<%BANNER%>

Pathways to Childlessness in the United States

Permanent Link: http://ufdc.ufl.edu/UFE0044847/00001

Material Information

Title: Pathways to Childlessness in the United States a Group-Based Analysis of Employment and Marital Union Trajectories
Physical Description: 1 online resource (122 p.)
Language: english
Creator: Tunalilar, Ozcan
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: childlessness -- life-course -- trajectories
Sociology and Criminology & Law -- Dissertations, Academic -- UF
Genre: Sociology thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The rate of permanent childlessness has been increasing in the United States for the last three decades. To identify distinct origins of childlessness, I examine lifetime patterns of education, employment and marriage between the ages of 18 and 44. Using data from National Longitudinal Survey of Youth (1979-2010), I identify trajectories of educational attainment, labor force attachment and marital status separately for men and women, and link them to likelihood of remaining childless. Results show that White, never-married, higher educated women, and women who are continuously employed are more likely to remain childless. White, never-married men, and men with lower paternal and higher maternal education are more likely to remain childless. Educational attainment and cumulative labor force experience are important predictors of remaining childless among women but not men. Later transition to first marriage increases the likelihood of remaining childless for men, but this is not the case for women. On the contrary, later transition to labor market increases women’s likelihood of remaining childless but not for men. Finally, men who have substantial decreases in their income during their prime years are more likely to remain childless. The distinct trajectories men and women follow to childlessness illustrate the lifelong patterns of accumulating risks for childlessness.  The discontinuous increases in risks for childlessness over different stages of the life course further reveal the critical periods for future fertility outcomes.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Ozcan Tunalilar.
Thesis: Thesis (M.A.)--University of Florida, 2012.
Local: Adviser: White, Robert G.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044847:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044847/00001

Material Information

Title: Pathways to Childlessness in the United States a Group-Based Analysis of Employment and Marital Union Trajectories
Physical Description: 1 online resource (122 p.)
Language: english
Creator: Tunalilar, Ozcan
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: childlessness -- life-course -- trajectories
Sociology and Criminology & Law -- Dissertations, Academic -- UF
Genre: Sociology thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The rate of permanent childlessness has been increasing in the United States for the last three decades. To identify distinct origins of childlessness, I examine lifetime patterns of education, employment and marriage between the ages of 18 and 44. Using data from National Longitudinal Survey of Youth (1979-2010), I identify trajectories of educational attainment, labor force attachment and marital status separately for men and women, and link them to likelihood of remaining childless. Results show that White, never-married, higher educated women, and women who are continuously employed are more likely to remain childless. White, never-married men, and men with lower paternal and higher maternal education are more likely to remain childless. Educational attainment and cumulative labor force experience are important predictors of remaining childless among women but not men. Later transition to first marriage increases the likelihood of remaining childless for men, but this is not the case for women. On the contrary, later transition to labor market increases women’s likelihood of remaining childless but not for men. Finally, men who have substantial decreases in their income during their prime years are more likely to remain childless. The distinct trajectories men and women follow to childlessness illustrate the lifelong patterns of accumulating risks for childlessness.  The discontinuous increases in risks for childlessness over different stages of the life course further reveal the critical periods for future fertility outcomes.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Ozcan Tunalilar.
Thesis: Thesis (M.A.)--University of Florida, 2012.
Local: Adviser: White, Robert G.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044847:00001


This item has the following downloads:


Full Text

PAGE 1

1 PATHWAYS TO CHILDLESSNESS IN THE UNITED STATES: A GROUP BASED ANALYSIS OF EMPLOYMENT AND MARITAL UNION TRAJECTORIES By OZCAN TUNALILAR A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Ozcan Tunalilar

PAGE 3

3 To my f amily

PAGE 4

4 ACKNOWLEDGMENTS I thank my t hesis committee, Dr. Robert White and Dr. Tanya Koropeckyj Cox, for their continuing help and interest in my learning This thesis would not be as well as it is right now witho ut their constructive comments and patience I also thank my parents, Penbe and Hasan Tunalilar, and my sister, Burcu Tunalilar, for their never ending e motional and financial support, and their trust on my success and abilities

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ 4 LIST OF TABLES ................................ ................................ ................................ ........... 6 LIST OF FIGURES ................................ ................................ ................................ ........ 8 LIST OF ABBREVIATIONS ................................ ................................ ............................ 9 ABSTRACT ................................ ................................ ................................ .................. 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ... 12 2 EXPLAINING CHILDLESSNESS ................................ ................................ ........... 21 3 TRAJECTORIES TOWARDS CHILDLESSNESS ................................ .................. 33 4 DATA AND METHOD ................................ ................................ ............................ 43 Dependent Variables ................................ ................................ ............................. 44 Trajectory Estimation ................................ ................................ ............................. 47 Regres sion Models and Control Variables ................................ ............................. 51 Missing Values ................................ ................................ ................................ ...... 54 5 RESULTS ................................ ................................ ................................ .............. 59 Correlates of Chil dlessness ................................ ................................ ................... 59 Trajectories ................................ ................................ ................................ ............ 69 Sensitivity Analysis ................................ ................................ ................................ 89 6 DISCUSSION ................................ ................................ ................................ ...... 104 LIST OF REFERENCES ................................ ................................ ............................ 111 BIOGRAPH ICAL SKETCH ................................ ................................ ......................... 122

