 39.5% * 30.8% 23.6% 32.1% 25.1% 17.4% 26.2% † Some college 24.7% 24.6% 24.7% 26.8% 27.2% 26.7% 27.7% 22.0% 28.5% † College 19.1% 36.5% 16.6% *** 27.7% 40.2% 25.4% * 37.0% 58.9% 33.7% * Race/ethnicity White 82.4% 87.2% 81.7% * 69.7% 77.7% 68.2% 67.8% 77.0% 66.4% † African American 10.6% 10.2% 10.7% 15.0% 15.2% 15.0% 14.3% 13.4% 14.4% Hispanic 7.0% 2.6% 7.7% *** 15.3% 7.1% 16.8% * 17.9% 9.6% 19.2% ** Family income in quintiles b 1st (Lowest) 13.1% 10.5% 13.4% 14.9% 8.2% 16.1% ** 13.3% 13.0% 13.3% 2nd 15.8% 19.1% 15.3% 13.7% 12.6% 14.0% 17.5% 15.4% 17.8% 3rd 17.1% 20.1% 16.7% 20.8% 24.0% 20.2% 23.9% 25.3% 23.6% 4th 21.8% 15.8% 22.7% 20.4% 20.3% 20.5% 21.2% 24.7% 20.7% 5th (Highest) 32.1% 34.6% 31.8% 30.1% 34.8% 29.2% † 24.2% 21.6% 24.6% Family background Mother has some college educ 14.8% 25.2% 13.3% *** 30.7% 35.9% 29.7% * 38.5% 54.9% 36.0% Mother worked 45.1% 41.7% 45.6% 62.1% 68.8% 60.8% † 70.5% 70.7% 70.5% Lived with both parents at 14 79.7% 83.4% 79.1% 65.4% 61.2% 66.1% 62.5% 64.4% 62.3% Fecundity Issues Reports fecundity problems 29.4% 39.5% 27.9% ** 18.5% 43.6% 13.8% ** 18.1% 27.0% 16.7% Attendance to religious services Never attends religious service 11.4% 11.3% 11.5% 19.3% 31.7% 16.9% * 24.0% 27.9% 23.4% Urbanicity Lives in urban area 77.0% 87.7% 75.4% *** 78.7% 88.8% 76.8% † 85.6% 89.0% 85.1% Unweighted sample size 1,874 211 1,663 1,592 335 1,257 712 134 578 a T -tests for differences in means between mothers and childless women: †p < .10; *p < .05; **p < .01; ***p < .001 b Income quintiles created using the entire population. Respondents aged 40–44 were overrepresented in the higher income quintiles. t - test a t - test a t - test a 2006–2010 1982–1988 2011–2013 72 shows that education increased for all women over time, yet, childless women are more likely to be college educated than mothers; in 2006–2010, 40.2% of childless women were college educated, compared with only 25.4% of mothers. Education continues increasing, by 2011–2013, 58.9% of childless women had a college degree as well as 33.7% of mothers. Whites comprised a significantly higher proportion of childless women than mothers in 1982–1988, but the difference in 2006–2010 was not statistically significant; while Hispanics exhibit the opposite pattern, comprising a smaller proportion of childless women in all periods; in 2006–2010 Hispanics represented 7.1% of all childless women, but 16.8% of all mothers, and this difference was statically significant. By 2011–2013, Hispanics were still significantly more likely to be mothers than to remain childless. Conversely, African Americans comprised similar proportions of childless women and mothers in each period. Although I did not find any significant differences between mothers and childless women by family income quintiles in the 1980s and in the preliminary 2011–2013 sample, in 2006–2010 childless women were slightly better-off; they were less likely than mothers to be in the lowest income quintile – 8.2% vs 16.1% – and slightly more likely to be in the highest income quintile, – 34.8% vs. 29.2% – although the latter difference was only marginally significant (p < .10). By family background characteristics, childless women were more likely to have had a mother with at least some college education in all periods, in 2006–2010 this percentage was 35.9% for childless women and 29.7% for mothers. These percentages increased by 2011–2013, however, the difference did not reach statistical significance. In the latter period, childless women were also more likely than mothers to have been raised by a mother who was employed, 68.8% vs. 60.8%, but this difference was only 73 marginally significant (p < .10) and no difference is observed in 2011–2013. In all periods, a higher percentage of childless women than mothers reported having ever experienced fecundity problems themselves or through their partners; in 2006–2010, 43.6% of childless women reported having ever experienced fecundity problems compared with 13.8% of mothers. The difference seems to have been reduced in 2011– 2013, but did not reach significance level. In the 1980s, I did not find a significant difference between mothers and childless women in the proportion who never attended religious services, my proxy for religiosity; by the late 2000s, this proportion increased for all women, but more so among childless women, 31.7% vs. 16.9%, probably as a result of the increasing secularization process characteristic of the Second Demographic Transition (Lesthaeghe 1995). By 2011–2013 the proportion of women who never attended religious services seems to have increased among mothers, and the difference with childless women was not significant. Finally, in both periods childless women were more likely to be living in urban areas; in the late 2000s, 88.8% of childless women lived in urban areas compared with 76.8% of mothers, however the preliminary results for 2011–2013 indicates that there may no longer be a difference. In what follows, I present my results comparing the 1982–1988 to the 2006–2010 period. Because, as of the writing of this dissertation, data collection has not been finalized for the 2011–2015 period, I will refer to the preliminary 2011–2013 results when relevant and will present those analyses separately. Figure 4.1 plots a bivariate comparison of predicted survival curves by age for the main independent variables for the 1982–1988 to the 2006–2010 periods. The y-axis indicates the estimated proportion of 74 Figure 4.1 Estimated survival curves by sociodemographic characteristics A. Marital status B. Educational attainment C. Race/ethnicity 75 D. Family income quintiles women remaining childless as they age. I included a reference line at age 40 for comparison purposes. Overall, Figure 4.1 shows that all survival curves shifted to the right across periods, revealing an increased postponement of childbearing, and higher proportions of women at each age remaining childless. The reduction in the slopes of the survival curves reflect the de-concentration of childbearing (Kirmeyer and Hamilton 2011; Musick 2002); in 1982–1988, first births were more heavily concentrated in the early 20s, as evidenced by the steeper slopes; by 2006–2010, first births were more spread out across the 20s and 30s, reflected by the flatter slopes. Panel A of Figure 4.1 shows the results by marital status. Across periods, we observe a decline in the proportion of never-married women who remain childless, reflecting the growth in nonmarital births over time. By contrast, among ever-married women, we observe higher delays in childbearing and a higher proportion of them remaining childless. Overall, this graph shows a smaller difference between never- married and ever-married women in their probability of remaining childless. I tested for significant changes over time by estimating a pooled model fully interacted with period, 76 the t-test revealed that this change was statistically significant (p < .001). The magnitude of this change can be appraised by the reduction in the vertical distance between the survival curves of ever-married and never-married women from 1982–1988 to 2006– 2010 at age 40, indicated by the reference line. Panel B shows that all women are postponing childbearing regardless of their level of education, but more so women with a college education. In general, higher education is associated with an increased probability of remaining childless; however, this pattern is more pronounced for the college educated. College educated women were significantly more likely to remain childless than high school graduates in both periods, and no significant change was found across periods. By race/ethnicity, see Panel C, we observe small differences in the 1982–1988 period, and although all racial/ethnic groups show delays in motherhood, the delay was more pronounced for Whites, evidenced by the flatter slope in their survival curve. Although in each period Whites were more likely to remain childless at the end of their reproductive lives than African Americans and Hispanics, the differences were not as pronounced as the ones observed by marital status and education. Panel D shows the bivariate results by family income quintiles. Overall, women with higher family incomes were more likely to remain childless by ages 40–44. In 1982– 1988, we observe some small differences by income quintiles, and only a few of them were significant; by 2006–2010, a more pronounced income gradient emerged, and the differences between the lowest income quintile and the top three income groups became statistically significant. These differences were also significant across periods. 77 4.4.2. Multivariate Results The previous results presented bivariate associations without control variables; Table 4.2 presents the results from the multivariate logistic regression models predicting the odds of being childless by ages 40–44, and Table 4.3 presents the preliminary results for 2011– 2013. As Table 4.2 shows, from 1982–1988 to 2006–2010, we observe a significant decline in the odds that never-married women remain childless relative to ever-married women, net of other factors. To better grasp these results, Figure 4.2 illustrates the predicted probability of remaining childless based on the logistic models presented in Table 4.2 by the main independent variables, setting all other variables to their means. Panel A in Figure 4.2 shows that in 1982–1988, 82.6% of never-married women were predicted to remain childless compared with only 6.7% of ever-married women, a difference of 76 percentage points, by 2006–2010 the difference declined to 51 points – 59.9% for never-married women were predicted to remain childless vs. 9.0% of ever- married women. These results suggest that although marital status is still a very important predictor of childlessness, its predicting power decreased by the late 2000s. The preliminary results for 2011–2013 (Table 4.3) suggest that this difference has further decreased in recent years and the difference no longer reaches statistical significance. In other words, net of other factors, ever-married and never-married women have become more alike in their probability of remaining childless. This convergence is at odds with the results on the timing of childbearing presented in Chapter 3, showing that a polarization in timing of first births; unmarried women tend to enter motherhood at younger ages than married women, especially among disadvantaged groups, and these 78 Table 4.2. Estimated coefficients and odds ratios from logistic regression of childlessness on sociodemographic characteristics. NSFG 1982–1988 and 2006–2010: Women aged 40 to 44. Odds Ratio Odds Ratio β p S.E. Exp β β p S.E. Exp β Marital status Ever-married (Ref.) --- --- --- --- --- --- Never-married 4.19 *** 0.41 65.83 2.71 *** 0.46 15.05 Education Less than high school -0.24 0.29 0.79 -0.12 0.68 0.89 High school (Ref.) --- --- --- --- --- --- Some college 0.30 0.30 1.35 0.48 *** 0.12 1.61 College 1.20 *** 0.27 3.31 0.82 ** 0.27 2.28 Race/ethnicity White (Ref.) --- --- --- --- --- --- African American -0.67 † 0.38 0.51 -0.40 *** 0.06 0.67 Hispanic -1.12 * 0.50 0.33 -0.95 *** 0.19 0.39 Family income in quintiles 1st (Lowest) (Ref.) --- --- --- --- --- --- 2nd 1.04 * 0.41 2.83 0.81 ** 0.29 2.25 3rd 1.01 ** 0.36 2.76 1.33 ** 0.42 3.77 4th 0.37 0.38 1.44 0.91 *** 0.23 2.49 5th (Highest) 0.70 * 0.35 2.01 0.94 * 0.45 2.56 Family background Mother has some college educ. 0.25 0.20 1.29 -0.11 0.09 0.90 Mother worked 0.07 0.20 1.07 0.33 0.22 1.38 Lived with both parents at 14 0.30 0.28 1.35 -0.14 0.28 0.87 Fecundity Issues Reports fecundity problems 1.06 *** 0.16 2.88 1.94 *** 0.20 6.95 Attendance to religious services Never attends religious services -0.18 0.29 0.83 1.00 *** 0.13 2.71 Urbanicity Lived in urban area 0.65 ** 0.23 1.92 1.03 *** 0.24 2.80 Intercept -4.48 *** 0.51 -4.74 *** 0.65 N 1874 1592 Log likelihood -583.2 -662.8 †p < .10; *p < .05; **p < .01; ***p < .001 1982–1988 2006–2010 79 6.7% 9.0% 82.6% 59.9% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1982–1988 2006–2010 Ever Married Never Married 5.1% 6.6% 6.4% 7.4% 8.4% 11.4% 18.4% 15.4% 0% 5% 10% 15% 20% 25% 30% 1982–1988 2006–2010 Less than HS High School Some college College 3.2% 5.1% 5.0% 8.5% 9.3% 12.1% 0% 5% 10% 15% 20% 1982–1988 2006–2010 Hispanic African American White 0% 5% 10% 15% 20% 1 (Lowest) 2 3 4 5 (Highest) 1982–1988 2006–2010 Figure 4.2 Predicted probability of remaining childless by sociodemographic factors based on multivariate logistic models a A. Marital status B. Educational attainment C. Race/ethnicity D. Family income quintiles a Predicted probabilities based on multivariate logistic regression holding all other variables to their means. trends diverged until the late 2000s (Ellwood and Jencks 2004; Hayford, Guzzo, and Smock 2014; McLanahan 2009). 80 Table 4.3. Estimated coefficients and odds ratios from logistic regression of childlessness on sociodemographic characteristics. Preliminary NSFG 2011–2013: Women aged 40 to 44. Odds Ratio Odds Ratio β p S.E. Exp β β p S.E. Exp β Marital status Ever-married (Ref.) --- --- --- --- --- --- Never-married 1.02 † 0.61 2.77 0.86 0.56 2.37 Education Less than high school -1.59 * 0.73 0.20 -1.70 * 0.73 0.18 High school (Ref.) --- --- --- --- --- --- Some college 0.28 0.30 1.33 0.32 0.36 1.38 College 1.12 ** 0.39 3.05 1.49 ** 0.47 4.45 Race/ethnicity White (Ref.) --- --- --- --- --- --- African American -0.05 0.29 0.95 -0.14 0.31 0.87 Hispanic -0.26 † 0.15 0.77 -0.28 * 0.13 0.76 Family income in quintiles 1st (Lowest) (Ref.) --- --- --- 2nd -0.01 0.19 0.99 3rd -0.22 0.34 0.81 4th -0.42 0.35 0.66 5th (Highest) -1.11 0.73 0.33 Family background Mother has some college educ. 0.61 † 0.33 1.83 0.66 † 0.35 1.93 Mother worked -0.07 0.39 0.93 -0.17 0.43 0.85 Lived with both parents at 14 -0.22 0.24 0.80 -0.18 0.20 0.84 Fecundity Issues Reports fecundity problems 0.82 0.53 2.27 0.69 0.55 1.99 Attendance to religious services Never attends religious services 0.31 * 0.14 1.36 0.40 † 0.21 1.50 Urbanicity Lived in urban area 0.18 0.25 1.20 0.30 0.22 1.34 Intercept -3.02 *** 0.44 0.05 -2.83 *** 0.63 0.06 N 712 712 Log likelihood -323.8 -318.0 †p < .10; *p < .05; **p < .01; ***p < .001 Model 1 Model 2 (+ Income) 81 The differences in the probability of remaining childless by educational attainment have also decreased over time, net of other factors. Overall, as education increases, so does the probability of remaining childless, as observed in Panel B of Figure 4.2. In 1982–1988, the differences in the probability of remaining childless among individuals without college degrees were not statistically significant, varying from 5.1% for women who did not graduate from high school, to 6.4% for high school graduates, and 8.4% for women with some college education; however, college graduates were 3 times more likely to remain childless than high school graduates 18.4% vs. 6.4% (p < .001). By 2006–2010, the probability of remaining childless increased for low educated women ranging from 6.6% for high school dropouts, to 11.4% for those with some college education, but, net of other factors, the probability of remaining childless for college graduates slightly decreased to 15.4%, reducing the overall educational gap. Given the increased disparities in the returns to education, I would have expected that college graduates would be even more likely to remain childless than less educated women; however, counter to this expectation, the educational differences were instead reduced. Again, in contrast to the increasing divergence by educational attainment in the timing and marital context of childbearing, I did not find increasing differences in the levels of childlessness by education; instead, the results suggest a converging pattern. Nonetheless, the initial results for 2011–2013 suggests that education may be gaining importance again, as the educational gap in the prevalence of childlessness seems to have increased in recent years, in line with the growing economic opportunity costs argument. Future research should confirm this preliminary finding once data collection has been finalized. As in previous studies (Hayford 2013; Lundquist, Budig, and Curtis 82 2009), I tested for interactions between marital status and educational attainment. However, the interactions were not statistically significant. Thus, I only present the more parsimonious model with main effects only shown in Table 4.2. Panel C in Figure 4.2 shows the results by race/ethnicity. In general, the probability of remaining childless has increased for all racial/ethnic groups. Net of other factors, Hispanic women aged 40–44 exhibit the lowest predicted probability of being childless, but it increased from 3.2% to 5.1% across periods; Whites exhibit the highest probability, 9.3% in the 1980s which increased to 12.1% by the late 2000s; while the probability for African Americans fell in between, increasing from 5% to 8.5% across periods. The proportional growth in the predicted probabilities was greater among African Americans, 69%, and Hispanics, 57%, although from a lower starting point than Whites, whose probability grew only by 30%. As seen in Table 4.2, the odds ratios exhibit a small decline – odds ratios approaching to 1 indicate a reduction in the difference between the two groups being compared – and although this decline was not statistically significant over time, it is congruent with a converging pattern. In 2011–2013 (Table 4.3), the racial gap appears to have been further reduced. In sum, net of other factors, the racial/ethnic differences in childlessness have not increased nor become substantially divergent as they have for early and nonmarital childbearing (Ellwood and Jencks 2004; McLanahan 2009). The differences in the probability of remaining childless by educational attainment have also decreased over time, net of other factors. Overall, as education increases, so does the probability of remaining childless, as observed in Panel B of Figure 4.2. In 1982–1988, the differences in the probability of remaining childless among 83 individuals without college degrees were not statistically significant, varying from 5.1% for women who did not graduate from high school, to 6.4% for high school graduates, and 8.4% for women with some college education; however, college graduates were 3 times more likely to remain childless than high school graduates 18.4% vs. 6.4% (p < .001). By 2006–2010, the probability of remaining childless increased for low educated women ranging from 6.6% for high school dropouts, to 11.4% for those with some college education, but, net of other factors, the probability of remaining childless for college graduates slightly decreased to 15.4%, reducing the overall educational gap. Given the increased disparities in the returns to education, I would have expected that college graduates would be even more likely to remain childless than less educated women; however, counter to this expectation, the educational differences were instead reduced. Again, in contrast to the increasing divergence by educational attainment in the timing and marital context of childbearing, I did not find increasing differences in the levels of childlessness by education; instead, the results suggest a converging pattern. Nonetheless, the initial results for 2011–2013 suggests that education may be gaining importance again, as the educational gap in the prevalence of childlessness seems to have increased in recent years, in line with the growing economic opportunity costs argument. Future research should confirm this preliminary finding once data collection has been finalized. As in previous studies (Hayford 2013; Lundquist, Budig, and Curtis 2009), I tested for interactions between marital status and educational attainment. However, the interactions were not statistically significant. Thus, I only present the more parsimonious model with main effects only shown in Table 4.2. 