PAGE 6

6 LIST OF TABLES Table page 5 1 Descriptive statistics by sex and parenthood status at age 44, National Longitudinal Survey of Youth 1979 2010* ................................ ......................... 90 5 2 T tests for differences on selected vari ables between parents and the childless by sex ................................ ................................ ................................ 91 5 3 Proportion tests for differences on selected percentages between parents and the childless by sex ................................ ................................ .................... 92 5 4 Mobility table with rates of childlessness based on average family income before and after 30 for women ................................ ................................ .......... 93 5 5 Mobility table with rates of childlessness based on average family income before and after 30 for men ................................ ................................ ............... 93 5 6 Odds ratios and standard errors of odds ratios from logistic regression models of remaining childless at or after age 44 on selected cov ariates by sex ................................ ................................ ................................ .................... 94 5 7 riage trajectories, National Longitudinal Survey of Youth 1979 2010 ................................ ................................ ........................... 96 5 8 Summary of fit statistics of school enrollment, labor force attachment and transition into first marital union trajectory models by number of trajectories for women ................................ ................................ ................................ ......... 97 5 9 Summary of f it statistics of school enrollment, labor force attachment and transition into first marital union trajectory models by number of trajectories for men ................................ ................................ ................................ .............. 97 5 10 Results from logistic regression models of remaining childless at or after age 44 on trajectory covariates for women ................................ ............................... 98 5 11 Descriptive statistics by school attendance, labor force attachment and Survey of Youth 1979 2010 ................................ ................................ ............. 100 5 12 Results from logistic regression models of remaining childless at or after age 44 on trajectory covariates for men ................................ ................................ 101 5 13 Cross tabulation of trajectory membership between models based on different operationalizations of labor force attachment for women ................... 102

PAGE 7

7 5 14 Cross tabulation of trajectory membership between models based on different operationalizations of labor force attachment for men ........................ 103

PAGE 8

8 LIST OF FIGURES Figure page 1 1 Parity distribution at completed fertility of birth cohorts: United States, 1912 1960 ................................ ................................ ................................ .................. 19 1 2 Rates of permanent childlessness for women by age groups in the Unites States, 197 6 2010 ................................ ................................ ............................. 20 4 1 Kernel density estimations of cumulative hours worked by age for men and women between ages 18 44 ................................ ................................ ............. 58 5 2 Estimated population prevalence and conditional role probabilities for latent trajectories for women, National Longitudinal Survey of Youth 1979 2010 ........ 95 5 3 Estimated population prevalence and conditional role probabilities for latent trajectories for men, National Lon gitudinal Survey of Youth 1979 2010 ............. 99

PAGE 9

9 LIST OF ABBREVIATION S AIC Akaike Information Criteria BIC Bayesian Information Criteria LLCA Longitudinal Latent Class Analysis NLSY79 National Longitudinal Survey of Youth 1979 Cohort

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts PATHWAYS TO CHILDLESSNESS IN THE UNITED STATES: A GROUP BASED ANALYSIS OF EMPLOYMENT AND MARITAL UNION TRAJECTORIES By Ozcan Tunalilar December 2012 Chair: Robert White Major: Sociology The rate of permanent childlessness has been increasing in the United States for the last three decades. To identify distinct origins of childlessness, I examine lifetime patterns of education, employment and marriage between the ages of 18 and 44. Using d ata from National Longitudinal Survey of Youth (1979 2010), I identify trajectories of educational attainment, labor force attachment and marital status separately for men and women, and link them to likelihood of remaining childless. Results show that Whi te, never married higher educated women and women who are continuously employed are more likely to remain childless. White, never married men, and men with lower paternal and higher maternal education are more likely to remain childless. Educational atta inment and cumulative labor force experience are important predictors of remaining childless among women but not men. Later transition to first marriage increases the likelihood of remaining childless for men, but this is not the case for women. On the con childless but not for men. Finally, men who have substantial decreases in their income during their prime years are more likely to remain childless. The distinct trajectories men

PAGE 11

11 and women fol low to childlessness illustrate the lifelong patterns of accumulating risks for childlessness. The discontinuous increases in risks for childlessness over different stages of the life course further reveal the critical periods for future fertility outcomes.

PAGE 12

12 CHAPTER 1 INTRODUCTION Over the last four decades, the developed world experienced declining fertility rates falling below replacement rates (Morgan & Taylor, 2006) Between 1970 and 2010, Total Fertility Rate (TFR) for all European countries dropped from 2.17 down to 1.53 (United Nations, 2011) During the eleven year period between 1990 and 2001, European countries with TFRs below 1.4 increased from 3 to 20, affecting 71.8% of all population in Europe (Sobotka, 2004) While the United States may stand at present as a rather special case because of its comparatively high er TFR, the decline in U.S. fertility followed the general European wide trend (Morgan & Taylor, 2006) In United States, the biggest declines in the fertility rates measured by TFR occurred during 1960s and 1970s, bottoming in 1976 with a TFR of 1.74 (World Bank, 2012) and it has leveled just above 2 since then (Martin et al., 2010) Although the effects from steadily increasing postponements in first birth to later ages are expected to diminish over time, many projections forecast long term sustained below replacement levels in the developed world (Alkema et al., 2011; Bongaarts & Feeney, 1998; Sobotka, 2004) While the decreasing rates of higher parity births and postponement of childbearing contributed most t o the declining TFRs in the United States as well as in many other industrialized countries (Bongaarts, 2002; Bongaarts & Feeney, 1998; Kohler et al., 2002) childlessness has steadily been an increasing component of low fertility. To illustrate the unfolding of this phenomenon in the U.S. context, I constructed parity distributions for women at completed fertility (at age 49) of birth cohorts from 1912 to 1960 using cumulative birth rates from National Vital Statistics (Figure 1 1 ). Starting with the early cohort of 1940s, the decline in higher parities was accompanied by

PAGE 13

13 increasing childlessness. Among women who were born in 1940, the proportion of childless women at the age of 49 was 8.14%. The corresponding proportion for women who were born in 1960 was 15.56%. Between birth cohorts of 1940 to 1960, the average growth rate of proportion of the childless was 3.07%. Annual growth rate ranged from 1.03% (from 1954 to 1955 cohort) to 11.89% (from 1942 to 1943 cohort) with a standard deviation of 4.59% across the 21 different birth cohorts between 1940 and 1960. Western European experience is comparable to the U.S. case: For women born between 1950 and 1954, France had the lowest rate of childlessness at age 45 (12%) while Austria had the highest with 17% (Rowland, 2007) With few exceptions in some years between 1976 and 2010, c hildlessness rates for women of age 40 44 increased from 10.2% to 18.8%, and from 10.5% to 19.7% for women of age 35 39 (U.S. Census Bureau, 2011) Figure 1 2 shows this increasing trend. Yearly data for men are not as easily available as they are for women. However, available data show that the estimated rates are even higher for men in the United States. From 1996 to 2006, the percentage of childless men age of 35 39 and 40 44 increased from 28.1% to 29.2% and from 18.9% to 22.4%, respectively (Bachu, 1996; Martinez et al., 2006) While childlessness is an increasingly visible phenomenon in the United States, it is not equally p revalent across different socioeconomic back grounds The higher educated women are still most likely to remain childless at the end of their fertile years but the rate of increase among the lower educated (high school dropouts or graduates) has been higher (Livingston & Cohn, 2010) Rates of childlessness also differ by family income. In 2008, 16.9% of women ages of 40 and 44 whose fami ly income were below

PAGE 14

14 $20,000 were childless while 14.8% of women with family income over $100,000 were childless (U.S. Census Bureau, 2010) The highest rate was 20.8% among women with family income between $ 30,000 and $ 49,999 Fertility decline and increasing childlessness have long term economic and social ramifications and remain a perennial concern for policy makers in low fertility countries (Clark et al., 2010; McDonald & Kippen, 2001) Declining fertility means that fewer people will be in the workforce in the next generation, which is expected to lead to problems associated with the labor market. For the United States, labor force projections by the Bureau of Labor Statistics show that declining fertility will translate into a decline in annual growth rate of labor force from 1.6% (1950 2000) to .06% (2000 2050) (Toossi, 2002) According to U.S. Census Bureau, with fewer people working to compensate for those who will not be working anymore, the dependency ratio is expected to increase from 22 in 2010 to 35 in 2030 (Crenshaw & Rab ison, 2010; Vincent & Velkoff, 2010) In other words, for every 100 people of working ages in the population, the number of people 65 and older will increase by 13 people. The importance of increasing dependency ratios is visible in the recent report by Social Security a nd Medicare Boards of Trustees (2011) which points out that Social Security will be p aying out more than it collects from the current workers. Similar or more dire increases in dependency ratios are expected in European labor markets (McDonald & Kippen, 2001) As a result of these changes in labor prospects, declining fertility portends an increasing need for immigran t workers for sustaining economies of industrialized countries, which may raise public policy concerns.

PAGE 15

15 The growth in childlessness also marks an important set of changes in American families with far reaching consequences. Increasing childlessness indicat es that a group of individuals may be emerging with certain vulnerabilities at older age. For instance, children can serve as valuable social resources by propelling parents to change the nature of their civic engagement. In absence of such motivation, the childless may feel less responsible for the problems associated with community and children, such as the schooling system. For instance, Keizer et al. (2010) recently found that childless men were less involved in their community than resident fathers such as engaging in volunteer work for local organizations (such as an association or a church) or providing help for the sick and the handicapped. In turn, the childless themselves may be less able to draw on community resources for assistance at older ages. Using data from the National Survey of Families and Households, Wegner et al. (2007) found that the main support network typ e for childless men and women in the United States was private, restricted networks associated mostly with little potential local support. Compared to the childless who live in other European countries included in the study (Australia, Finland, Germany, th e Netherlands, Spain, the United Kingdom), American men had the highest indicators of social isolation at age 65 and older. In addition, evidence suggests that the elder childless are more likely to be institutionalized (i.e., living in hostels and nursing homes) and live alone (Koropeckyj Cox & Call 2007; Rowland, 1998) and to have smaller family networks (Dykstra, 2006) The many events unfolding over the long period of the life course when one might have children suggest potentially numerous origins of childlessness. Life course theorists have long attributed the timing of major life transitions for lasting

PAGE 16

16 consequences for many dimensions of family life (Elder, 1994; Elder, 1995) The timing of marriage relative to one's remaining years for childbearing has obvious consequences for the opportunity for a first childbirth. However, the timing of other transitions may be just as important. Delays in leaving home and establishing an independent household imply children remaining closely tied to their families into a later period of adulthood. Longer durations of schooling similarly imply a longer period in a state commonly associated with low risk for first childbirth. However, the b iggest effects may come from labor market experiences. Early joblessness and extended periods of joblessness during early adulthood when marriage rates are highest may have particularly large consequences for transitioning to marriage, the remaining normat ive gateway to first childbirth. The destabilizing effects from episodes of joblessness for cohabiting couples may also bring a postponement in first birth that permanently shifts the lifetime risk for childlessness. Finally, t he re is growing evidence of p opulation wide declines in both marriage and fertility during economic recession (Sobotka et al., 2011) Early adulthood is a period of time when interrelated transition s in several domains of life take place, such as finishing school, starting work, experiencing unemployment, and entering marriage. On the one hand, institutionalized expectations for completing education coupled with a concern to find a stable job promisi ng a career during younger ages characterized by high fecundity can result in postponement of other life events, such as union formation and childbearing. Furthermore, career investment in early adulthood can increase opportunity costs of exiting labor mar ket in mid adulthood. Additionally, long spells of unemployment in early adulthood may hamper transition to parenthood as future parents may find having a stable job and flow of income as a