84 Let us now focus our attention on the differences by economic well-being assessed by family income quintiles. Most previous studies of childlessness have used education and family background characteristics as proxies for socioeconomic status, and have not directly analyzed differences by income (Bloom and Trussell 1984; Hayford 2013; Livingston and Cohn 2010; Lundquist, Budig, and Curtis 2009; S. P. Martin 2000). I expand existing research by incorporating a measure of income and evaluating its association with childlessness net of education and other sociodemographic factors. To test whether family income was independently associated with childlessness, I ran the models excluding income, and compared them with those presented in in Table 4.2. The models excluding income are presented in Appendix B in Table B.1. In both periods, family income was significantly and independently associated with childlessness. To my surprise, incorporating income into the 1982–1988 model barely changed the coefficients for education, marital status, race/ethnicity, and the other covariates. In 2006–2010, family income slightly reduced the education coefficients by 58% for less than a high school education, and by 23% for college education. In the results for 2011–2013 shown Table 4.3, Model 1 does not include family income, and Model 2 does; although none of the coefficients for income were significant in 2011–2013, including income increased the coefficient for college education by 34%. In sum, although most of these changes were modest, these results suggest that leaving family income out of the model, may lead us to erroneously attribute to other variables, such as education, what is partly an income effect. The last panel of Figure 4.2 shows the predicted probabilities by income quintile based on the models shown in Table 4.2. Overall, more affluent women are more likely to 85 be childless than those in the bottom income quintile. Across periods, we observe a significant increase in the predicted probabilities for women in the top three income quintiles; while the probabilities for women in the two lowest income quintiles remained stable, net of other factors. While most of the differences by income quintile in 1982– 1988 were small and not significant, by 2006–2010 the top three income quintiles became statistically significantly different from the lowest quintile, and a more pronounced gradient emerged. As a sensitivity analysis, I tested a different measure of economic well-being using family income as a proportion of the poverty threshold. These results are presented in Appendix B in Table B.2. 7 The alternative measure consists of five categories: one for those below poverty, and four categories for 100–199%, 200–299%, 300–399%, and 400% or more above the poverty line. Using this measure, the results show an even higher divergence by family income, with a pronounced increase over time in the odds of remaining childless for women in the top income category. The results I present using family income quintiles show a more modest divergence, and thus, are more 7 Data for family income as a percentage of the poverty threshold were not included in the data released by the IFSS for cycle 3 of the NSFG (1982). This variable originally had a high level of missing values; however, the National Center for Health Statistics (NCHS) imputed this variable using a special imputation procedure. After imputation, the values for this variable were comparable to poverty data from the 1982 Current Population Survey, reported in the Current Population Reports (NCHS 1986). I obtained these imputed values directly from the NCHS (Copen 2014). 86 conservative. 8 Overall, the results show that family income is positively associated with childlessness independently of marital status, education, race/ethnicity, and other factors. According to these analyses, income, measured either way, is the only dimension of socioeconomic inequality that grew divergent among childless women between the 1980s and early 2000s, and in this sense, it is in line with the literature on the divergence on childbearing behaviors (Edin and Kefalas 2005; Ellwood and Jencks 2004; McLanahan 2009). I argue that income may be gaining importance as a dimension of stratification in fertility behaviors; differentiating not only mothers but also those who ultimately remain childless. Future research with new data should test this finding and evaluate new trends in childlessness along income lines. 4.5. Conclusions Childlessness among women ending their reproductive lives has become a more common phenomenon in recent decades. Because fecundity decreases rapidly after age 40, the vast majority of women who have not had children by this age are expected to become permanently childless. For some women, childlessness may be an involuntarily outcome resulting from age-related infertility due to the continuous postponement of childbearing; 8 Because the 2006–2010 period includes a time of economic downturn, I tested a dummy variable for the years 2007–2008 to evaluate whether the economic recession affected the findings. However, not only this variable was not significant, but also the other coefficients did not substantially change. I also tested a dummy variable for the years 2008–2009 and found no significant effects. 87 while for others it may be a voluntary choice to facilitate the pursue of education, career, and other interests (Abma and Martinez 2006; Chandra, Copen, and Stephen 2013; Heaton, Jacobson, and Holland 1999; McQuillan et al. 2003). Some women who initially intended to have children, but did not for different reasons, become accustomed to their child-free lifestyles over the life course and decide to remain childless (Heaton, Jacobson, and Holland 1999; McQuillan et al. 2012). Despite the higher prevalence of childless women, relatively few studies have analyzed the trends in childlessness and the extent to which their sociodemographic correlates have changed over time, and, to my knowledge, none of the recent studies has directly compared the trends in the patterns of childlessness with the trends of childbearing. This study fills this gap by analyzing these changes and comparing the converging trend in childlessness to the diverging trends in childbearing, using data from two recent cohorts, one who turned age 40–44, approaching the end of their reproductive span during 1982–1988, and a more recent cohort who reached ages 40–44 during 2006– 2010. I also present preliminary results for the 2011–2013 data release, part of the 2011– 2015 wave, which as of the writing of this dissertation has not been released yet. This study also incorporates an analysis of the differences by income levels, showing its importance as a dimension of inequality, net of education and other sociodemographic factors. Given the growth in economic inequality, which disproportionally increases the opportunity costs for more advantaged individuals, we would have expected growing disparities in the prevalence of childlessness by demographic and socioeconomic characteristics. However, contrary to this expectation, I found that childlessness has 88 instead increased faster among disadvantaged women. Although significant differences remain; I found evidence of a converging trend in the prevalence of childlessness by marital status, education, and race/ethnicity from the 1980s until the early 2000s. The preliminary results for 2011–2013 suggest that the levels of childlessness may have continued converging along marital status and racial/ethnic lines. The converging levels of childlessness by marital status is in great part a result of the increase in nonmarital births and the decrease in marital fertility. It is worth noting that marital status is still the major predictor of childlessness. As Hayford (2013) puts it “marriage still matters”; however, its power to predict childlessness has decreased over time. These findings support the thesis of the decoupling of marriage and childbearing; not only are unmarried women more likely to enter motherhood than in the past, but also married women have become more likely to remain childless, indicating that marriage has lost some of its centrality as the social milieu for reproduction. The convergence in the levels of childlessness by education suggests that education has become less of a differentiating factor in women’s likelihood of remaining childless. This study does not address the mechanisms that have led to this convergence. Some scholars have argued that although college educated women are more likely to be postponers, they are more likely to realize their fertility intentions at older ages than less educated women (Heaton, Jacobson, and Holland 1999; S. P. Martin 2000). This could also be the result of offsetting effects; education increases the opportunity costs of childbearing, but also increases the probability of marriage and, thus, of entering motherhood (Musick, Brand, and Davis 2012; Schwartz and Mare 2012). These opposing forces could explain the weaker relationship between education and childlessness in 89 recent times. However, it is also possible, as other scholars have suggested, that as education has expanded and become less selective of individual socioeconomic characteristics, it has lost some of its differentiating effect for certain behaviors (Musick, Brand, and Davis 2012). Because college educated women now come from more heterogeneous socioeconomic and demographic backgrounds, having a college education may not have the same meaning it had before as a stratifying factor among childless women. It is worth noting that the results for the preliminary data release for 2011–2013 indicates that an educational gradient may be emerging again. The results by race/ethnicity indicate that White women are still more likely to remain childless than African American and Hispanic women; however these differences have not grown divergent over time, instead my results show slightly decreasing odds ratios, which are more compatible with the convergence argument. More recently, childlessness among African Americans has been linked to their low marriage rates, driven by a reduced pool of “marriageable men” (Lundquist, Budig, and Curtis 2009; Marsh et al. 2007). Some scholars suggest that African American professional women are choosing to stay single and childless as a strategy to achieve or retain middle class status (Marsh et al. 2007). The growing similarities found by education and the small disparities by race/ethnicity suggest a declining significance of education and race in the probability of remaining childless. As some scholars have suggested, it is possible that women of different educational and racial/ethnic backgrounds may now be more similarly pressed by the increasing time demands of work and the practical difficulties of combining work and family (S. P. Martin 2000). 90 My findings also suggest that the socioeconomic factors affecting the probability of remaining childless may not be completely captured by education and family background characteristics, a more thorough picture can be portrayed by incorporating family income, which exhibits a different relationship with the levels of childlessness. Differences by income rarely have been analyzed in previous research on fertility; but, as my study indicates, net of education and other factors, the disparity in the rates of childlessness by income increased from the 1980s until the late 2000s. The results of the 2006–2010 wave show that although women are increasingly postponing childbearing and remaining childless, a higher proportion of affluent women are doing so, differentiating themselves from their less affluent counterparts. This result could be driven by the larger growth in the incomes of individuals at the higher end of the earnings distribution (Autor, Katz, and Kearney 2008), which may have disproportionally increased the economic opportunity costs of childbearing for the more affluent. However, this finding could also be the result of changes in population composition if the rise in income inequality significantly reduced the proportion of mothers in the top income quintile. Nonetheless, these results indicate a growing importance of economic factors as a dimension of the stratification in fertility behaviors, and urges scholars to consider this dimension in addition to education, family background characteristics, and other variables that are more frequently included in studies of fertility. Future research should continue investigating income as a dimension of difference using more recent data, as the coefficients for income in the 2011–2013 wave did not reach statistical significance, indicating that the relationship between income and childlessness may be changing again. 91 This study has some shortcomings. First, the NSFG sample is limited to respondents aged 15 to 44 at the time of screening. Given this limitation, this study misses a few births to women past this age. However, because the probability of having a first birth in this age group is relatively small (J. A. Martin et al. 2015), I expect that this limitation does not significantly bias the estimates. In addition, this analysis focuses only on biological childlessness; it does not consider differences between voluntary and involuntary childlessness, nor does it distinguish childless women who have adopted children or who are raising step children. It is possible that social mothers may be closer in their characteristics to biological mothers than to child-free women (McQuillan et al. 2003). This issue remains a promising area for future research. Another limitation is that I was not able to identify women who have cohabited but have never married given that data on cohabitation is only available as of the 1988 cycle. As the previous literature shows, cohabitation has gained importance as a milieu for childbearing (Kennedy and Bumpass 2008; Manning 2001). Future research should evaluate how childlessness has varied among women who cohabit separately from single and married women. Finally, this study does not take into consideration changes in the population composition, such as the decreasing proportion of women marrying, increases in women’s education, or the growing proportion of ethnic minorities. For an account on changes in childlessness due to changes in population composition until 2002, see Hayford (2013). For example, the convergence in educational attainment among childless women could be the result of an overall increase in women’s education, which reduces the pool of low-educated women, decreasing the denominator, thus, even if the number of childless women among the low-educated remains the same, they would represent a 92 higher proportion; alternatively, this result could be due to a real increase in the number of low-educated women remaining childless. Moreover, the growth in income inequality among childless women could be due not to increases in childlessness among better off women, but to a reduced pool of mothers in the high income quintile. The descriptive analysis does not indicate a significant decline in the proportions of mothers in the top income quintile. To better assess the effect of changes in the income distribution, a decomposition analysis would be necessary. 93 Chapter 5 Racial Variations in the Effect of Fertility on Women’s Employment: Declining or Enduring Effects? 