PAGE 17

17 prerequisite for family formation. As time passes, the postponemen t of childbearing due to labor force attachment may increase the risk of facing fecundity problems, unavailability of a suitable partner, and union dissolution, leading to remaining permanently childless This thesis examines the importance of lifetime exp eriences in education, employment, and marriage for remaining childless into mid adulthood. I use the National Longitudinal Survey of Youth 1979 (NLSY79) to examine the importance of men and women's individual trajectories through schooling, the labor mark et and marriage for their risk of remaining childless in middle adulthood. I adapt a novel method for describing individual life course trajectories that accounts for the timing of school leaving, unemployment and marriage as well as the duration in full t ime employment, part time employment and joblessness. I first identify 9 types of life course trajectories for women and 8 types of life course trajectories for men distinguished by unique timings and durations in schooling, labor force participation and m arriage. I consider men and women separately to account for the differences in both their life course experiences and upper age limits for fertility. I then examine the relationship between membership in these different trajectory groups and the probabilit y of remaining childless at age 44. Introducing individual membership in a group of closely related trajectories into a standard logistic regression model presents a tractable method for evaluating the importance of the heterogeneity in individual biograph ies for later life outcomes. This thesis proceeds in chapter 2 by reviewing theories of childlessness and the relevance of general theories of fertility for understanding childlessness. Chapter 3

PAGE 18

18 introduces an overview of the arguments as to how timing of transitions may be influential on the likelihood of remaining childless. Chapter 4 describes data, method, and analytical strategy and explains operationalization of each variable used in the analysis. Chapter 5 represents detailed analysis of results. Cha pter 6 summarizes results and includes discussion of the broad implications of these results for the state of knowledge in the field.

PAGE 19

19 Figure 1 1 Parity distribution at completed fertility of birth cohorts: United States, 1912 1960

PAGE 20

20 Figure 1 2 Rates of permanent childlessness for women by age g roups in the Unites States, 1976 2010

PAGE 21

21 CHAPTER 2 EXPLAINING CHILDLESS NESS Childlessness may result from choice, missed opportunities or diminished fecundity. While the distinction between these statuses is mostly made within a framework of agency in sociological literature (voluntary vs. involuntary), it is hard to draw the line between the two concepts (Sobotka, 2010) Admittedly, issues of intentions, expectations and desires mig ht be relevant factors in studies of fertility. However, there are few young people who intend to remain permanently childless (Houseknecht, 1987; Morgan & Rackin, 2010) and fertility intentions are unreliable predictors of completed fertility behavior (Morgan & Rackin, 2010; Quesnel Valle & Morgan, 2003) In addition, fertility intentions are subject to change over the life course (Hayford, 2009) Finally, people adjust their fertility expectations in response to events such as forming a new romantic relationship or losing a p artner (Iacovou & Tavares, 2011) Considering the tentative nature of fertility intentions my conce ptualization of childlessness in this thesis is not characterized around the question of agency. Instead, While adopting a child is a different pathway to becoming a pa rent, it has always been a very rare instance in the United States (National Center for Health Statistics, 2008) After weighting the option of dropping them from the sample, I considered adoptee parents as childless in line with my original definition. Finally, those whose children died before them can s till be considered as parents from a biological perspective and by the above de finition are classified as such in this thesis.

PAGE 22

22 Empirical studies of fertility point to a common set of correlates with permanent childlessness. M ost of these earlier studies focus on women while only a few includes male samples. C hildlessness is higher among Whites than Blacks and Hispanics (Abma & Martinez, 2006; Livingston & Cohn, 2010) positively associated with education for women (Abma & Martinez, 2006; Gonzlez & Jurado Guerrero, 2006; Livingsto n & Cohn, 2010) and more prevalent among the unmarried and never married than currently married and ever married for both women and men (Abma & Martinez, 2006; Keizer et al., 2007; Koropeckyj Cox & Call, 2007; Parr, 2010) In addition, currently working women are more likely to be childless than non working women (Abma & Martinez, 2006) Finally, continuity of employment seems to lead to higher likelihood of remaining childless for wo men and lower likelihood for men (Keizer et al., 2007) While there is consistent evidence regarding most of the characteristics of the childless, there is less agreement about the factors causing childlessness. Studies of childlessness commonly attribute the same factors identified in prevailing theories of low fertility to account for childlessness. These include characteristics of and changes in social institutions, market institutions and indiv idual socioeconomic circumstances. In addition, cultural norms related to life transitions including marriage and fertility have long been accorded a primary role in theories of fertility decline. Finally, technological advances that allow women to gain co ntrol over their reproduction and the incidence of biological infertility have also been argued as relevant factors in explaining low fertility and childlessness. For a large number of individuals, and especially for younger people, the decision as it conc erns their fertility is about whether to remain childless for the moment or to

PAGE 23

23 have the first child. In modern societies, this decision is mostly made within the institutional environment formed by labor markets, families established through marital and co habiting unions, and religious institutions. Although education is generally considered an individual level factor as years of schooling, remaining in school also involves an important set of institutional influences. Specifically, remaining in school req uires a patterning of social behaviors related to completing coursework and preparing for the job market which reduce the risk of a first birth both directly and indirectly through prevailing social norms which do not favor childbirth during educational ca reer. The last four decades witnessed the expansion of college and higher education among women, both in terms of degrees awarded and enrollment. Female enrollment to 24 year olds in the United States more tha n doubled over the course of 42 years, increasing from 19.2% in 1967 to 44.2% in 2009 (U.S. Department of Education, 2010) This extended participation in educational institutions during the transition into adulthood has been one of the leading factors of fertility postponement in the United States by later school leaving since education demands time commitment, and energy (Brand & Davis, 2011; Rindfuss et al., 1996) In addi tion to the likely difficulty of reconciling studen t and mother roles, the normative supposition that schooling and motherhood are not compatible with each other may further enhance the delaying effects of later schooling (Blossfeld & Huinink, 1991) The latter is best evident in debates over teenage childbearing levels in the United States (Morgan, 1996)

PAGE 24

24 the relationship between childlessness and education might be more complicated among men For instance, Kneale and Joshi (2008) found that childlessness was highest among British men with no qualifications ( e.g., elementary education and high school dropout) or tertiary qualifications ( e.g., college graduate and above ). Similar to their British counterparts American men who are high school dropout s might be more likely to remain childless than high school graduates (Marsiglio & Hinojosa, 2007) While h igher educated men may also remain childless due to reasons akin to higher educated women the reason lower educated men might also be more likely remain childless may be their inability to find a suitable partner since success in the marriage market depends on future prospects and status for men given their perceived primary role as provider. In addition and mostly due to this extended educational a ttainment during this period, women also started to work more, for longer periods of time, occupying higher status positions (Bureau of Labor Statistics, 2011; Percheski, 20 08) In the United ncreased from 43.3% up to 60% in 1999 and has remained around this level since then (Bureau of Labor Statistics, 2011) In addition, the proportion of marriages in w hich both husband and wife work increased from 43.6% in 1967 to 55.3% in 2009 with a peak of 60.4% in 1996. Finally, wives were the sole earners in the household for 6.6% of the married couples in 2009, the highest percent during the last four decades.