5.1. Introduction Women’s labor force participation in the U.S. substantially increased during the last half of the 20 th century. Most of the growth in women’s employment rates was primarily driven by the increase in the number of mothers who remained in or returned to the labor force. In 1970 only 10.8% of married women with a child under the age of 6 were in the labor force, by 1996 this percentage increased to 62.7% (Jacobsen 1999; U.S. Department of Labor 2014). Despite gains towards gender equity in labor market institutions, women’s labor force experiences continue being greatly affected by the gendered nature of family life (Bianchi 2000; Jacobs and Gerson 2004). Women still bear a significantly higher share of childrearing and household work that competes with their time investments in paid labor, leaving them with constrained decisions (Hochschild 1989; Jacobs and Gerson 2004; Sayer, Cohen, and Casper 2005; Stone 2007). The unequal gender distribution of domestic labor coupled with institutional barriers to combining work and family life result in an negative effect of fertility on women’s employment (Budig and England 2001; Glauber 2007; Kahn, García-Manglano, and Bianchi 2014). However, most prior research has evaluated the effect of motherhood on labor force 94 participation while women are still in their childbearing years (Budig 2003; Hynes and Clarkberg 2005; Reid 2002). But, does this effect persist once women end their reproductive lives? How does it vary over the life course? Moreover, how do the experiences of racial minority women differ from those of Whites? Little research has attempted to answer these questions, and they are the focus of this paper. Prior research in this topic has had some limitations. A recent study found evidence indicating that the motherhood penalty attenuates and may even dissipate as women reach their 40s and 50s (Kahn, García-Manglano, and Bianchi 2014). However, this study used a gross measure of employment: labor force participation on the week prior to the survey. Would the effect of motherhood be different if we distinguish between full-time and part-time employment? Because women often resort to part-time employment when they have children (Budig 2003; Hynes and Clarkberg 2005; Klerman and Leibowitz 1999; Laughlin 2011), we have good reasons to expect that the effect of fertility will vary by employment type as women age. In addition, would we find similar results on the effect of motherhood if we use a more enduring measure of women’s involvement in paid labor? None of the recent research has tested these hypotheses. Drawing from the life course perspective and the intersectionality framework, this study fills these gaps using data for White, Hispanic and Black women from the National Longitudinal Survey of Youth (NLSY) 1979–2012, to assess the effect of fertility on two different measures of employment (1) labor force participation, a temporal measure, and (2) cumulative work experience, a long-term measure, further disaggregating the analyses by full-time and part-time employment. By contrast to most prior research, this study analyzes women’s employment over time, following women from their early 20s until 95 their late 40s and early 50s, when they have virtually ended their reproductive lives, providing a more holistic understanding of the different ways that fertility affects employment across racial/ethnic groups. Assessing the effect of fertility on women’s employment across racial/ethnic groups is important because it has major implications for women’s and children’s economic well-being, shaping broader patterns of social inequality (McCall 2001). Beyond wages, employment confers many benefits for most individuals, including a sense of purpose and satisfaction, improved family stability, as well as health, retirement, and other cumulative benefits (Bianchi and Milkie 2010; Sayer, Cohen, and Casper 2005). Therefore, motherhood can have immediate consequences as well as long-lasting effects, as individuals who work less accrue less wealth and fewer old-age benefits over their life time, which can be detrimental to their economic well-being as they enter retirement ages (Budig and England 2001; Hynes and Clarkberg 2005; Staff and Mortimer 2012). By analyzing the intersectionality of gender and race over time, the results of this study can shed light not only on the system of privilege and disadvantage that hinders equality in labor market outcomes, but also on how it contributes to economic hardship and inequality among women at advanced ages. 5.2. Background Feminist scholars have long argued that parenthood is a gendered experience that affects social organization, and in particular, individuals’ relationship with the labor market (England 2005; Glauber 2007; Ridgeway and Correll 2004a). For women, having children implies less time for work outside the home. Despite gains towards gender 96 equality, women still bear a higher share of childrearing and household work that competes with their time investments in paid labor (Bianchi et al. 2012; Hochschild 1989; Jacobs and Gerson 2004; Sayer, Cohen, and Casper 2005; Sayer 2005; Stone 2007). By contrast to fatherhood, motherhood elicits the role of primary caregiver, activating gender stereotypes that conflict with cultural expectations of ideal worker, leaving mothers at a disadvantaged position in the labor market (Glauber 2007; Ridgeway and Correll 2004a, 2004b). Because mothers are expected to give priority to their children, motherhood also signals lower productivity and commitment to work for employers (Browne and Kennelly 1999; Hays 1998). Prior research has found evidence indicating that motherhood biases evaluations of performance and competence in the workplace (Benard and Correll 2010; Correll, Benard, and Paik 2007; Ridgeway and Correll 2004b). Some scholars have argued that when a status is particularly salient, it affects not only other’s evaluations, but also self-evaluations and behaviors, sometimes becoming a self-fulfilling prophecy (Ridgeway and Correll 2004b). Cultural expectations coupled with institutional barriers to combining work and family result in a negative effect of fertility on women’s employment (Budig 2003; England, Garcia-Beaulieu, and Ross 2004; Hynes and Clarkberg 2005). 5.2.1. Gender and Race in the Labor Market However, gender is also intertwined with race, and their confluence forms a new dimension of social organization (Baca Zinn and Dill 1996; Browne and Misra 2003; McCall 2005). For example, prior research indicates that the stereotype of women as less competent than men in the workplace better suits the experiences of Whites; among 97 African Americans, women are instead seen as more competent than their male counterparts (Ridgeway and Correll 2004b). Feminists scholars have claimed that minority women have different family and labor market experiences that, until recently, have been greatly ignored in social discourse and academic research (Collins 1990; Crenshaw 1989; McCall 2005; Ridgeway and Correll 2004a). Research in this area has indicated that the factors that lead women to get involved in the labor force, as well as the meaning of employment and motherhood, vary across race and class. Given historical conditions of slavery, Black women have always been workers and have had higher employment rates than Whites. Thus, whereas for middle-class White women, who have historically been dependent of her parents or husband, employment may signify independence, for Black women labor has been a source of oppression that later became a necessity (Harnois 2005). However, in the last decades this relationship has reversed; White women now exhibit higher employment rates than Black and Hispanic women (England, Garcia-Beaulieu, and Ross 2004; Jacobs and Gerson 2004; Reid 2002). Motherhood also has different meanings that elicit distinct expectations across racial groups. Whereas for Whites and presumably Hispanics, good mothers are expected to stay home to care for her children, for Blacks, good mothers are also expected to be economic providers (Harnois 2005). Consequently, White and Hispanic women may be more compelled to withdraw from the labor market when they become mothers, whereas Blacks may be instead pressed to increase their attachment to the labor market. Young Black women are often perceived as single mothers and in need of employment, and, for the same reason, they are sometimes seen as more committed to work (Browne and 98 Kennelly 1999). However, the interaction between race and gender in the labor market may not be static, but change over time. 5.2.2. Intersectionality across Stages of the Life Course Contrary to the static view that sees intersectionality as ubiquitous, i.e. active at all times and in all places, an alternative perspective states that the salience of intersectionality varies depending on the context, being primed under certain circumstances (Browne and Misra 2003; Ridgeway and Correll 2004b). According to the latter perspective, the intersectionality of race and gender is theorized as being situational, becoming active in certain contexts, while remaining latent in others (Browne and Misra 2003; Ridgeway and Correll 2004a, 2004b). In line with this rationale, the life course perspective states that individuals’ lives are affected by different sets of factors at each stage of the life cycle, suggesting that the salience of intersectionality changes as women age (Dannefer 2003; Ridgeway and Correll 2004b). Although the intersection of race and gender has been theorized as fluid and changing, most empirical research has approached it from a fairly static view, providing main effects or aggregate racial differences at one point in time or over a period of time (Glass 1999; McCall 2001, 2005). This study expands current research by using the intersectionality and life course perspectives as both, theoretical frameworks and analytical tools, to investigate the effect of motherhood on women’s employment, analyzing changes in the salience of the intersection of race and gender in the labor market over women’s reproductive lives. We could expect motherhood and the intersection of race and gender to be more salient while women have dependent children, that is, during the 20s and 30s, the prime 99 childbearing years for women in the U.S. (Desai and Waite 1991; Ridgeway and Correll 2004a). However, as women approach the end of their childbearing years, children get older and become less dependent on mothers, making motherhood a less salient status in the workplace. Based on this premise, many studies have analyzed the impact of fertility on women’s employment during young adulthood (Budig 2003; Desai and Waite 1991; Hynes and Clarkberg 2005; Klerman and Leibowitz 1999). Using data from the early baby-boom cohort, a recent study found evidence indicating that the deterrent effect of motherhood on women’s employment is stronger at younger ages, but attenuates as women transition to later stages in their lives (Kahn, García-Manglano, and Bianchi 2014; Moen 1991). Yet, how the intersectionality of race and gender in the labor market occurs across stages of the life course, or whether it becomes less salient as women reach the end of their reproductive lives have received little attention. Using data from the late baby-boom period, this study addresses these gaps, investigating how the effect of children on women’s employment changes across stages of the life cycle, whether the intersection of race and gender is more salient during the prime childbearing years, and whether it persists once women reach their late 40s and 50s. Given that motherhood has different meanings for racial minority women, we could expect different trajectories in the effect of motherhood on employment across racial/ethnic groups. White and Hispanic mothers have traditionally depended economically from their husbands, and thus have been seen as secondary earners (Browne and Misra 2003; Ridgeway and Correll 2004a). Recent evidence indicates that White mothers are more likely to have financial support from their husbands and families (Damaske 2011). Although greater support may facilitate White women’s continued 100 employment, it may, at the same time, reduce their need to participate in the labor market during childbearing years. The same argument applies to Hispanic mothers, although to a lesser extent, given that Latino men have lower earnings potential than White men. By contrast, motherhood among young Black women often means single-motherhood. Although single motherhood implies more family responsibilities and greater potential for job absenteeism, it also implies a higher need for employment and, for some employers, a greater commitment to work (Browne and Kennelly 1999; Ridgeway and Correll 2004a). Moreover, because the cost of children increases as they get older, mothers, regardless of race, may be more compelled to work for pay as they end their reproductive lives to help cover children’s expenses. This mechanism could make the intersectionality of gender and race less relevant at older ages. Based on these premises, I expect the impact of motherhood and the salience of race to be stronger during young adulthood, but weaker as women transition to later stages in the life course. 5.3. Measuring Employment: Temporal and Long-Term Measures Most studies analyzing the impact of motherhood have measured employment using labor force participation (LFP) (Budig, 2003; Hynes and Clarkberg 2005; Kahn, García- Manglano, and Bianchi 2014; Moen 1991). Although LFP is an important outcome, it is highly volatile. Prior research has found considerable movement into and out of employment among mothers, particularly surrounding the time of childbirth (Budig 2003; Hynes and Clarkberg 2005; Jacobs and Gerson 2004; Klerman and Leibowitz 1999; Laughlin 2011). By contrast, work experience is cumulative in nature and, thus, represents a more enduring measure of employment, providing a better assessment of 101 continuity in individuals’ involvement in paid labor over the life course. Yet, work experience has often been used as a control variable, but has rarely been the focus of recent research (Budig and England 2001; Budig 2003; Kahn, García-Manglano, and Bianchi 2014; Staff and Mortimer 2012). Consequently, little attention has been paid to whether the racial gaps in cumulative work experience widen or narrow over the life course. A prior study found that mothers work fewer weeks per year than nonmothers, most notably White and Hispanic mothers (England, Garcia-Beaulieu, and Ross 2004). However, because this study was based on cross-sectional data, it provided average racial differences, but could not assess whether the gaps widen or narrow over time. The advantage/disadvantage perspective posits that disadvantages experienced early in life accumulate and generate greater disadvantages at later stages, thus implying widening gaps over time (Dannefer 2003). Do mothers in recent cohorts catch up with nonmothers in work experience by the time they end their reproductive lives or do these gaps widen over time? Furthermore, does the gap in work experience between mothers and nonmothers evolve differently among minority women? This study addresses these questions. By using two different measures of employment, and distinguishing between part-time and full-time work, this study provides a better understanding of the temporal and long-term effects of fertility and the relevance of the intersectionality of gender and race in the labor market. 5.4. Contribution to the Literature This study contributes to the literature in several ways. First, it provides new empirical evidence of the temporal and long-term effects of fertility on women’s employment using 102 recent data for women born in the late-baby boom period throughout their reproductive span. Second, it evaluates the extent to which the intersectionality between race, gender, and age affects women’s labor market outcomes, and demonstrates that the intersectionality of gender and race is dynamic, depending on the outcome analyzed, and evolving over the life course. As the evidence of this study shows, gender and race interact in different ways over women’s life cycle, having at times offsetting effects. Researchers should be cautioned that averaging effects over time may obscure these variations, making the salience of intersectionality go undetected. I argue that age is not just another dimension of difference, but by placing individuals in different statuses over the life course, age exposes individuals to varying, compound sources of advantage and disadvantage. This study advances gender theory by providing evidence showing that the intersectionality of race and gender in mother’s labor force participation is more relevant during prime childbearing years, but become less salient at later stages; however, the effect of fertility and the intersection of race and gender in cumulative work experience remains salient even as women end their reproductive lives. The results indicate that motherhood differently shapes inequality among women over the life course. 5.5. Data, Measures, and Methods 5.5.1. Data As described in Chapter 2, for this section I use data for female respondents from the National Longitudinal Survey of Youth, NLSY79. As a reminder NLSY79 respondents were between the ages of 14 and 22 in 1979. By 2012, respondents were between the ages of 47 to 55, an age when most women have ended their reproductive lives. Because 103 this study approaches stages in the life course using age decades, the analysis begins with the age interval 20-29 and ends with 50-55, as the oldest respondent was age 55 in the last survey period. Consequently, the analysis is restricted to ages 20-55. This age period provides representative data for women who are in their prime childbearing years and participating in the labor force at relatively high rates. All multivariate analyses use fixed-effects models, which require valid data for at least two periods for each individual, thus, 46 respondents who did not meet this requirement were dropped from the study. The final sample consists of 4,880 women, including 1,463 Black, 966 Hispanic or Hispanic, and 2,451 non-Hispanic, non-Black women, who are primarily White. Together, they contribute to 93,525 person years of data analysis. 5.5.2. Measures The two dependent variables analyzed in this study are labor force participation and cumulative years of work experience, both measured repeatedly over time. Labor force participation is measured with a dichotomous variable that takes the value of 1 if the respondent was working for pay a positive number of hours during the calendar year prior to the survey. Full-time employment is a dichotomous variable that takes the value of 1 if the respondent was working 35 hours or more per week, whereas part-time employment takes the value of 1 if the respondent was working a positive number of hours but fewer than 35 per week. In each case, the reference category is not working. Years of work experience is a constructed variable based on the cumulative number of weeks respondents worked for pay since the last time they were interviewed divided by 52. This 104 variable is further subdivided into cumulative years of full-time and part-time work experience. Motherhood is measured by parity, the cumulative number of children respondents reported at each interview, and is divided into four categories for zero (referent), one, two, and three or more children. Women’s age is a continuous time- varying covariate. Age is introduced in the models in 10-years intervals that mirror decades in the life course. This strategy also allows for non-linearities in the effect of age at each decade. The intervals include ages 20–29 (reference category), 30–39, 40–49, and 50–55, the oldest possible age interval in 2012. I include interactions between parity and age decades to allow the effect of number of children to vary as women transition to different stages in the life course. Race/ethnicity is coded into two mutually exclusive dummy variables for Blacks and Hispanics, the referent category is Whites. The dummy variables for Blacks and Hispanics are interacted with age decades to assess how racial differences evolve over time. Control variables include women’s education measured in four categories, including less than high school (omitted category), high school, some college, and college education. A dummy variable also indicates whether the respondent was enrolled in school at the time of the survey. The model predicting years of work experience controls for the hourly wages of the respondents' most current job at each survey year. Wages are adjusted for inflation using the Consumer Price Index (CPI) reported by the U.S. Bureau of Labor Statistics and expressed in 2012 real values. To redress skewness in the wage distribution, I use transformed natural logarithm values. 105 Marital status is measured with a dummy variable indicating whether the respondent was married at the time of each survey. I control for economic resources available to women using a variable for net family income, measured by family income minus the respondent’s own earnings, if any. This measure captures not only husband’s earnings, but also other sources of income in the family that may influence women’s decision to work for pay. Net family income is also adjusted for inflation using the CPI and expressed in 1,000s of dollars of 2012 real values. The presence of preschoolers is measured with a time-varying dichotomous variable indicating whether the youngest child in the home is younger than six. The models predicting cumulative years of work experience also include controls for women’s age at first birth, measured with two dichotomous variables, one for having had a first birth by age 24, and the other one, for first births occurring between ages 25–29. The referent category is not having had any children by age 29. 5.5.3. Methods I approach this research from a longitudinal perspective, using multi-level fixed-effects models to examine individual’s change in labor force participation and cumulative years of work experience from age 20 to 55. Level 1 comprises change over time measured by calendar years, whereas Level 2 comprises the individuals. Fixed-effects models investigate individual variation in the outcome variable as a function of within-individual changes over time in the independent variables. Fixed-effects models implicitly estimate individual’s fixed characteristics, making each woman serve as her own control. By doing so, these models control for unobserved heterogeneity or omitted-variable bias 106 derived from time-invariant individual-level characteristics, such as individual’s motivation causing selectivity into motherhood and paid labor. Consequently, fixed- effects models allow for a better estimation of causal effects of motherhood on employment (Rabe-Hesketh and Skrondal 2012). One limitation of fixed-effects models is that they do not control for time-varying unobserved individual’s characteristics, thus the results of this study may still be biased by this source of heterogeneity. Another limitation of fixed effects is that the effects of covariates that do not change over time, such as, family background, ability, or fertility expectations at young ages, are dropped from the model and, although they are implicitly taken into account, they cannot be estimated. Nonetheless, the advantage of fixed effects is that it provides consistent estimates of within-individual effects or changes over time controlling for all observed and unobserved individual characteristics that do not vary over time (Rabe-Hesketh and Skrondal 2012). The models predicting labor force participation use conditional fixed- effects logistic regression because the outcome is a dichotomous variable, whereas the models predicting cumulative years of work experience use ordinary least squares (OLS) fixed-effects regression, because years is a continuous variable. All analyses are weighted. To avoid artificially inflating the standard errors, I use transformed weights that average to 1. 5.6. Results 5.6.1. Descriptive Results Table 5.1 shows weighted means and percentages describing the analytic sample. The descriptive results show increasing rates of full-time employment from the 20s through 107 All women Childless 1 Child 2 Children 3+ Children Number of persons (unweighted sample size) 4,880 864 815 1,625 1,576 Number of person-years 93,525 30,904 18,882 24,239 19,500 Total years of work experience by race a All women 22.1 22.9 22.5 23.2 19.9 White 22.6 23.0 22.7 23.6 20.9 Hispanic 19.9 21.3 21.3 21.7 17.4 Black 20.0 22.1 22.1 21.1 17.3 Years of full-time work experience by race a All women 14.0 16.3 15.2 14.3 11.2 White 14.1 16.3 15.1 14.2 11.3 Hispanic 13.1 15.2 14.7 14.3 10.7 Black 13.9 16.4 15.9 14.9 10.9 Years of part-time work experience by race a All women 7.5 6.1 6.8 8.3 8.1 White 8.0 6.2 7.1 8.8 8.9 Hispanic 6.3 5.4 6.0 7.0 6.2 Black 5.7 5.3 5.7 5.8 5.8 Percent of years b Average 21–29 30–39 40–49 50–55 White Employed full time 47.8 43.2 49.6 54.4 54.1 Employed part time 36.6 43.6 33.5 27.9 24.0 Not employed 15.6 13.3 16.9 17.8 21.9 Hispanic Employed full time 44.5 35.5 50.3 55.2 51.4 Employed part time 33.0 41.1 28.7 23.5 20.4 Not employed 22.6 23.5 21.0 21.3 28.3 Black Employed full time 44.4 33.0 51.6 58.2 51.0 Employed part time 31.5 40.9 27.0 19.8 20.7 Not employed 24.1 26.2 21.5 22.1 28.3 b Percentage of years respondents spent in each employment status. Table 5.1. Weighted means and percentages of employment measures from ages 20 through 55 for women in the sample. National Longitudinal Survey of Youth (NLSY) 1979–2012. Age a Mean for the last observation available for each respondent. 108 the 50s, and declining part-time employment rates. During ages 50–55, women in the sample were employed full time 54 percent of the time compared with 41 percent during their 20s. White women exhibit a slightly higher average employment rate than Hispanics and Blacks. However, while the rates of part-time employment are greater for Whites at all times, the rates of full-time employment are greater for Whites only during the 20s and 50s; during the 30s and 40s, Black and Hispanic women exhibit slightly higher rates of full-time employment. The descriptive results also indicate that mothers with one or two children do not differ much from childless women in total years of cumulative work experience by the time they were last surveyed. Instead, women with two children exhibit one more year of cumulative work experience than childless women. Nonetheless, disaggregating by cumulative years of full-time and part-time work reveals that part-time work experience is responsible for mothers’ apparent advantage. Women with three or more children accumulate 2 more years of part-time work experience than childless women. Conversely, childless women accumulate 2 more years of full-time work experience than women with two children, and 5 more years of full-time work experience than women with three or more children. White women exhibit a 2-year advantage in total years of work experience over minority women; however, this advantage disappears when we compare only years of experience in full-time work. 5.6.2. Multivariate Results 5.6.2.1. Motherhood and Women’s Labor Force Participation The descriptive results presented above do not adjust for differences in education and other characteristics affecting women’s employment. Table 5.2 shows the adjusted results 109 from fixed-effects logistic regressions predicting labor force participation. The first two columns show the coefficients and odds ratios predicting overall employment, and the next columns disaggregate the results by full-time and part-time employment. In an effort to overcome potential endogeneity, or reversed causality (Budig 2003), parity is lagged by one period in all the models. Table 5.2 shows significant negative effects of the number of children on women’s labor force participation with considerable variations by employment type and over the life course. To better grasp these variations, Figure 5.1 illustrates the net effects of parity as women age, estimated by adding the coefficients for parity and their interactions with age decade. Figure 5.1 (A) shows that children are strongly associated with a reduction in mothers’ odds of being employed during the 20s and 30s, a stage in mothers’ life cycle when children are young and demand intensive physical care; but these associations are weaker during the 40s. By the early 50s, the associations even become positive for women with three of more children. Nonetheless, disaggregating the results by full-time and part-time work shows different patterns across age decades in the association between the number children and women’s employment. As shown in Figure 5.1 (B) and (C), during the 20s and 30s, children significantly decrease women’s odds of full-time and part-time employment. These effects are significantly reduced during the 40s. However, by the early 50s, children are positively associated with full-time employment, and negatively associated with part-time employment, resulting in offsetting effects that cannot be discerned when labor force participation is aggregated. The results indicate that children deter mothers’ employment in full-time work during the prime childbearing years, but encourage full-time employment at later stages of the life cycle. 110 Table 5.2. Fixed-effects logistic regression predicting women's employment vs. not employed. National Longitudinal Survey of Youth (NLSY) 1979–2012: Women, ages 20 through 55. Coefficient Odds Ratio Coefficient Odds Ratio Coefficient Odds Ratio Number of children Childless (reference) --- --- --- --- --- --- 1 Child -1.12 *** 0.33 -1.38 *** 0.25 -0.98 *** 0.37 2 Children -1.53 *** 0.22 -1.70 *** 0.18 -1.41 *** 0.24 3 or more children -1.62 *** 0.20 -1.88 *** 0.15 -1.53 *** 0.22 Age decade 20–29 (reference) 30–39 -0.96 *** 0.38 -0.99 *** 0.37 -1.05 *** 0.35 40–49 -3.51 *** 0.03 -4.35 *** 0.01 -3.25 *** 0.04 50–55 -5.71 *** 0.00 -7.36 *** 0.00 -4.67 *** 0.01 No. children × Age Childless × 20–29 (reference) --- --- --- --- --- --- One child × 30–39 0.05 1.05 -0.06 0.94 0.16 1.18 One child × 40–49 0.68 *** 1.97 0.93 *** 2.54 0.55 *** 1.72 One child × 50–55 1.19 *** 3.28 2.01 *** 7.48 0.00 1.00 Two children × 30–39 0.11 1.12 -0.11 0.90 0.26 * 1.30 Two children × 40–49 0.95 *** 2.58 1.20 *** 3.30 0.82 *** 2.27 Two children × 50–55 1.46 *** 4.31 2.17 *** 8.73 0.73 ** 2.07 Three or more children × 30–39 0.36 *** 1.43 0.37 * 1.44 0.44 *** 1.56 Three or more children × 40–49 1.54 *** 4.65 2.10 *** 8.20 1.26 *** 3.53 Three or more children × 50–55 2.48 *** 11.88 3.61 *** 36.82 1.47 *** 4.33 Race × Age White × 20–29 (reference) --- --- --- --- --- --- Hispanic × 30–39 0.31 ** 1.37 0.68 *** 1.97 0.15 1.16 Hispanic × 40–49 0.46 *** 1.59 0.90 *** 2.46 0.28 1.32 Hispanic × 50–55 0.45 * 1.57 0.74 * 2.10 0.44 1.56 Black × 30–39 0.45 *** 1.56 0.78 *** 2.17 0.32 *** 1.38 Black × 40–49 0.63 *** 1.87 1.04 *** 2.83 0.35 *** 1.42 Black × 50–55 0.53 *** 1.69 0.50 * 1.65 0.70 *** 2.01 Family characteristics Net family income (in 1,000s of 2012 dollars) 0.00 *** 1.00 0.00 *** 1.00 0.00 *** 1.00 Married -0.62 *** 0.54 -0.86 *** 0.42 -0.45 *** 0.64 Married × Hispanic 0.74 *** 2.10 0.85 *** 2.35 0.57 *** 1.77 Married × Black 0.85 *** 2.34 1.11 *** 3.04 0.62 *** 1.85 Youngest child is <6 years old -1.04 *** 0.35 -1.47 *** 0.23 -0.83 *** 0.44 Married × Youngest child <6 -0.10 0.91 -0.21 ** 0.81 -0.04 0.96 Education Less than high school (reference) --- --- --- --- --- --- High school 0.17 * 1.19 0.56 *** 1.74 0.10 1.11 Some college -0.10 0.90 0.51 *** 1.67 -0.23 0.80 College 0.14 1.14 1.74 *** 5.67 -0.38 * 0.69 Enrolled in college -0.51 *** 0.60 -1.14 *** 0.32 -0.32 *** 0.73 Years of part-time work experience 0.31 *** 1.36 0.27 *** 1.31 0.38 *** 1.45 Years of full-time work experience 0.16 *** 1.17 0.26 *** 1.30 0.00 1.00 Number of persons 3,160 2,850 3,136 Number of person-years 61,953 35,852 40,100 Log likelihood -19,616 -9,666 -14,662 *p < .05, **p < .01, ***p < .001 Employed Full-time Part-time 111 *p < .05. **p < .01. ***p < .001. Denote significant main effects. †p < .05. ††p < .01. †††p < .001. Denote significant interactions. A. Employed vs. not employed B. Employed full time vs. not employed C. Employed part time vs. not employed Figure 5.1. Net motherhood effect on women's employment by age -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 20-29 30-39 40-49 50-55 1 Child 2 Children 3+ Children *** *** *** ††† ††† ††† ††† ††† ††† ††† -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 20-29 30-39 40-49 50-55 1 Child 2 Children 3+ Children *** *** *** † ††† ††† ††† ††† ††† ††† -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 20-29 30-39 40-49 50-55 1 Child 2 Children 3+ Children *** *** *** †† ††† ††† ††† ††† † ††† 112 5.6.2.2. Race/Ethnicity and Women’s Labor Force Participation over the Life Course By contrast to the descriptive results showing higher average rates of full-time and part- time employment for Whites, the multivariate analysis in Table 5.2 indicates that, controlling for family and human capital characteristics, Hispanic and Black women have higher odds of being employed than Whites at each age decade. Disaggregating the results by full-time and part-time employment shows that relative to their 20s, Hispanics and Blacks have higher odds of being employed full-time during the 30s and 40s than Whites. The difference between Whites and Blacks is somewhat reduced by the early 50s. This reduction in the odds could be the result of White women’s increasing rates of entrance into full-time employment towards the end of their childbearing years, or Black women’s higher exit rates from full-time employment relative to Whites, as a prior study has suggested (Reid 2002), or a combination of both forces. Conversely, the difference between Whites and Blacks in part-time employment widens into the 50s. While the difference between Whites and Hispanics follows a similar pattern, it does not reach statistical significance. Because being full-time employed implies a stronger commitment to work and is more consequential for individuals’ economic well-being (Budig 2003; Cohen and Bianchi 1999; Percheski 2008; Reid 2002), I ran models separately by race/ethnicity predicting full-time employment to better grasp racial variations in the effect of motherhood. The findings are illustrated in Figure 5.2. See Appendix C, Table C.1 for table with full results. To maximize the sample size for each racial/ethnic group, in these analyses the omitted category consists of not working or working part-time; however, using only not working as the referent category yields similar results.. The results 113 *p < .05. Denotes significant main effects. †p < .05. Denotes significant interactions. Figure 5.2. Net motherhood effect on women's full-time employment by race and age -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 20-29 30-39 40-49 50-55 20-29 30-39 40-49 50-55 20-29 30-39 40-49 50-55 White Hispanic Black 1 Child 2 Children 3+ Children † † † † † † † † * * * * * * † † † † 114 indicate a higher deterrent effect of motherhood on full-time employment for Whites than for Hispanics during their 20s and 30s, but no deterrent effect for Black women; instead, among Black women, having children is associated with higher rates of full-time employment at all age decades. The analyses on full-time employment indicate that the intersectionality of gender and race is more salient during the prime childbearing years, but decreases in relevance during the 40s and early 50s. Although the effects of having fewer than three children do not reach statistical significance among Hispanics, overall the results suggest that during these later stages in women’s life cycle, children encourage full-time employment among mothers, regardless of race. 5.6.2.3. Motherhood and Cumulative Work Experience By contrast to labor force participation, cumulative work experience provides a long-term assessment of involvement in paid labor. Table 5.3 shows the results from fixed-effects models predicting cumulative years of work experience. The first column shows the results for total years of work experience and the next two columns disaggregate the results by full-time and part-time work experience. In these models, parity and marriage are lagged by one period. Figure 5.3 summarizes the effects of the number of children on cumulative years of full-time work across the life course based on the results presented in Table 5.3. Figure 5.3 (A) shows that mothers exhibit 2 to 4 more years of work experience than childless women during the 20s. However, this advantage is reduced to 7 to 10 months during the 30s, and by the 40s and 50s mothers lag behind childless women in years of work experience. The results also show that the association between number of children and work experience is not proportional, but stronger at higher parities. For 115 Total years Full-time years Part-time years Number of children Childless (reference) --- --- --- 1 Child 2.08 *** 1.23 *** 0.84 *** 2 Children 3.17 *** 1.68 *** 1.43 *** 3 or more children 4.09 *** 2.13 *** 1.90 *** Age decade 20–29 (reference) --- --- --- 30–39 6.58 *** 4.93 *** 1.51 *** 40–49 15.62 *** 12.24 *** 3.10 *** 50–55 21.44 *** 16.51 *** 4.55 *** No. children × Age Childless × 20–29 (reference) --- --- --- One child × 30–39 -1.49 *** -1.32 *** -0.15 ** One child × 40–49 -2.58 *** -2.58 *** 0.04 One child × 50–55 -2.72 *** -2.12 *** -0.56 *** Two children × 30–39 -2.37 *** -2.18 *** -0.15 ** Two children × 40–49 -4.09 *** -4.42 *** 0.40 *** Two children × 50–55 -3.91 *** -4.61 *** 0.74 *** Three or more children × 30–39 -3.44 *** -3.31 *** -0.15 * Three or more children × 40–49 -6.30 *** -6.74 *** 0.45 *** Three or more children × 50–55 -7.13 *** -7.77 *** 0.63 *** Race × Age White × 20–29 (reference) --- --- --- Hispanic × 30–39 -0.32 ** 0.05 -0.36 *** Hispanic × 40–49 -0.44 *** 0.56 *** -0.95 *** Hispanic × 50–55 -1.09 *** 0.40 * -1.42 *** Black × 30–39 -0.35 *** 0.07 -0.39 *** Black × 40–49 -0.51 *** 0.63 *** -1.08 *** Black × 50–55 -0.73 *** 0.96 *** -1.62 *** Education Less than high school (reference) --- --- --- High school -0.34 *** -0.77 *** 0.32 *** Some college 1.33 *** 0.33 ** 0.88 *** College 2.93 *** 0.94 *** 1.87 *** Enrolled in college -0.56 *** -0.24 *** -0.31 *** Table 5.3. Fixed-effects regression coefficients predicting years of cumulative work experience. National Longitudinal Survey of Youth (NLSY) 1979–2012: Women, ages 20 through 55. 