PAGE 25

25 ticipation has been proposed as the other major causal factor of fertility decline by sociologists (Bianchi, 2011; E sping Andersen, 1996) social demographers (Brewster & Rindfuss, 2000; McDonald, 2000) and economists (Becker, 1981; Becker & Barro, 1986) While family sociologists and social partici pation, economists focus on the opportunity costs of leaving paid employment and of childbearing. (2000) explains low fertility by pointing out the incompatibility between nonfamilial and familial institutions. His framework posits that a secular move occurred in the institutions of education and labor market from the br eadwinner model where men earn income and women manage the household to the gender equity model which favors equity in spouses' contributions to income and household management. Yet this change was not accompanied by compatible changes in familial institutions. As a result, in countries where women are forced to choose between familial responsibilities (e.g., housework, childbearing and child rearing) and nonfamilial activities (e.g., labor force participation), they respond by decreasing the number of children they have or by remaining childless to reconcile the two domains. has not been accompanied by changes in gender normative allocation of housework and c hildcare in the United States. Although women reduced their housework hours almost by half between 1965 and 1995 they continue to spend on average 1.8 times more time on overall housework, 3.8 times more time on housecleaning, and 2.8 times more time on c ooking than men (Bianchi et al., 200 0) I n addition, t he steady increase in

PAGE 26

26 women's hours of labor force participation has not been matched with a commensurate decline in hours of household management, with particularly little adjustment in time spent on child care (Bianchi, 2000) s to families increase their opportunity cost s of leaving the labor market as well as the costs of childbearing in terms of wages forgone, interruptions in job career, loss of experience due to passing time and decrease of leisure time (Becker, 1981) Simply delaying motherhood may lead to substantial increases in earnings, as highly as 9% per year of delay, and this is especially true for college educated women and those in professional and managerial occupations (Miller, 2009; Staff & Mortimer, 2011) Labor market characteristics, such as flexibility, may create heterogeneity in labor force participation ov er the life course among women. While role incompatibility between labor market participation and motherhood is evident in the United States, flexible labor markets providing part time job opportunities and availability of childcare may help women reconcil e labor market requirements and familial responsibilities by minimizing role incompatibility (Rindfuss et al., 2003) W omen who want to have children may adapt a diverse set of strategies to reconcile familial responsibilities and labor market req uirements by adjusting levels of their labor force participation instead of fully exiting the labor market. In this context, Esping Andersen (1999) argues that welfare state regimes offer unique institutional environments for individuals to decide on whether and when to parent. He posits that there can be different combinations of the type of labor markets, the state an d the family, which negotiate responsibilities for dealing with social risks and

PAGE 27

27 problems. For instance, if the state absorbs the responsibilities of childcare by subsidizing it on the market, monetary pressures on individuals who want to have children may ease as children in industrialized countries can be quite costly in monetary terms (Lino, 2008) Yet United States has one of the lowest public spending among OECD countries with $19,700 per each child from birth to the age of 5 while the average is $36,000 (OECD Fami ly Database, 2011) Alongside steady increases in women's educational attainment and labor force attachment, changes in marriage and cohabitation have brought comparable shifts in the factors related to fertility. While marriage remained an important pr ecursor to first birth throughout the latter half of the 20th century, marriage rates steadily declined and age at first marriage increased (Cherlin, 2010) For instance, between 2000 and 2010, marriage rate declined from 8.2 to 6.8 in the national level (CDC/NCHS National Vital Statistics, 2012) Although divorce rates somewhat declined during 2000s after two decades of increasing trend, the risk of lifetime divorce remains relatively the same (Cherlin, 2010; Raley & Bumpass, 2003) Looking at the trends in union formation behavi ors in the United States during 1970s and 1980s, Bumpass (1990) had previously argued that nonmarriage and increasi ng divorce rates weakened pronatalist pressures of the family on individuals, especially on women, leading to a different set of constraints on fertility decisions. However, the notion that marriage is the only social place in which childbearing and child raising can occur has been altered during the last three decades. The number of births to unmarried women reached at a peak level in 2008, 40.6% of all births to unmarried women from a level of 18.4% in 1980 (Martin et al., 2010) Yet, these results

PAGE 28

28 do not mean that all unmarried women are ne cessarily single. Musick (2002) found that approximately 40 percent of nonmarital births occurred within cohabiting unions. Finally, although cohabitation might be a step towards marriage for some, it seems to serve as a desirable end point for others (Musick, 2007) As a result of these changes, while the effect of marriage may still be cont inu ing, cohabitation has become increasingly influential on fertility outcomes In addition to the effects of institutional factors, demographic l iterature on low fertility often emphasizes the effects of modernization and economic development on changing cultural values and declining influences of religious institutions (Davis, 1963; Notestein, 1945) Increasing secularization coupled with individualization in the developed world has b een offered as a causal force behind fertility decline since r (Uecker & Stokes, 20 08) and indirectly through its deterrent effects on nonmarital fertility (Wildeman & Percheski, 2009) For instance, Zhang (200 8) found that the str ength of religious beliefs is positively associated with the number of children respondents had regardless of religious denomination they belong and the effects were similar for both men and women ( see also, Frejka & Westoff, 2007) C hanging conceptions of gender and family in the developed world may also have been influential in increasing levels of childlessness P ost materialist value theory points to the effects of cultural shift from collectivist materialistic to individualistic values on industrialist societies (Inglehart, 2008; Kaa, 1987; Kaa, 2001) This view suggests that p arenthood is no longer viewed as the main goal or a