116 example, during the early 50s having one child is associated with 8 fewer months of work experience, having two children, with 9 fewer months, but having three or more children is associated with 3 fewer years of work experience. Although these differences may not seem large, disaggregating the results by full-time and part-time work shows a stronger disadvantaging effect of children on women’s cumulative years of full-time work experience. As Figure 5.3 (B) shows, the negative association between the number of children and years of full-time work experience becomes evident during the 30s, and grows stronger during the 40s. Uniparous women are the least affected and seem to recoup from some of the loss in full-time work experience by the time they reach their 50s, a stage in Total years Full-time years Part-time years Employment characteristics Hourly wages (natural log) 0.20 *** 0.11 *** 0.08 *** Working part-time 1.78 *** 1.80 *** 0.06 Working full-time 0.67 *** 0.07 0.67 *** Family characteristics Net family income (in 1,000s) 0.00 *** 0.00 *** 0.00 * Married 0.99 *** 0.60 *** 0.39 *** Married × Hispanic -0.58 *** -0.55 *** -0.01 Married × Black -0.24 ** -0.04 -0.20 *** Youngest child is <6 years old -0.15 ** 0.17 *** -0.27 *** Married × Youngest child <6 -0.80 *** -0.28 *** -0.52 *** Had first child by age 24 1.31 *** 0.01 1.20 *** Had first child by between 25–29 0.97 *** 0.30 *** 0.63 *** Constant 0.89 *** 0.50 *** 0.33 *** Number of persons 4,880 4,880 4,880 Number of person-years 93,525 93,525 93,525 Log likelihood -232,225 -234,893 -190,556 Table 5.3. Continued… *p < .05, **p < .01, ***p < .001 117 Figure 5.3. Net motherhood effect on women's cumulative years of work experience by age *p < .05. **p < .01. ***p < .001. Denote significant main effects. †p < .05. ††p < .01. †††p < .001. Denote significant interactions. A. Overall years of work experience B. Years of full-time work experience C. Years of part-time work experience -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 20-29 30-39 40-49 50-55 1 Child 2 Children 3+ Children *** ††† *** *** ††† ††† ††† ††† ††† ††† ††† ††† -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 20-29 30-39 40-49 50-55 1 Child 2 Children 3+ Children *** *** *** ††† ††† ††† ††† ††† ††† ††† ††† ††† 0.0 1.0 2.0 3.0 20-29 30-39 40-49 50-55 1 Child 2 Children 3+ Children *** *** *** †† †† † ††† ††† ††† ††† ††† 118 women’s life cycle when children are older and demand less physical care. However, women with two or more children are less able to do so. Overall, the results indicate that motherhood has enduring effects on full-time work experience. Net of other factors, by the early 50s, having one child is associated with nearly 11 fewer months of full-time work experience, however, having two children is associated with nearly 3 fewer years, and having three or more children with 5.6 fewer years of full-time work experience. By contrast, children are associated with more years of part-time work experience at all stages in the life course, as illustrated in Figure 5.3 (C). As these analyses demonstrate, the effects of parity on full-time work experience are larger than the ones found for the number of years of full-time and part-time work experience combined. 5.6.2.4. Cumulative Work Experience and the Persistent Salience of Intersectionality The descriptive results indicated that Hispanic and Black women accumulate fewer years of work experience than Whites by the end of their reproductive lives. The multivariate analyses presented in Table 5.3 showed that this is the result of widening racial gaps in work experience as women age. Net of number of children, marital status, and other factors, by the early 50s, Hispanic and Black women exhibit nearly 1 fewer year of cumulative work experience than Whites. These results are in line with prior literature showing White women’s advantage in employment (England, Garcia-Beaulieu, and Ross 2004; Reid 2002). However, differentiating between full-time and part-time employment reveals that White women’s advantage is mainly due to the accumulation of more years of part-time employment. In fact, although the differences are small, Hispanics and Blacks accumulate more years of experience in full-time employment. By the early 50s, 119 Hispanic women accumulate nearly 5 more months, while Black women accumulate nearly 1 more year of full-time work experience than Whites with similar characteristics. In order to better assess how the effect of motherhood varies by race/ethnicity, I ran the models predicting total and full-time years of work experience separately for Whites, Hispanics and Blacks. Table 5.4 summarizes the effects of motherhood by race/ethnicity, estimated by adding the coefficients for number of children and their interactions with age. The results confirm that the salience of the intersectionality of gender and race varies by age and depends on the specific outcome analyzed. The effect of children on the total number of years of work experience is similar for Whites and Hispanics, for whom having up to two children is associated with a 7 to 10 months Age White Hispanic Black White Hispanic Black 1 Child 20-29 2.1 1.9 1.9 1.3 0.8 1.2 30-39 0.6 0.5 0.1 -0.1 -0.5 0.1 40-49 -0.6 0.1 -0.5 -1.5 -1.4 -0.4 50-55 -0.6 -0.6 -0.9 -0.7 -2.1 -1.1 2 Children 20-29 3.2 3.2 3.0 1.6 1.6 2.0 30-39 0.8 0.8 0.4 -0.5 -0.7 0.0 40-49 -1.0 -0.5 -0.8 -3.0 -2.4 -1.3 50-55 -0.7 -0.8 -1.1 -3.1 -3.4 -1.7 3+ Children 20-29 4.0 4.1 4.2 1.9 2.5 2.7 30-39 0.7 0.5 0.2 -1.3 -1.1 -0.6 40-49 -2.1 -2.2 -2.6 -4.8 -4.2 -3.4 50-55 -2.9 -3.3 -3.9 -5.7 -5.9 -4.8 All years Full-time years Table 5.4. Net motherhood effect on cumulative years of work experience by race a a Based on fixed-effect models fully interacted with race/ethnicity. Controls for education, school enrollment, employment status, wages, marital status, having a preschooler in the household, and age at first birth. All main effects of number of children and its interactions with age decade are significant at p < .001. 120 reduction in work experience by the early 50s, and having three or more children with 3 fewer years of work experience. The effect is slightly larger for Blacks, for whom having two or fewer children is associated with one fewer year of work experience, and having three or more children with 4 fewer years of work experience. Although those effects do not seem large, the disadvantaging effects of having children on the accumulation of years of full-time employment are stronger, particularly among Whites and Hispanics with two or more children, and smaller, but still significant among Black women. These effects accumulate, increasing from the 30s until the 50s. As Table 5.4 shows, by the early 50s, Whites and Hispanics with two children, accumulate over 3 fewer years of full-time work experience than childless women, whereas Black women with two children are disadvantaged only by 1.7 fewer years of experience in full-time work. More importantly, having three or more children is associated with nearly 6 fewer years of full-time work experience for White and Hispanic women, and nearly 5 fewer years for Black women. 5.6.2.5. A Note on Marriage and Race It is also worth noting that marriage is not significantly associated with part-time employment, regardless of race/ethnicity, but its effects on full-time employment vary by race. The results in Table 5.2 show that marriage is associated with lower odds of employment, particularly hindering full-time employment for women with a preschooler in the home, but only among White women. However, marriage does not deter full-time employment among Hispanic or Black women. By contrast, the results using work experience reveal that marriage is associated with more work experience, especially more years of full-time work experience among White and Black women, but it has no effect 121 among Hispanic women. The implications of these findings are discussed in the concluding section. In sum, by contrast to the declining effect of fertility on women’s labor force participation, the results reveal instead enduring effects of fertility on cumulative work experience – a long-term measure of involvement in paid labor – that persist as women reach their early 40s. Nonetheless, in all the measures of employment used, Black women stand out, being differently affected by motherhood in the labor market than White and Hispanic women. Although the analyses of labor force participation revealed disadvantaging effects of motherhood for Whites and Hispanics during the 20s and 30s, but not for Blacks, the results using years of work experience showed that all mothers, regardless of race, accumulate fewer years of work experience by the early 50s. Nevertheless, White and Hispanic women’s involvement in full-time work seems to be more impacted by having children than Black women’s. Yet, the effect of motherhood and the salience of race on this latter outcome remain strong even as women end their reproductive lives. 5.7. Conclusions Although a vast literature has analyzed the effect of motherhood on women’s employment (Budig 2003; Desai and Waite 1991; Hynes and Clarkberg 2005; Klerman and Leibowitz 1999), less attention has been paid to how this effect evolves over the life course, how it varies by race, how the intersectionality of race and gender in this domain changes over time, and whether it remains salient as women end their reproductive lives. This study fills these gaps using fixed-effects models and recent data from the National 122 Longitudinal Survey of Youth (NLSY) 1979–2012, comparing the experiences of White, Hispanic, and Black women from the late baby-boom cohort over their reproductive span, from their early 20s until their early 50s. Assessing these variations is important because it sheds light on how the system of privilege and disadvantage affects women’s employment over the life course. In addition, this study uses two different measures of employment, labor force participation (LFP), which is a temporal measure, and cumulative years of work experience, which is an enduring outcome, further differentiating full-time and part-time employment, thus, providing a more holistic understanding of the effect of motherhood among women from different racial/ethnic backgrounds. The results confirm that motherhood deters women’s LFP, having stronger effects among White and Hispanic women (Budig 2003; England, Garcia-Beaulieu, and Ross 2004; Kahn, García-Manglano, and Bianchi 2014). However, the effect of motherhood and the salience of intersectionality both depend on the measure of employment analyzed. The results using LFP indicate diminishing negative effects of motherhood among Whites and Hispanics during the 20s and 30s, but these effects dissipate and become positive by the time women end their reproductive lives, in line with prior research (Kahn, García-Manglano, and Bianchi 2014). However, among Black women, motherhood is instead associated with greater labor force participation at most stages of the life course. By contrast, using cumulative work experience reveals instead persistent disadvantaging effects of motherhood for all women regardless of race, primarily reducing years of full-time work experience for mothers with two or more children. 123 Although the results confirm a significant intersectionality of gender and race/ethnicity (Baca Zinn and Dill 1996; Ridgeway and Correll 2004a), in this case, it appears to be to White women’s disadvantage. White women’s employment is more greatly impacted by having children than Hispanic’s, but the racial differences in the effect of motherhood attenuate as women age. Conversely, motherhood does not hurt Black women’s labor force participation, but it depresses their cumulative years of work experience. This study contributes to the literature by drawing from the intersectionality and life course perspectives, and using them as theoretical frameworks and analytical tools to elucidate the mechanisms that hinder equality in the labor market, adding age, an often neglected dimension of difference and disadvantage, to the study of motherhood and intersectionality (Moen 1991). The results confirm that the intersection of race and gender is contingent not only on the outcome of study, but also on the stage in the life course. The findings also reveal that motherhood has offsetting effects on full-time and part-time employment towards the end of women’s reproductive lives that cannot be detected when labor force participation is aggregated. The analyses of full-time labor force participation show that the intersectionality of gender and race in the labor market is more relevant during women’s prime childbearing years, but become less salient at older years. But, when the outcome of interest is cumulative years of work experience, the results indicate that the salience of race persists even as women enter their 50s. While these results could suggest greater disadvantages in the labor market for White women, the results could also reflect privileges that White women enjoy more than Hispanics or Blacks, such as more support from their husbands and families that allow White women to reduce their investments in paid labor during their prime childbearing 124 years (Damaske 2011; England, Garcia-Beaulieu, and Ross 2004). By contrast, Hispanic and Black women tend to have husbands who are less supportive at home and have lower earnings power (Damaske 2011; Glauber 2007). Thus, minority women, and Blacks in particular, will be more compelled to work for pay when they become mothers to secure an income, and obtain other benefits, given that these benefits are less likely to be available through their partners. The results suggest that family processes confer different privileges and disadvantages to women from diverse racial backgrounds at different stages of the life course. This research is not without limitations. First, this study does not investigate the reasons for mothers’ withdrawal from employment. Motherhood may not necessarily constitute a source of disadvantage for women who prefer to devote their time to care for their children or other family members, or who withdraw from the labor market to pursue other interests. Nonetheless, prior research has indicated that more often women would prefer to be employed despite family responsibilities, but the practical difficulties of combining work and family leave them with constrained decisions (Budig 2003; Jacobs and Gerson 2004; Percheski 2008; Sayer, Cohen, and Casper 2005; Stone 2007). Another limitation is that this study does not evaluate the effect of policies in the workplace that may moderate the effect of motherhood on employment, such as maternity leave, introduced in 1993 (Klerman and Leibowitz 1999). Unfortunately, in the early waves of the NLSY79, it is not clear whether some women coded as not working were taking unpaid maternity leave, thus, the results could be biased if the effect of motherhood found in this study were altered by treating those cases differently. Similarly, other labor market structural conditions, such as child care, or even discriminatory practices are not 125 considered in this study, but they could also moderate the effect of motherhood on women’s involvement in paid labor (Reid 2002). Despite these limitations, the results of this study provide evidence indicating that the intersectionality of race and gender in the labor market evolves as women transition to later stages of the life course. The results suggest that motherhood still constitutes a mechanism that hinders equality in labor market outcomes by reducing Whites’ and Hispanics’ labor force participation during young adulthood, a period in an individual’s life cycle normally characterized by high growth in wages and fast occupational mobility (Budig and England 2001; Budig 2003; Hynes and Clarkberg 2005). Motherhood also reduces women’s cumulative years of work experience, regardless of race, and these effects are strong even as women end their childbearing years. Work experience is important because it is used by employers to assess individuals’ potential productivity, assign wages, and grant promotions (Budig and England 2001; Kilbourne et al. 1994; Staff and Mortimer 2012). In sum, the results of this study indicate that motherhood still carries important implications for women’s future careers and economic well-being. As women in this cohort transition to later stages in their lives and more data become available, future research should evaluate how these effects evolve as women approach retirement ages. 126 Chapter 6 Conclusion and Discussion 6.1. Summary Fertility patterns in the U.S. radically changed since the 1970s; nonmarital births increased, individuals began increasingly postponing childbearing, while others eschewed parenthood altogether (Casper and Bianchi 2002; Morgan 1996). Fertility behaviors have important consequences for family structure and socioeconomic well-being. Prior research has indicated that social inequality has been exacerbated by fertility patterns (Ellwood and Jencks 2004; McLanahan 2004; Western, Bloome, and Percheski 2008). In this dissertation, I explored the link between fertility and socioeconomic factors in the U.S. using nationally representative data for non-Hispanic Whites, non-Hispanic Blacks, and Hispanics, investigating the trends and patterns of fertility from the 1970s until the late 2000s, and evaluating the consequences of fertility for women’s socioeconomic outcomes. Demographers and policy makers have raised concerns about the growing divergence in childbearing trends. While Whites and better off women are increasingly delaying childbearing after getting married and have fewer children, disadvantaged racial minorities are still having children at relatively young ages, often out-of-wedlock, and bear more children, compounding their disadvantages (Ellwood and Jencks 2004; J. A. 127 Martin et al. 2015; McLanahan 2009). I use data from the National Survey of Family Growth (NSFG) to investigate the trends in maternal age, fertility rates, and the prevalence of childlessness, evaluating racial and educational differences in childbearing behaviors among recent cohorts. The trends presented in Chapter 3 confirm that although all demographic groups are delaying parenthood, low-educated ethnic minorities still enter motherhood at relatively younger ages and once they do, they bear children at higher rates than low-educated Whites. However, in this dissertation, I demonstrate that the divergent trends in maternal age by race/ethnicity mainly occurred during the 1980s and 1990s, but the trends stopped diverging since the late 2000s, although significant differences remain. By 2006–2013, the average age at first birth in the U.S. was 24.7, ranging from 25.7 among Whites, 22.9 among foreign-born Hispanics, 22.5 among native-born Hispanics, to 22.4 among African Americans, with