PAGE 29

29 necessary step towards happiness and life satisfaction in these countries. Empirical evidence s upports this view. In their analysis of historical trends in attitudes towards childlessness in U.S. from 1960 to 1990, Thornton and Young DeMarco (2001) found that parenthood is no longer viewed as a necessary component of being a couple in the United States. In addition, their results suggest that childlessness have become more positive More recently, Koropeckyj Cox and Pendell (20 07) childless adults could have fulfilling indicate that childless individuals face an increasingly more tolerant social context regarding their fertility Alongside the changes in institutional environment follow t echnological advances allowing y oung women's control over their fertilit y as a facilitating condition of low fertility N ew contraception opportunities such as the pill make it easier for women t o control the consequences of sexual activity. Goldin and Katz (2000) suggest that there are direct and indirect effects of the pill for the prospects of women. The direct effect is that it enables women to be able to focus on their education and careers without the fear of pregnancy, which can disrupt both of these pursuit s. Indirectly, it enables women to delay marriage by reducing the necessity of marrying to have sex and lowering the possibility of shotgun marriages (Goldin & Katz, 2000) Aside from factors associated with the institutions and changing culture, literature emphasizes the role of i ndividual racial and socioeconomic circumstances on the outcome of low fertility and childlessness. While Hispanic and Black women aged 40 to 44 had replacement level fertility in 2008, non Hispanic White and Asian women had a rate of 1.8 children per women that is more similar to rates experienced in Europe (Dye,

PAGE 30

30 2010) There are also differences in rates of childlessness across racial g roups within the United States: Although W hite women are historically more likely to remain childless and this is still the case, the racial gap has been narrowing in more recent years (Abma & Martinez, 2006; Livingston & Cohn, 2010) In this context, f amily background can also be an other important indicator of remaining childlessness directly (Murphy & Wang, 2001; Murphy & Knudsen, 2002; Parr, 2005) and indirectly through its effect on family bui lding behavior (Amato et al., 2008; Landale et al., 2010; Schoen et al., 2009) For instance Murphy and Wang (2001) argue that fertility patterns of parents and children may be associated in the form of intergenerational continuities through socialization genetic and available kin support Using the National Survey of Families and Households they found that sibship size explained 2% to 4% of the variance in number of children among ever married parents after controlling for parental and individual level factors such as sex, age, race, education and region Similarly, Parr (2005) found that higher number of siblings is associated with lower likelihood of remaining childless for Australian women c ontroll ing for current marital status, highest level of education, type of school atte nded and region of birth childl essness. A later study by Parr (2010) showed no significant relationship between number of siblings and the likelihood of remaining childless a mong Australian men. Compared to men whose fathers had no occupation when they were 14, men whose fathers had high status or low status jobs were less likely to remain childless. On the other hand, men whose mothers were working in professional jobs had hi gher likelihood of remaining childless compared to those whose mothers were not working when they

PAGE 31

31 were 14. Finally, once controlling for marital status, the effect of level of education lost its significance. Apart from studies that focus on individual fac tors quantitatively scholars who do qualitative work adopted an analytical strategy by which experiences of the childless help defining who they are and why they are childless. For most part, this literature individuals who is assumed to make a decision to remain childless, meaning that the person in question has the ultimate control over his/her fertility. This line of research has identified reasons reported by respondents (during face to face interviews) ab out why they are currently childless. For example, Houseknecht (1987) identified nine possible motives towards childlessness: freedom, more satisfactory marital relationship, female career considerations, monetary advantages, concern about population growth, general dislike of children, doubts about spect of childbirth and finally, concern for c hildren given world conditions. Later qualitative findings do not disagree with her initial enumeration. An attempt by Somers (1993) childless individuals. More recently, Carmichael and Whittaker (2007) found many rejecting motherhood identity completely, selfish identity and unsuitability for parenthood role was themes that prevailed in the interviews. Lack of partner or reluctance of the partner for parenthood was exogenous reasons for women of why they ended up being

PAGE 32

32 childless. Finally, Park (2005) interviews identified (observed) parenting models, (mostly negative) feelings abou t children and population growth concerns as some of the other reasons suggested by voluntarily childless of why they are childless. What these qualitative studies reveal might be after the fact rationalizations of their current situation by respondents si nce all of them are retrospective accounts. The notion that individuals choose childlessness over having children is not well to remain childless for the rest of their l ives, are a small group (Houseknecht, 1987; Morgan & Rackin, 201 0) F or instance, Morgan and Rackin (2010) found that only about 5 percent of 20 year old respondents intended to remain childlessness. Yet, even if they were a larger group, the fact remains that fertility intentions do not always correspond with actual behavio r. Most Americans actually underachieve the number of children they intended to have due to postponing childbearing and marrying late (Quesnel Valle & Morgan, 2003) In addition, fertility intentions are highly su bject to change. Heaton et al. (1999) found that almost 62% of the respondents who wanted to remain childless in the first interview switched to wanting children in a period of six years. The youngest respondent in the second interview was 25 years old. Finally, Hayford (2009) concluded recently that there is not a subgroup of women who always want to remain childless over the life course.

PAGE 33

33 CHAPTER 3 TRAJECTORIES TOWARDS CHILDLESSNESS The large literature concerning fertility consistently identifies factors related to education, labor markets, marriage and cultural norms which may contribute to low fertility. Yet, there are limits to the relevance of this evidence for understanding childlessness. The main problem is that many o f the identified correlates with age at first birth vary over time. Changes in schooling, labor market experiences, living arrangements and family structure present a shifting set of risks over the life course. Risks associated with childlessness may also change at certain points over the course of fertile years defining critical periods where effects of transitions, and lack of transition thereof, can drastically increase. Consequently, many of the correlates with childlessness reflect the cumulative effec ts of a lifetime of transitions associated with va rying risks for a first birth. Empirical studies of the predictors of later life childlessness risk ignoring the intermediate accumulation of these risks leading up to eventual childlessness Studies which emphasize contemporaneous correlates with childlessness during the years beyond the average biological fertil e period may similarly underestimate the importance of the timing of prior major life transitions. This study considers the importance of a lifetim e of transitions for completed fertility by evaluating the distinct trajectories individuals follow through schooling, the labor market and family life. Transitions occur when life events such as leaving school or marrying take place and they change indivi Considering multiple transitions over an extended period of time provides a temporal may eventually lead to remaining childl ess.

PAGE 34

34 M ost of the studies that focus on childlessness or age at first birth do not take into account the life stage at which individuals actually spent time in the marita l union or in the labor market. Put differently, the effects of timing of transition to labor force and union formation on completed fertility mostly remain an uncharted territory in the literature. On the other hand, the complexity of life events that occur during transition to adulthood is well studied (Elder, 1994; E lder, 1995) M any people leave home go to school, enter the labor force experience short or long spells of unemployment, and get married or have a partner before even considering having children However, more important is the fact that n ot all these transitions are in harmony with each other at each point in life and countervailing forces may be the reason behind individuals not having children at all or delaying having u ntil after their fertile years. For instance, l eaving school may directly increa se the risk for first childbirth as well as indirectly through inc reasing the risk for marriage or cohabitation (Guzzo, 2006; Landale et al., 2010) On the contrary prolonged schooling may lead to postponement of marriage and having children. Furthermore, a spell of unemployment may directly decrease the risk of having a first child for men as w ell as indirectly through its negative effect s on marriage prospects (Carlson et al., 2004; Guzzo, 2006; Xie et al., 2003) Finally, t he effect of marriage on the risk of first birth also li kely strengthens over time, with effects from marriage occurring when women are in their late 30s being much greater than when they are in their mid 20s (Baizan et al., 2 003; Martin, 2000) As the complex of lif e events compete for individual s limited time and capacity of commitment, they may simply delay making a decision regarding whether to have a child. Instead of seeing

PAGE 35

35 childlessness as a one time decision individ uals make, it should be considered as the cumulative outcome of many decisions, limited by socioeconomic constraints. Consideration of the entire trajectory of transitions provides an alternate method for describing the heterogeneity in individual biograph ies. Examining transitions presents a challenge because of the complexity of life course s (Shanahan, 2000) This complexity is embedde d in the number of transitions and possible sequencing of these transitions. While data related to many transitions are readily available, researchers generally lack necessary analytical tools to study all available data. In this context, t rajectories can be considered as representations of life histories or biographies, narrating role acquisitions over the life course with a focu s on the timing of transitions. They are capable of capturing timing and sequences of transitions in addition to providing a novel way of visually representing these changes over time. More importantly, d efining trajectories allows classifying individuals by the type of timing, duration and sequences occurring through a defined set of transitions. This classification provides analytical leverage to examine the likely effects of tran sitions on fertility outcomes. For this reason, i t is equally important to analyze an optimal number of trajectories that may represent the differences in timing of transitions. Family scholars have been interested in trajectories of labor force participation and marital status since the 1960s. These earlier discussions focused o n the increasing transitions, and methodological challenges to analyzing such deviations (Rindfuss et al., 1987; Uhlenberg, 1974) More recent studies extended the scope of these earlier studies by restricting the analysis to certain life course periods, such as late

PAGE 36

36 adolescence or young adulthood (Amato et al., 2008; Cooksey & Rindfuss, 2001; Macmillan & Copher, 2005; Oesterle et al., 2010) S ome of these studies consider tr ajectories as precedence of some transitions over other transitions (Cooksey & Rindfuss, 2001) while others focus on transitions during early adulthood to represent increas ing diversity in timings of transitions to adulthood among newer cohorts (Amato et al., 2008; Macmillan & Copher, 2005) Transition to adulthood generally occurs through transitions between educational, and market and familial institutions. Although attending to school may be accompani ed with earlier union formation and part time employment marriag e and labor force participation generally follows after the initial completion of schooling (Amato et al., 2008; Landale et al., 2010) Individuals with prolonged schooling exit school not only with a larger number of years of schooling, positively associated with higher age at first birth for women (Brand & Davis, 2011; Rendall et al., 2010) but also having spent a larger number of their fertile years in an environment that poses more constraints to childbirth. Higher educational attainment make s it necessary to postpone marriage due to t ime conflict and can lead to adapting to a life style that is not compatible with having children (Blossfeld & Huinink, 1991; Carmichael & Whittaker, 2007; Thornton et al., 1995) Landale et al. (2010) found that attending school decreased the likelihood of all family transitions (i.e., marriage, cohabitation, and birth) among young women. Guzzo (2006) found similar results for both men and women: school enrollment was associated with higher likelihood of no union formation compared to marriage or cohabitation. In sum, trajectories characterized by higher educational attainment would be expected to have later transitions to marriage than those who show earlier school leaving patterns.

PAGE 37

37 While later school leaving has an initial delaying effect on family formation and having children, it may also increase likelihood of finding a partner and having a child through increased human capital in the long run (Thornton et al., 1995) For example, Ono (2003) found that each year of education increased the likelihood of first marriage among never married women in the United States. In addition, Landale et al. (2010) con c luded that completing education increased the likelihood of transition to marriage. These results suggest that higher educational attainment might just be shifting the timing of transition to marriage t owards older ages among the highly educated, but not h amper the eventual transition. Apart from the timing effect s of school leaving, education may have differential direct effects on completed fertility for men and women Keizer et al. (2007) found that level of education predicted permanent childlessness for w omen but not for men. Each year of education increased the likelihood of remaining childless by 14% for women. sc (2010) results show a similar nonsignificant relationship f or me n. L abor force participation may necessitate postponement of having a first child for women (Blau & Robins, 1989) C urrently working women are more likely to be childless than non working women (Abma & Martinez, 2006; U.S. Census Bureau, 2010) Unemployment, on the other hand, can actually decrease the risk of having a first child amo ng men while it may increase the likelihood of having children for women (Brewster & Rindfuss, 2000; Sobotka et al., 2011) For example, Rindfuss et al. (1998) found that unemployment discour aged men from e arly parenthood.