most of the racial differences concentrated among low-educated women. The differences in fertility rates vary from 1.8 for Whites, 2.2 for Blacks, 2.4 for native-born Hispanics, to 2.8 for foreign- born Hispanics, and are greatly explained by levels of education. I evaluate whether the results support each of the current theories explaining racial/ethnic differences in fertility outcomes, including the structural and cultural perspectives, the classic assimilation and segmented assimilation theories, the racial stratification theory, and the minority group status hypothesis. Supporting the structural perspective, which states that racial inequality derives from differences in socioeconomic and demographic characteristics (Musick et al. 2009; Schoen et al. 2009; Wilson 1987), I found that part of the racial differences in fertility behaviors among low-educated women can be explained by sociodemographic factors; moreover, I did not find significant racial 128 differences in the fertility behaviors of college educated women. These findings leave little room for cultural arguments, which pose that norms and pronatalist values drive racial/ethnic inequality (Edin and Kefalas 2005; Forste and Tienda 1996; Hartnett and Parrado 2012; Landale and Oropesa 2007; Wilson and Neckerman 1987). The results are more in line with the racial stratification perspective that places the social structure at the base of racial disparities, but also acknowledges the role of culture and historical conditions constraining the opportunities of disadvantaged ethnic minorities, who have fewer resources to overcome the legacy of historical discrimination (Frank and Heuveline 2005; Parrado and Morgan 2008; Telles and Ortiz 2008). Partly supporting the classic assimilation theory, predicting higher fertility among immigrants that converges to the dominant group behaviors over time (Alba and Nee 1997; Choi 2014; Parrado and Morgan 2008; Parrado 2011), I found that foreign-born Hispanics exhibit higher fertility rates than Whites. However, the fertility behaviors of native-born Hispanics more closely resemble those of Blacks, giving more credibility to the segmented assimilation theory, which criticizes the assumption of assimilation to White normative behaviors, claiming that the context of reception, and place of settlement play important roles shaping immigrant groups behaviors (Portes and Zhou 1993). The minority group status hypothesis received some support. This hypothesis predicts that upwardly mobile racial minorities may restrict their fertility more than Whites, in an effort to overcome barriers for social mobility (Goldscheider and Uhlenberg 1969). Consistent with this hypothesis, the results revealed greater postponements of motherhood among Blacks and native-born Hispanics with a college degree, a proxy for upward social mobility. Nonetheless, once college educated racial 129 minorities enter motherhood, they do not differ from Whites in the number of children they have, probably because fertility rates among college graduates are already well below replacement levels, at 1.6 children per woman. Lastly, I found that age at first birth is still an important factor determining fertility rates among low-educated groups, attesting to the importance of providing opportunities that encourage delaying entrance into motherhood for disadvantaged groups. In Chapter 4, I use data from the National Survey of Family Growth (NSFG), 1982–1988, and 2006–2010 to analyze trends in childlessness among women aged 40– 44, who are at the end of their reproductive lives. Childlessness is a fertility outcome that has received little attention in fertility research, but that has become more popular among recent cohorts, increasing from a low 10% among women born in the 1930s to 16% among those born in the 1960s. Given the growth in wage inequality since the 1970s, which increased the opportunity costs of childbearing for women with higher earnings potential, we would expect a faster growth in childlessness among college educated women. Contrary to this expectation, the results in Chapter 4 shows that childlessness instead increased faster among less educated women and racial minorities, resulting in a convergence in the prevalence of childlessness by race and class. By the late 2000s, White women were still more likely to remain childless by the end of their reproductive lives (17.6%) than Blacks (14.1%) and Hispanics (7.2%), however the racial differences in the probability of doing so have decreased. I found that even though education, race, and marital status are still strong predictors of childbearing behaviors, these factors have become weak predictors of childlessness over time, from the 1980s to the 2000s. Nonetheless, the results also 130 suggest that economic factors may be gaining importance as a dimension of the stratification in fertility behaviors, and urges scholars to consider income in addition to education, family background characteristics, and other variables that are more frequently included in studies of childlessness. I argue that the convergence in childlessness by sociodemographic factors has partly offset the growth in socioeconomic inequality derived from the divergence in childbearing behaviors. In Chapter 5, I explore the consequences of fertility for women’s opportunities for social mobility. Because employment constitutes one of the most important means for social mobility, I analyze the effect of parity, or number of children, on women’s employment. Women constitute nearly half of the U.S. labor force, and the vast majority of them eventually become mothers. In 2012, 70% of mothers participated in the labor force (U.S. Department of Labor 2013). Thus, it is crucial to investigate whether motherhood constitutes a mechanism that inhibits employment opportunities for women in recent cohorts. Using data from the National Longitudinal Survey of Youth 1979-2012, I assess the extent to which motherhood affects women’s employment, using two different measures, (i) labor force participation, a temporal measure of employment, and (ii) cumulative work experience, a long-term measure. In this chapter, I assess the magnitude of the effect of fertility on women’s employment, I evaluate whether this effect persists or declines as women age, and how it varies across racial/ethnic groups. Drawing from the life course and intersectionality perspectives, I followed a cohort of White, Black, and Hispanic women born in the late baby-boom period from their early 20s until their early 50s, covering most of their reproductive span. 131 My findings reveal declining effects of motherhood on labor force participation, but enduring effects on cumulative years of work experience. Having children reduces women’s employment rates during the 20s and 30s, when children are young and require intense physical attention, but encourages employment by the end of the reproductive years, presumably as children grow up and demand more economic resources. My research also shows that the effect of fertility is greater for White women during their prime childbearing years, and smaller but still significant for Hispanic women, but surprisingly, motherhood does not hurt African American women’s labor force participation. In line with feminist arguments, this result suggests that Black mothers may not enjoy some of the privileges that White and Latina mothers do, such as having the economic support of their husbands and families that may allow them to temporarily withdraw from the labor market when they have children (Collins 1990; McCall 2001). Nonetheless, children are associated with a reduction of work experience for all mothers regardless of race. Although uniparous women are able to recoup from some of the loss in work experience by the time they end their reproductive lives, closing the gap with childless women proves more challenging for women with two or more children. Net of education, family characteristics, and other socioeconomic factors, by the early 50s, having one child is associated with nearly 1 fewer year of work experience in full- time employment, however, having two children reduces work experience in full-time employment by nearly 3 years, and having three or more children by nearly 6 years. I argue that despite progress towards gender equality in the labor market, motherhood still constitutes a mechanism that hinders women’s opportunities for social mobility. 132 6.2. Contributions to Current Research This research contributes to the literature in demography, sociology of the family, social inequality, and stratification by linking lines of research that have previously been addressed separately, including an analysis of motherhood and childlessness, and revealing how changing fertility trajectories affect women’s socioeconomic outcomes, differentially shaping opportunities for social mobility across racial groups. More specifically, I build on prior research by analyzing ethnic differentials in fertility behaviors, using recent nationally representative data that extend until 2013. By contrast to most prior studies, I include data for non-Hispanic Whites, non-Hispanic Blacks, native-born Hispanics, and foreign-born Hispanics, broadening extant research that has primarily focused on explaining Black and White inequality. As Hispanics have become the largest minority group in the U.S., analyzing their childbearing behaviors is important to provide a more holistic account of fertility patterns at a national level. Moreover, I use data for women who are at the end of their reproductive lives, whereas earlier research has analyzed the fertility behaviors of women who are still in their childbearing years. By doing so, this study provides a more accurate depiction of fertility outcomes, and of the long-term effects of motherhood on women’s employment. My analyses also go beyond descriptive statistics, introducing innovative applications of advanced quantitative methodologies, including logistic regression, event history models, Poisson regression, and fixed-effects models. In Chapter 3, I use a variation of Hurdle models to analyze the timing of childbearing and total fertility, using models fully interacted with education. Hurdle models combine logistic regression to predicting zero outcomes, and zero truncated Poisson (ZTP) regression to predict the rate 133 of a positive outcome, conditional on having overcome the zero to one ‘hurdle’ (Long and Freese 2006). Applied to fertility, Hurdle models allow to model separately the probability of remaining childless (a zero outcome), and fertility rates conditional on having entered motherhood. Although Poisson models have been widely used to analyze fertility rates (Brand and Davis 2011; Choi 2014; Telles and Ortiz 2008), Hurdle models and ZTP have not been previously applied in this field. Comparing the results from the Poisson regression to the ZTP regression reveals cases in which racial differences in fertility rates are due to differences in the probability of overcoming the hurdle of being childless rather than the probability of having more children. I innovate by using discrete- time logistic regression, rather than simply logistic regression proposed in Hurdle models, as a strategy to also account for variations in the timing of first births. I include other innovative methodological applications. In Chapter 5, I take advantage of the richness of longitudinal data to apply fixed-effects models to evaluate the impact of fertility on women’s employment over the life cycle. Fixed-effects models have been previously used to assess the causal effects of fertility (Budig and England 2001; Budig 2003; Kahn, García-Manglano, and Bianchi 2014). However, most studies using fixed-effects do not evaluate racial differences because covariates that do not change over time, such as race, are dropped from the model. I advance research by including an interaction of race with age, which is a time-varying covariate, not only to retain the race indicators, but most importantly, to test for significant changes in the effect of parity over the life course across racial groups, a strategy that has not been previously used in this area of research. Moreover, by using two different measures of employment, namely, labor force participation, and cumulative work experience, and 134 including and interaction of parity with age, my research provides new empirical evidence of the temporal and long-term effects of fertility on employment over women’s entire reproductive span. Using this innovative strategy, I was able to demonstrate that the effect of motherhood and the salience of the intersection of race and gender are dynamic, changing over the life cycle, and dependent on the outcome analyzed. Aggregating these effects across racial groups and over the life cycle, as most studies have done, obscure these important variations. In my dissertation, I also explore a wide range of theoretical arguments, evaluating the extent to which current data support the predictions of these theories. By doing so, this dissertation sheds light on the potential causal mechanisms responsible for the outcomes observed, revealing the importance of education and social structure shaping fertility behaviors while also highlighting the persistent salience of race among low educated groups at the turn of the twenty first century. By identifying disadvantaged groups and evaluating potential mechanisms of social inequality, the results of this research can assist policy makers to target vulnerable populations and devise effective policy strategies to aid women and children with limited opportunities for social mobility. 6.3. Limitations This study is not without limitations. A major limitation is the inability to acurately assess culture. Although I used an indicator of religiosity, an important dimension of culture shaping fertility behaviors (Hayford and Morgan 2008), the effect of culture is assessed primarily by the residual racial/ethnic differences after structural factors are taken into account, an approach that has been previously criticized (Forste and Tienda 135 1996; Parrado and Morgan 2008; Telles and Ortiz 2008). Using race as a proxy for culture assumes that culture is fixed. Conversely, several scholars have argued that cultural norms and values are continuously evolving, shaping and being shaped by social interaction with other demographic groups (Forste and Tienda 1996; Frank and Heuveline 2005; Telles and Ortiz 2008). Culture also goes beyond the values and practices pertaining to ethnic groups, including other dimensions such as cultural capital. In this dissertation, structural and cultural explanations have been primarily treated as mutually exclusive in this study; however, cultural factors may develop as a response to structural factors, and structural factors can be constrained by cultural norms. Furthermore, the extent to which education is a measure of structure or culture is subject to debate. This study is not able to assess the changing nature of culture or disentangle these intertwined forces with the data and methods used. Another caveat is that although the NSFG provides a sufficiently large sample of Hispanics to separate them by nativity, given sample size restrictions, I was not able to distinguish groups by country of origin. As I mentioned in Chapter 3, this distinction would have allowed a more appropriate test of the theory of segmented assimilation. Prior research has shown significant behavioral differences by national origin (Alba and Nee 1997; Portes and Zhou 1993; Telles and Ortiz 2008), thus, using Hispanics as an umbrella categorization obscures important subgroup variation. In addition, given data limitations, I was not able to include an analysis of the fertility behaviors of Asians, another important minority group that, by contrast to Hispanics, exhibits the greatest maternal age and the lowest fertility among all racial/ethnic groups (J. A. Martin et al. 2015). Future research should explore the extent to which distinguishing Hispanics by 136 country of origin and including an analysis of the fertility behaviors of Asian groups may enrich our understanding of the current theories that attempt to explain racial inequality. An important drawback is the lack of time-varying data for socioeconomic characteristics in the NSFG, which assesses these characteristics at the time of the survey. Consequently, the analyses that rely on data from the NSFG ignore whether women completed their education before or after becoming mothers. This is an important limitation that I tried to mitigate by using data for women aged 32 and older, most of whom have already completed their education. However, the problem of potential endogeneity remains. Given this limitation, in Chapters 3 and 4, I was only able to evaluate associations and not causal effects. Finally, my analysis of the effects of fertility on women’s employment does not take into consideration the reasons for which women withdraw from the labor market or take part-time jobs. It is possible that part of the effect of fertility revealed in this study is due to other factors, such as, having a family member at home in need of intensive care, or pursuing personal interests. Thus, the readers are cautioned not to interpret the results in Chapter 5 as a motherhood ‘penalty’ per se, or as an indicator of discriminatory practices in the labor market. Moreover, the analyses in Chapter 5 also ignore the influence of policies that may moderate the effect of fertility on women’s employment, such as the availability of childcare, the Family and Medical Leave Act introduced in 1993, and other policies that were enforced after data from the NLSY started being collected. It is possible that the declining effect of motherhood on labor force participation could partly reflect the effect of work policies, or the rise in the availability of childcare services over time. Future research should explore the extent to which these 137 policies moderate the effect of fertility on women’s employment and occupational outcomes. 6.4. Discussion and Implications The results of this research indicate that although race has lost some of its relevance to predict fertility behaviors in recent decades, its influence is still noticeable among the lower classes. However, the fact that there is little variation in age at first birth and fertility rates among college graduates regardless of race speaks to the power of class and structure, more than culture, in shaping fertility outcomes. Moreover, I found evidence that college educated minorities delay entrance into motherhood more often than college educated Whites, presumably as a strategy to achieve or maintain middle-class status as predicted by the minority status hypothesis. But, why are the fertility behaviors of the highly educated similar regardless of race? I have argued that college education not only opens up alternative adult roles to motherhood, but also constitutes a medium of social contagion, spreading norms and values about desirable fertility outcomes. By either mechanism, I have argued that college education homogenizes fertility behaviors across racial/ethnic groups, and thus, constitutes an important tool for policy implementation to reduce racial inequality. Nevertheless, in this research I have assumed that reducing the rates of early childbearing among disadvantaged populations would contribute to ameliorate socioeconomic disadvantage among Blacks and Hispanics. But, not all racial/ethnic groups may equally benefit from the postponement of fertility. In a recent study, Philip Cohen (2016) has argued that delaying childbearing may be more beneficial for White 138 women because child mortality, which follows a U-shape pattern among Whites, reaches the lowest point during the early thirties. Conversely, among Blacks child mortality instead increases monotonically with maternal age. A similar pattern is observed among Mexican women. Thus, delaying childbearing implies an increased risk in child mortality for Black and Mexican women. Based on this evidence, Cohen has argued that early childbearing among Blacks may be an adaptive behavior to counter the effects of deteriorating health conditions as they age (Cohen 2016). I contend that educational and occupational opportunities should be offered in hand with improved health care services to minimize the negative effects of delayed fertility among minority groups. Similarly, in line with the argument of the decoupling of marriage and childbearing (Hayford, Guzzo, and Smock 2014), my results indicate a significant association between marital status and childbearing behaviors, but primarily among the highly educated. Marital and partnership status seems to have a gradational effect; with no significant effects among women with less than high school education, increasing in predictive power at higher levels of education, and having the strongest effect in the childbearing behaviors of college educated women. The results suggest that policies aimed at encouraging marriage among disadvantaged populations, such as the Healthy Marriage Initiative, may have a more limited effect to reduce inequality in childbearing behaviors than policies aimed at improving education and job opportunities among the disadvantaged. To conclude, if motherhood still reduces women’s employment, hindering their socioeconomic outcomes (Budig and England 2001; Budig and Hodges 2010; Kahn et al. 2014), remaining childless may, as Marsh and colleagues (2007) have suggested, 139 constitute an avenue to achieve or maintain middle-class status, at least for some women. My research indicates that fertility has lifelong consequences for women’s lives; by reducing work experience, motherhood lowers life-time savings and accumulated benefits that can adversely impact women’s well-being at older ages. In this sense, childlessness may represent an equalizing mechanism among women from different marital, educational, and racial/ethnic backgrounds. However, the persistent racial and educational differences in the timing and marital context of childbearing work in the opposite direction, stratifying women and their children by these same characteristics. Given the limited socioeconomic resources for disadvantaged families, the prevalence of early and higher fertility among the lower class and racial minorities, implies that disadvantaged women will continue to face a myriad of challenges for social mobility, compounding inequality among the future generations (Brown 2004; McLanahan 2009). I claim that fertility still constitutes a source of social inequality, and that the current policies supporting working mothers, such as paid maternity leave and affordable childcare, are still inadequate. Enhancing these policies will play a crucial role in reducing inequality in the labor market and equalizing the chances for social mobility for the current and future generations. 140 Bibliography Abendroth, Anja-Kristin, Matt. L. Huffman, and Judith Treas. 2014. “The Parity Penalty in Life Course Perspective: Motherhood and Occupational Status in 13 European Countries.” American Sociological Review 79(5): 993–1014. Abma, Joyce C., and Gladys M. Martinez. 2006. “Childlessness among Older Women in the United States: Trends and Profiles.” Journal of Marriage and Family 68(November): 1045–56. Alba, Richard, and Victor Nee. 1997. “Rethinking Assimilation Theory for a New Era of Immigration.” International Migration Review 31(4): 826–74. Anderson, Deborah J., Melissa Binder, and Kate Krause. 2002. “The Motherhood Wage Penalty Revisited: Experience, Heterogeneity, Work Effort, and Work-Schedule Flexibility.” Industrial and Labor Relations Review 56(2). Autor, David H., Lawrence F. Katz, and Melissa S. Kearney. 2008. “Trends in U.S. Wage Inequality: Revising the Revisionists.” Review of Economics and Statistics 90(2): 300–323. Baca Zinn, Maxine, and Bonnie Thornton Dill. 1996. “Theorizing Difference from Multiracial Feminism.” Feminist Studies 22(2): 321–32. Benard, Stephen, and Shelley J. Correll. 2010. “Normative Discrimination and the Motherhood Penalty.” Gender & Society 24(5): 616–46. Bianchi, Suzanne M. 2000. “Maternal Employment and Time with Children: Dramatic Change or Surprising Continuity?” Demography 37(4): 401–14. Bianchi, Suzanne M., and Melissa A. Milkie. 2010. “Work and Family Research in the First Decade of the 21st Century.” Journal of Marriage and Family 72(3): 705–25. Bianchi, Suzanne M., Liana C. Sayer, Melissa A. Milkie, and John P. Robinson. 2012. “Housework: Who Did, Does or Will Do It, and How Much Does It Matter?” Social Forces 91(1): 55–63. Bloom, David E., and James Trussell. 1984. “What Are the Determinants of Delayed Childbearing and Permanent Childlessness in the United States?” Demography 21(4): 591–611. Bongaarts, John. 2001. “Fertility and Reproductive Preferences in Post-Transitional Societies.” Population and Development Review 27(2001): 260–81. 141 Bourdieu, Pierre. 1984. Distinction: a social critique of the judgment of taste Distinction: A Social Critique of the Judgment of Taste. Cambridge: Harvard University Press. ———. 1986. “The Forms of Capital.” In Handbook of Theory and Research for the Sociology of Education, , 241–58. Boyd, Robert L. 1989. “Racial Differences in Childlessness: A Centennial Review.” Sociological Perspectives 32(2): 183–99. Brand, Jennie E, and Dwight Davis. 2011. “The Impact of College Education on Fertility: Evidence for Heterogeneous Effects.” Demography 48(3): 863–87. Browne, Irene, and Ivy Kennelly. 1999. “Stereotypes and Realities: Images of Black Women in the Labor Market.” In Latinas and African American Women at Work: Race, Gender, and Economic Inequality, ed. Irene Browne. New York, NY: Russell Sage Foundation, 302–26. Browne, Irene, and Joya Misra. 2003. “The Intersection of Gender and Race in the Labor Market.” Review Literature And Arts Of The Americas 29(1997): 487–513. Buchmann, Claudia., and Thomas A. DiPrete. 2006. “The Growing Female Advantage in College Completion: The Role of Family Background and Academic Achievement.” American Sociological Review 71(4): 515–41. Budig, Michelle J. 2003. “Are Women’s Employment and Fertility Histories Interdependent? An Examination of Causal Order Using Event History Analysis.” Social Science Research 32(3): 376–401. Budig, Michelle J., and Paula England. 2001. “The Wage Penalty for Motherhood.” American sociological review 66(2): 204–25. Budig, Michelle J., and Melissa J. Hodges. 2010. “Differences in Disadvantage: Variation in the Motherhood Penalty across White Women’s Earnings Distribution.” American Sociological Review 75(5): 705–28. Bureau of Labor Statistics. 2015. “National Longitudinal Survey of Youth, NLSY 1979. Retention & Reasons for Noninterview.” National Longitudinal Surveys, Bureau of Labor Statistics. https://www.nlsinfo.org/content/cohorts/nlsy79/intro-to-the- sample/retention-reasons-noninterview. Calvo, Rocío, and Natalia Sarkisian. 2014. “Racial/ethnic Differences in Post-Migration Education among Adult Immigrants in the USA.” Ethnic and Racial Studies. Casper, Lynne M., and Suzanne M. Bianchi. 2002. Continuity and Change in the American Family. Thousand Oaks, CA: Sage Publications. Chandra, Anjani et al. 2005. Fertility, Family Planning, and Reproductive Health of U.S. Women: Data From the 2002 National Survey of Family Growth. 142 Chandra, Anjani, Casey E. Copen, and Elizabeth Hervey Stephen. 2013. National health statistics reports Infertility and Impaired Fecundity in the United States, 1982-2010: Data from the National Survey of Family Growth. Hyattsville, MD. Cherlin, Andrew J. 2004. “The Deinstitutionalization of American Marriage.” In Journal of Marriage and Family 66: 848–61. ———. 2010. The Marriage-Go-Round: The State of Marriage and the Family in America Today. New York, NY: Vintage Books. Choi, Kate. 2014. “Fertility in the Context of Mexican Migration to the United States.” Demographic Research 30: 703–38. Cleves, Mario, William Gould, Roberto Gutierrez, and Yulia Marchenko. 2010. An Introduction to Survival Analysis Using Stata. Third. College Station, TX: Stata Press. Cohen, Philip N. 2016. “Maternal Age and Infant Mortality for White, Black, and Mexican Mothers in the United States.” Sociological Science 2: 32–38. Cohen, Philip N., and Suzanne M. Bianchi. 1999. “Marriage, Children, and Women’s Employment: What Do We Know?” Monthly Labor Review 122(12): 22–31. Collins, Patricia Hill. 1990. Black Feminist Thought: Knowledge, Consciousness, and the Politics of Empowerment. New York, NY: Routledge. Correll, Shelley J., Stephen Benard, and In Paik. 2007. “Getting a Job: Is There a Motherhood Penalty?” American Journal of Sociology 112(5): 1297–1339. Crenshaw, Kimberlé. 1989. “Demarginalizing the Intersection of Race and Sex: A Black Feminist Critique of Antidiscrimination Doctrine, Feminist Theory and Antiracist Policies.” The University of Chicago Legal Forum: 139–67. Damaske, Sarah. 2011. For the Family?: How Class and Gender Shape Women’s Work. Oxford University Press. Dannefer, Dale. 2003. “Cumulative Advantage/disadvantage and the Life Course: Cross- Fertilizing Age and Social Science Theory.” The journals of gerontology Series B: Psychological sciences and social sciences 58: S327–37. Desai, Sonalde, and Linda J. Waite. 1991. “Women’s Employment During Pregnancy and After the First Birth: Occupational Characteristics and Work Commitment.” American Sociological Review 56(4): 551–66. Dye, Jane Lawler. 2010. “Fertility of American Women: 2008. ” Current Population Reports. Washington, D.C.: U.S. Census Bureau. Dykstra, Pearl A. 2009. “Childless Old Age.” In International Handbook of Population 143 Aging, ed. Peter R. Uhlenberg. Springer Science & Business Media, 671–90. Edin, Kathryn, and Maria Kefalas. 2005. Berkeley, CA: University of California Promises I Can Keep: Why Poor Women Put Motherhood before Marriage. Berkely: University of California Press. Ellwood, David T., and Christopher Jencks. 2004. “The Uneven Spread of Single-Parent Families: What Do Know? Where Do We Look for Answers?” In Social Inequality, ed. Kathryn M. Neckerman. New York, NY: Russell Sage Foundation, 3–77. England, Paula. 2005. “Gender Inequality in Labor Markets: The Role of Motherhood and Segregation.” Social Politics: International Studies in Gender, State & Society 12(2): 264–88. England, Paula, Carmen Garcia-Beaulieu, and Mary Ross. 2004. “Women’s Employment among Blacks, Whites, and Three Groups of Latinas: Do More Privileged Women Have Higher Employment?” Gender & Society 18(4): 494–509. England, Paula, Lawrence L Wu, and Emily Fitzgibbons Shafer. 2013. “Cohort Trends in Premarital First Births: What Role for the Retreat from Marriage?” Demography 50(6): 2075–2104. Forste, Renata, and Marta Tienda. 1996. “What’s Behind Racial and Ethnic Fertility Differentials?” Population and Development Review 22: 109–33. Frank, Reanne, and Patrick Heuveline. 2005. “A Crossover in Mexican and Mexican- American Fertility Rates: Evidence and Explanations for an Emerging Paradox.” Demographic Research 12: 77–104. Frost, Jennifer J, Susheela Singh, and Lawrence B Finer. 2007. “Factors Associated with Contraceptive Use and Nonuse, United States, 2004.” Perspectives on sexual and reproductive health 39(2): 90–99. Furstenberg, Frank F. Jr. 1996. “The Future of Marriage.” American Demographics 18(June): 34–40. Glass, Jennifer L. 1999. “The Tangled Web We Weave: Editorial Introduction to the Special Issue on Ethnicity, Race, and Gender in the Workplace.” Work and Occupations 4(26): 415–21. Glauber, Rebecca. 2007. “Marriage and the Motherhood Wage Penalty among African Americans, Hispanics, and Whites.” Journal of Marriage and Family 69(4): 951–61. Goldscheider, Calvin, and Peter R. Uhlenberg. 1969. “Minority Group Status and Fertility.” American Journal of Sociology 74(4): 361–72. Goldstein, Joshua R., and Catherine T. Kenney. 2001. “Marriage Delayed or Marriage.” American Sociological Review 66: 506–19. 144 Guzzo, Karen Benjamin, and Sarah R. Hayford. 2014. “Fertility and the Stability of Cohabiting Unions: Variation By Intendedness.” Journal of Family Issues 35(4): 547–76. Hagestad, Gunhild O., and Vaughn R. A. Call. 2007. “Pathways to Childlessness: A Life Course Perspective.” Journal of family issues 28(10): 1338–61. Hall, John R. 1992. “The Capital(s) of Cultures: A Nonholistic Approach to Status Situations, Class, Gender, and Ethnicity.” In Cultivating Differences: Symbolic Boundaries and the Making of Inequality, eds. Michèle Lamont and Marcel Fournier. Chicago, IL: University of California Press, 257–88. Harnois, Catherine E. 2005. “Different Paths to Different Feminisms? Bridging Multiracial Feminist Theory and Quantitative Sociological Gender Research.” Gender & Society 19(6): 809–28. Hartnett, Caroline Sten, and Emilio A. Parrado. 2012. “Hispanic Familism Reconsidered: Ethnic Differences in the Perceived Value of Children and Fertility Intentions.” The Sociological Quarterly 53: 636–53. Hayford, Sarah R. 2013. “Marriage (Still) Matters: The Contribution of Demographic Change to Trends in Childlessness in the United States.” Demography 50: 1641–61. Hayford, Sarah R., Karen Benjamin Guzzo, and Pamela J. Smock. 2014. “The Decoupling of Marriage and Parenthood? Trends in the Timing of Marital First Births, 1945-2002.” Journal of Marriage and Family 76(3): 520–38. Hayford, Sarah R., and S. Philip Morgan. 2008. “Religiosity and Fertility in the United States: The Role of Fertility Intentions.” Social Forces 86(3): 1163–88. Hays, Sharon. 1998. The Cultural Contradictions of Motherhood. New Haven, CT: Yale University Press. Heaton, Tim B., Cardell K. Jacobson, and Kimberlee Holland. 1999. “Persistence and Change in Decisions to Remain Childless.” Journal of Marriage and the Family 61(2): 531–39. Hochschild, Arlie. 1989. Viking The Second Shift: Working Parents and the Revolution at Home. New York, NY. Hynes, Kathryn, and Marin Clarkberg. 2005. “Women’s Employment Patterns during Early Parenthood: A Group-Based Trajectory Analysis.” Journal of Marriage and Family 67(1): 222–39. Jacobs, Jerry A., and Kathleen Gerson. 2004. The Time Divide: Work, Family, and Gender Inequality. Cambridge: Harvard University Press. Jacobsen, Joyce P. 1999. “Labor Force Participation.” The Quarterly Review of 145 Economics and Finance 39(5): 597–610. Kahn, Joan R, Javier García-Manglano, and Suzanne M. Bianchi. 2014. “The Motherhood Penalty at Midlife: Long-Term Effects of Children on Women’s Careers.” Journal of Marriage and Family 76(1): 56–72. Kalmijn, Matthijs, and Gerbert Kraaykamp. 1996. “Race, Cultural Capital, and Schooling: An Analysis of Trends in the United States.” Sociology of Education 69: 22–34. Kennedy, Sheela, and Larry Bumpass. 2008. “Cohabitation and Children’s Living Arrangements: New Estimates from the United States.” Demographic Research 19: 1663–92. Kilbourne, Barbara S. et al. 1994. “Returns to Skill, Compensating Differentials, and Gender Bias: Effects of Occupational Characteristics on the Wages of White Women and Men.” American Journal of Sociology 100(3): 689–719. Kirmeyer, Sharon E., and Brady E. Hamilton. 2011. “Transitions between Childlessness and First Birth: Three Generations of U.S. Women.” Washington, D.C.: National Center for Health Statistics, Vital Health Stat 2. Klerman, Jacob Alex, and Arleen Leibowitz. 1999. “Job Continuity among New Mothers.” Demography 36(2): 145–55. Koropeckyj-Cox, T., and V. R. a. Call. 2007. “Characteristics of Older Childless Persons and Parents: Cross-National Comparisons.” Journal of Family Issues 28(10): 1362– 1414. Koropeckyj-Cox, T., and G. Pendell. 2007. “Attitudes About Childlessness in the United States: Correlates of Positive, Neutral, and Negative Responses.” Journal of Family Issues 28(8): 1054–82. Landale, Nancy S., and R. S. Oropesa. 2007. “Hispanic Families: Stability and Change.” Annual Review of Sociology 33: 381–405. Landale, Nancy S., Robert Schoen, and Kimberly Daniels. 2010. “Early Family Formation among White, Black and Mexican American Women.” Journal of family issues 31(4): 445–74. Lareau, Annette. 2003. Unequal Childhoods. Berkeley: University of California Press. Laughlin, Lynda. 2011. “Maternity Leave and Employment Patterns of First-Time Mothers: 1961-2008. ” Current Population Report. Washington, D.C: U.S. Census Bureau. Lesthaeghe, Ron. 1995. “The Second Demographic Transition in Western Countries: An Interpretation.” In Gender and Family Change in Industrialized Countries, eds. 146 Karen Oppenheimer Mason and An-Margrit Jensen. Pp, 17–62. Livingston, Gretchen, and D’Vera Cohn. 2010. “Childlessness up among All Women; down among Women with Advanced Degrees. ” A Social & Demographic Trends Report, Pew Recearch Center. Lloyd, Kim M. 2006. “Latinas’ Transition to First Marriage: An Examination of Four Theoretical Perspectives.” Journal of Marriage and Family 68(4): 993–1014. Long, J. Scott, and Jeremy Freese. 2006. Regression Models for Categorical Dependent Variables Using Stata, Second Edition. College Station, TX: Stata Press. Lundquist, Jennifer H., Michelle J. Budig, and Anna Curtis. 2009. “Race and Childlessness in America, 1988–2002.” Journal of Marriage and Family 71(August): 1988–2002. Manning, Wendy D. 2001. “Childbearing in Cohabiting Unions: Racial and Ethnic Differences.” Family planning perspectives 33(5): 217–23. Marini, Margaret M. 1984. “Women’s Educational Attainment and the Timing of Entry into Parenthood.” American Sociological Review 49(4): 491–511. Marsh, Kris, William A. Darity, Philip N. Cohen, Lynne M. Casper, and Danielle Salters. 2007. “The Emerging Black Middle Class: Single and Living Alone.” Social Forces 86(2): 735–62. Martin, Joyce A., Brady E. Hamilton, Michelle J. K. Osterman, Sally C. Curtin, and T. J. Mathews. 2013. “Births: Final Data for 2012.” National Vital Statistics Reports. Hyattsville, MD: National Center for Health Statistics ———. 2015. “Births: Final Data for 2013.” National Vital Statistics Reports. Hyattsville, MD: National Center for Health Statistics Martin, Steven P. 2000. “Diverging Fertility Among U.S. Women Who Delay Childbearing Past Age 30.” Demography 37(4): 523–33. Martinez, Gladys M., Kimberly Daniels, and Anjani Chandra. 2012. “Fertility of Men and Women Aged 15-44 Years in the United States: National Survey of Family Growth, 2006-2010.” National Health Statistics Reports. Hyattsville, MD: National Center for Health Statistics. McCall, Leslie. 2001. Complex Inequality: Gender, Class, and Race. New York, NY: Routledge. ———. 2005. “The Complexity of Intersectionality.” Signs: Journal of Women in Culture and Society 30(3): 1771–1800. McLanahan, Sara. 2004. “Diverging Destinies: How Children Are Faring under the 147 Second Demographic Transition.” Demography 41(4): 607–27. ———. 2009. “Fragile Families and the Reproduction of Poverty.” The ANNALS of the American Academy of Political and Social Science 621: 111–31. McLanahan, Sara, and Christine Percheski. 2008. “Family Structure and the Reproduction of Inequalities.” Annual Review of Sociology 34: 257–76. McQuillan, Julia et al. 2012. “Does the Reason Matter? Variations in Childlessness Concerns Among U.S. Women.” Journal of Marriage and Family 74(5): 1166–81. McQuillan, Julia, Arthur L. Greil, Lynn White, and Mary Casey Jacob. 2003. “Frustrated Fertility: Infertility and Psychological Distress among Women.” Journal of Marriage and Family 65(November): 1007–18. Moen, Phyllis. 1991. “Transitions in Mid-Life: Women’s Work and Family Roles in the 1970s.” Journal of Marriage and the Family 53(1): 135–50. Montgomery, Mark R., and John B. Casterline. 1996. “Social Learning, Social Influence, and New Models of Fertility.” Population and Development Review 22(SUPPL. 1): 151–75. Morgan, S. Philip. 1996. “Characteristic Features of American Fertility.” Population and Development Review 22(1996): 16–63. ———. 2003. “Is Low Fertility a Twenty-First-Century Demographic Crisis?” Demography 40(4): 589–603. Morgan, S. Philip, and Ronald R. Rindfuss. 1999. “Reexamining the Link of Early Childbearing to Marriage and to Subsequent Fertility.” Demography 36(1): 59–75. Musick, Kelly. 2002. “Planned and Unplanned Childbearing among Unmarried Women.” Journal of Marriage and Family 64(4): 915–29. ———. 2007. “Cohabitation, Nonmarital Childbearing, and the Marriage Process.” Demographic Research 16: 249–86. Musick, Kelly, Jennie E Brand, and Dwight Davis. 2012. “Variation in the Relationship Between Education and Marriage: Marriage Market Mismatch?” Journal of marriage and the family 74(1): 53–69. Musick, Kelly, Paula England, Sarah Edgington, and Nicole Kangas. 2009. “Education Differences in Intended and Unintended Fertility.” Social Forces 88: 543–72. National Center for Health Statistics. 2015. “National Survey of Family Growth.” Public- use data file and documentation. http://www.cdc.gov/nchs/nsfg/nsfg_questionnaires.htm. 148 Oropesa, R. S., and Nancy S. Landale. 2004. “The Future of Marriage and Hispanics.” Journal of Marriage and Family 66(4): 901–20. Parrado, Emilio A. 2011. “How High Is Hispanic/Mexican Fertility in the United States? Immigration and Tempo Considerations.” Demography 48(3): 1059–80. Parrado, Emilio A., and S. Philip Morgan. 2008. “Intergenerational Fertility among Hispanic Women: New Evidence of Immigrant Assimilation.” Demography 45(3): 651–71. Percheski, Christine. 2008. “Opting out? Cohort Differences in Professional Women’s Employment Rates from 1960 to 2005.” American Sociological Review 73(3): 497– 517. Popenoe, David. 1993. “American Family Decline, 1960-1990: A Review and Appraisal.” Journal of Marriage and the Family 55(3): 527–42. Portes, A, and M Zhou. 1993. “The New 2nd-Generation - Segmented Assimilation and Its Variants.” Annals of the American Academy of Political and Social Science 530: 74–96. Quesnel-Vallée, Amélie, and S. Philip Morgan. 2003. “Missing the Target? Correspondence of Fertility Intentions and Behavior in the US.” Population Research and Policy Review 22: 497–525. Rabe-Hesketh, Sophia, and Anders Skrondal. 2012. Multilevel and Longitudinal Modeling Using Stata: Continuous Responses, Third Edition, Volume 1. Stata Press. Reid, Lori L. 2002. “Occupational Segregation, Human Capital, and Motherhood: Black Women’s Higher Exit Rates from Full-Time Employment.” Gender & Society 16(5): 728–47. Ridgeway, Cecilia L., and Shelley J. Correll. 2004a. “Motherhood as a Status Characteristic.” Journal of Social Issues 60(4): 683–700. ———. 2004b. “Unpacking the Gender System: A Theoretical Perspective on Gender Beliefs and Social Relations.” Gender & Society 18(4): 510–31. Rindfuss, Ronald R., S. Philip Morgan, and Kate Offutt. 1996. “Education and the Changing Age Pattern of American Fertility: 1963–1989.” Demography 33(3): 277– 90. Rowland, Donald T. 2007. “Historical Trends in Childlessness.” Journal of Family Issues 28(10): 1311–37. Rumbaut, Rubén G. 2005. “Turning Points in the Transition to Adulthood: Determinants of Educational Attainment, Incarceration, and Early Childbearing among Children of Immigrants.” Ethnic and Racial Studies 28(6): 1041–86. 149 Sandler, Lauren, and Kate Witteman. 2013. “None Is Enough.” Time 182(7): 38–46. Sayer, Liana C. 2005. “Gender, Time and Inequality: Trends in Women’s and Men's Paid Work, Unpaid Work and Free Time.” Social forces 84(1): 285–303. Sayer, Liana C., Philip N. Cohen, and Lynne M. Casper. 2005. “Women, Men, and Work.” In The American People: Census 2000, eds. Reynolds Farley and John Haaga. New York, NY: Russell Sage Foundation, 76–106. Schoen, Robert, Nancy S. Landale, Kimberly Daniels, and Yen-Hsin Alice Cheng. 2009. “Social Background Differences in Early Family Behavior.” Journal of Marriage and Family 71(2): 384–95. Schwartz, Christine R., and Robert D. Mare. 2012. “The Proximate Determinants of Educational Homogamy: The Effects of First Marriage, Marital Dissolution, Remarriage, and Educational Upgrading.” Demography 49: 629–50. Smock, Pamela J., and Fiona Rose Greenland. 2010. “Diversity in Pathways to Parenthood: Patterns, Implications, and Emerging Research Directions.” Journal of Marriage and Family 72(3): 576–93. Staff, Jeremy, and Jeylan T Mortimer. 2012. “Explaining the Motherhood Wage Penalty during the Early Occupational Career.” Demography 49(1): 1–21. Stange, Kevin. 2011. “A Longitudinal Analysis of the Relationship between Fertility Timing and Schooling.” Demography 48(3): 931–56. Stone, Pamela. 2007. Opting Out?: Why Women Really Quit Careers and Head Home. Berkeley: University of California Press. Telles, Edward M., and Vilma Ortiz. 2008. Generations of Exclusion: Mexican- Americans, Assimilation, and Race. New York, NY: Russell Sage Foundation. U.S. Department of Labor. 2013. “Employment Status of the Civilian Noninstitutional Population by Age, Sex, and Race.” Bureau of Labor Statistics. http://www.bls.gov/cps/cpsaat03.htm (December 9, 2014). ———. 2014. “Women’s of Working Age - Recent Facts.” Bureau of Labor Statistics. http://www.dol.gov/wb/stats/stats_data.htm. Western, Bruce, Deirdre Bloome, and Christine Percheski. 2008. “Inequality among American Families with Children, 1975 to 2005.” American Sociological Review 73(6): 903–20. Wilson, William Julius. 1987. 19 The Truly Disadvantaged. Chicago: University of Chicago Press. Wilson, William Julius, and Kathryn M. Neckerman. 1987. “Poverty and Family 150 Structure: The Widening Gap between Evidence and Public Policy Issues.” In Wilson, The Truly Disadvantaged, Chicago, IL: The University of Chicago Press, 63–92. Wu, Lawrence L., and Brian C. Martinson. 1993. “Family Structure and the Risk of a Premarital Birth.” American Sociological Review 58(2): 210–32. Yang, Yang Claire, and S. Philip Morgan. 2003. “How Big Are Educational and Racial Fertility Differentials in the U.S.?” Biodemography and Social Biology 50(3-4): 167–87. 151 Appendix A: Additional Findings for Chapter 3 Findings on Marital and Partnership Status and Fertility The multivariate analyses in Chapter 3 also show a significant effect of marital and partnership status on fertility, however, this effect varies by education. Model 2 in Table 3.3 shows that among women who did not complete high school and high school graduates, single women exhibit a higher risk of becoming mothers than married women, 25% and 39% greater risks respectively, but among college graduates their risk is instead 68% (0.32 - 1 = -0.68) lower, net of other factors. Although, cohabiting women without college degrees exhibit slightly greater risks of becoming mothers than married women, most of these differences are not statistically significant, suggesting that cohabitation has become suitable family context to start childbearing. As expected, previously married childless women, including divorced, separated, or widowed women, have lower risk than married women of becoming mothers, indicating that union dissolution has a negative effect on fertility. In sum, among the low educated, single women are more likely to become mothers than married women, but the opposite is true among college educated women, while cohabitation is more similar to marriage than to singlehood. The results in Table 3.4 indicate that the differences in achieved fertility by marital and partnership status are small, and mostly non-significant among the low educated, but large and mostly significant among the college educated. As expected, single women tend to have fewer children than ever-married women. Nonetheless, some of those differences are explained by the probability of entering motherhood in the first place. For example, among the college educated, Model 1 shows that single women have 82% fewer children (0.19 - 1 = - 0.82, rounding up) than ever-married women, however 152 conditional on having entered motherhood, the difference is reduced and no longer significant in Models 2 through 4. Similarly, among high school dropouts I found no significant difference in achieved fertility between ever-married women and those who have ever cohabited but never married, but among the college educated, women who have ever cohabited but never married have 47% fewer children (0.53 - 1 = - 0.47) than ever-married women, controlling for other factors in Model 4. Overall, these results confirm a severe decoupling of marriage and childbearing among low-educated women. Marriage is a weak predictor of fertility rates for low- educated women, but it has large impact on the reproductive behaviors of women with higher education. In line with prior research, the findings also suggest that cohabitation and marriage have similar cultural meanings as contexts for childbearing among the low educated, but they are more distinct among the more educated (Hayford, Guzzo, and Smock 2014). I conclude that marriage still has a distinctive meaning as a milieu for childbearing but primarily for more educated women. The results also indicate a strong interaction between marital status and education. Relative to married women, single women have a higher risk of becoming mothers among the low-educated, but a lower risk among the college educated. Similarly, marriage is a weak predictor of achieved fertility among the less educated, but has a stronger predictive power among the college educated. Conversely, cohabitation is closer to marriage as a family context for childbearing. In other words, marriage matters for fertility behaviors, but primarily for more educated women. These results provides new evidence of a strong decoupling of marriage and childbearing at low levels of education (Hayford, Guzzo, and Smock 2014). 153 Appendix B: Additional Tables for Chapter 4 Odds Ratio Odds Ratio β p S.E. Exp β β p S.E. Exp β Marital status Ever-married (Ref.) --- --- --- --- --- --- Never-married 3.97 *** 0.37 53.06 2.61 *** 0.35 13.59 Education Less than high school -0.28 0.29 0.76 -0.29 0.60 0.75 High school (Ref.) --- --- --- --- --- --- Some college 0.29 0.29 1.33 0.51 ** 0.18 1.66 College 1.19 *** 0.26 3.29 0.86 *** 0.16 2.37 Race/ethnicity White (Ref.) --- --- --- --- --- --- African American -0.72 † 0.38 0.48 -0.52 *** 0.09 0.60 Hispanic -1.05 * 0.49 0.35 -1.01 *** 0.11 0.36 Family background Mother has some college educ. 0.27 0.19 1.31 -0.12 0.09 0.89 Mother worked 0.02 0.21 1.02 0.38 † 0.23 1.46 Lived with both parents at 14 0.31 0.28 1.36 -0.12 0.27 0.88 Fecundity Issues Reports fecundity problems 1.06 *** 0.15 2.87 1.88 *** 0.16 6.53 Attendance to religious services Never attends religious services -0.22 0.29 0.80 0.92 *** 0.10 2.50 Urbanicity Lived in urban area 0.66 ** 0.22 1.94 0.98 ** 0.24 2.66 Intercept -3.77 0.46 -3.77 *** 0.61 N 1874 1592 Log likelihood -591.0 -662.8 †p < .10; *p < .05; **p < .01; ***p < .001 1982–1988 2006–2010 Table B.1. Estimated coefficients and odds ratios from logistic regression of childlessness on sociodemographic characteristics (excluding family income). NSFG 1982–1988 and 2006–2010: Women aged 40 to 44. 154 Odds Ratio Odds Ratio β p Exp β β p Exp β Marital status Ever-married (Ref.) --- --- --- --- Never-married 4.41 *** 82.34 2.95 *** 19.02 Education Less than high school -0.08 0.93 0.00 1.00 High school (Ref.) --- --- --- --- Some college 0.10 1.11 0.32 1.38 College 0.96 *** 2.62 0.15 1.16 Race/ethnicity White (Ref.) --- --- --- --- African American -0.51 0.60 -0.28 *** 0.76 Hispanic -0.94 0.39 -0.63 *** 0.53 Family income, % of poverty Below poverty (Ref.) --- --- --- --- 100-199% above poverty 0.72 2.05 0.85 2.33 200-299% above poverty 1.58 ** 4.83 1.00 ** 2.72 300-399% above poverty 1.16 * 3.19 1.04 2.83 400%+ above poverty 1.92 *** 6.79 2.82 *** 16.84 Family background Mother has some college educ. 0.29 1.33 -0.04 0.96 Mother worked 0.08 1.08 0.21 1.23 Lived with both parents at 14 0.28 1.32 -0.25 0.78 Fecundity Issues Reports fecundity problems 1.10 *** 2.99 1.97 *** 7.15 Attendance to religious services Never attends religious services -0.22 0.80 0.78 *** 2.19 Urbanicity Lived in urban area 0.55 * 1.74 1.02 *** 2.78 Intercept -5.163 *** -4.998 *** N 1875 1592 Log likelihood -572.2 -594.0 †p < .10; *p < .05; **p < .01; ***p < .001 1982–1988 2006–2010 Table B.2. Sensitivty analyis. Estimated coefficients and odds ratios from logistic regression of childlessness on sociodemographic characteristics, measuring family income as a percentage of the poverty threshold. NSFG 1982–1988 and 2006–2010: Women aged 40 to 44. 155 Appendix C: Additional Tables for Chapter 5 Number of children Childless (reference) --- --- --- 1 Child -0.55 *** -0.02 0.07 2 Children -0.77 *** -0.20 0.29 * 3 or more children -1.10 *** -0.60 * 0.38 * Age decade 20–29 (reference) --- --- --- 30–39 -0.40 *** 0.15 0.09 40–49 -2.77 *** -1.66 *** -1.62 *** 50–55 -4.94 *** -3.24 *** -3.45 *** No. children × Age Childless × 20–29 (reference) --- --- --- One child × 30–39 -0.10 -0.39 -0.28 One child × 40–49 0.51 *** 0.39 0.41 + One child × 50–55 1.42 *** 0.61 0.43 Two children × 30–39 -0.25 *** -0.18 -0.32 Two children × 40–49 0.98 *** 0.52 0.20 Two children × 50–55 2.07 *** 1.05 0.52 Three or more children × 30–39 0.24 * 0.34 -0.08 Three or more children × 40–49 1.85 *** 1.60 *** 1.14 *** Three or more children × 50–55 3.36 *** 2.24 *** 1.35 *** Education Less than high school (reference --- --- --- High school 0.42 *** 0.26 1.14 *** Some college 0.51 *** 0.15 1.28 *** College 1.80 *** 1.49 *** 2.13 *** Enrolled in college -0.96 *** -0.87 *** -0.84 *** Table C.1. Fixed-effects logistic regression coefficients predicting full-time employment by race/ethnicity. National Longitudinal Survey of Youth (NLSY) 1979–2012. White Hispanic Black 156 Years of part-time work experienc -0.02 ** -0.05 * -0.02 Years of full-time work experienc 0.27 *** 0.20 *** 0.22 *** Family characteristics 0.00 -0.02 -0.01 Net family income (in $1,000s) 0.00 *** 0.00 * 0.00 ** Married -0.30 *** -0.09 -0.05 Youngest child is <6 years old -0.79 *** -0.83 *** -0.56 *** Married × Youngest child <6 -0.42 *** -0.11 0.04 Number of persons 2,219 846 1,296 Number of person-years 43,822 16,071 25,540 Log likelihood -26,954 -2,147 -4,951 *p < .05, **p < .01, ***p < .001 White Hispanic Black Table C.1. Continued...
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Marital matching in West Africa: an examination of interethnic and interreligious first marriages in Benin
PDF
Improvements in women's status, decision-making and child nutrition: evidence from Bangladesh and Indonesia
PDF
Friendship network position on adolescent behaviors: an examination of a broker position and the likelihood of alcohol and cigarette use
PDF
A biodemographic approach to understanding sociodemographic disparities in kidney functioning on three dimensions: individual, population, and cross-national
Asset Metadata
Creator
Florian, Sandra M.
(author)
Core Title
Diverging fertility patterns? Racial and educational differences in fertility behaviors and their implications for socioeconomic mobility
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Sociology
Publication Date
04/15/2016
Defense Date
04/15/2016
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
age at first birth,fertility,Motherhood,OAI-PMH Harvest,socioeconomic inequality,women’s employment
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Casper, Lynne M. (
committee chair
), Biblarz, Tim (
committee member
), Crimmins, Eileen M. (
committee member
), Hook, Jennifer (
committee member
)
Creator Email
sandraf@usc.edu,squarechi2@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-229012
Unique identifier
UC11278088
Identifier
etd-FlorianSan-4258.pdf (filename),usctheses-c40-229012 (legacy record id)
Legacy Identifier
etd-FlorianSan-4258.pdf
Dmrecord
229012
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Florian, Sandra M.
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
age at first birth
fertility
socioeconomic inequality
women’s employment