PAGE 38

38 While there is agreement in the literature on the likely effect of current labor force participation on fertility it is less clear whether the effects change over time and depending on levels of labor force participation. In this context, a temporal consideration of l abor force attachment and cumulative labor force experience can be more revealing to analyze the likelihood of remaining childless Women who start a full time job upon leaving school may find it harder to have and raise child ren at the beginning of their career due to higher opportunity costs associated with exiting the labor market for childbearing. On the contrary, early employment may actually decrease the lik elihood of remaining childless for men i n the bread winner model of family relations, where men are exp ected to provide for the family Over time, continued full time employment may change attitudes of individuals towards family and children, thus leading them to a life style not compatible with having children (De Ollos & Kapinus, 2002) With a focus on employment history, Keizer et al. (2007) found that continuity of employment led to higher likelihood of remaining childless for women and lower likelihood for men. Contrarily after controlling for current occupation, age, income and early life course variables such as paternal and maternal education at age 14, number of siblings and country of birth, Parr (2010) recently found that time spent working was not significant in changing the likelihood of remaining childless for Australian men. It is evident that both labor force participation and attachment have different effects on fertility outcomes for men and women. In addition, w omen and men are likely to follow distinct trajectories due to their different educational attain ment, labor force participation, and biological deadlines. Considering distinct effects of certain trajectories

PAGE 39

39 on remaining childless depending on gender, it is essential to analyze trajectories of men and women separately. At the time of school leaving the higher educated participate in the labor market more persistently, that is, remain highly attached to the labor market and reemployed more easily when they are unemployed (Neumark, 2002; Riddell & Song, 2011) As a result, a later but rapid transition would be expected among those who have college education or more. Additionally, those who have lower educational attainment may find it harder to find a job with a stable flow of income. This debilitating effect of ear ly spell of unemployment may have a long lasting effect on employment stability, especially for the cohorts under study (Bureau of Labor Statistics, 2010; Neumark, 2002) Long term labor force attachment is also related to the transition to a marital union. Marriage increases the likelihood of exiting the labor market for women, depending on Transition to a marital union would also indirectly increase the likelihood of exiting the lab or market since marriage increases the risk of childbearing (Brewster & Rindfuss, 2000) On the other hand, marriage and especially transition to fatherhood is associated with higher labor force attachment among men (Knoester & Eggebeen, 2006; Lundberg & Rose, 2002) However, there are two main possibilities that can explain the latter phenomenon. First, men might be waiting unt il establishing themselves to get marriag e and transition to fatherhood (May, 1982) If that is the case, their labor force attachment would be expected to have reached a higher level before transition to marriage. Second, marriage may be initia ting a more compatible life with the requirements of higher labor force

PAGE 40

40 attachment for men (Knoester & Eggebeen, 2006) In that case, m en would be expected to increase their labor force attachment after the transition to marriage. The s chool to work transition can facilitate transition to a marital union through increased earnings for men, but the effects may be more complicated for women Losing a job or starting to work is related to decisions individuals make about marrying or cohabiting (Guzzo, 2006) Xie et al. (2003) found that earnings potential positively affect the likeli hood of getting married for men, but t his was not the case for women. Their operationalization of earnings potential includes predicted current earnings, earnings over the next five years, past earnings, future earnings, and lifetime earnings. Similarly, C arlson et al. (2004) found that among unmarried couples who just had concluded that women's higher education was positively associated with the likelihood of future marriage. C ontrary to these findings, Ono (2003) had both direct and indirect effects (t hrough increasing chances of cohabitation) on le d to increases in the likelihood of first marriage in the next period after controlling for education, local area sex rat io, and In addition to the effects of timing of school leaving and cumulative labor force attachment, the timing of marriage may be influential for completed fertility. T hose who delay union formation to later may feel the biological clock ticking more loudly and rush behavior during 1975 to 1995, Martin (2000) concluded that there is a trend of increasing first birth rates after the age of 30 onl y for four year college graduates.

PAGE 41

41 However, other educational groups were more likely to end up being childless if they 30. In addition, Heaton et al. (1999) found that higher education was associated with belonging to eit Contrarily, the enough to decrease the likelihood of remaining childless. In support of this view, Keizer et al. (2007) has recently shown that age at first marriage does not have any significant effect on the likelihood of remaining childless but every year withou t a partner increases Recently, Parr (2010) found similar results regarding years of marriage and cohabitation for Australian men. These results also stress t he importance of the effect of duration in marital union on childlessness. In this thesis, I cla ssify individual biographies into groups of common trajectories through school leaving, types of labo r force attachment and marriage. The crucial part of the analysis is the comparison of trajectories with different timing of transitions controlling for ot her factors identified in the literature. A s I discussed above, behaviors in each of these trajectories may have different effects on the likelihood of remaining childless for men and women so I analyze them separately. Overall, literature suggests several working hypotheses that I proceed to test in this thesis. First, there will be pathways to childlessness for both men and women. These could be considered as the trajectories with highest likelihood of remaining childle ss. For women, they are expected to be never married full time employed women I also expect that earlier school to work transition will increase the likelihood of

PAGE 42

42 remaining childless for women while earlier u nion formation will decrease the likelihood of remaining childless compared to those who are never married independe nt of labor force participation. Contrarily, for men, those who experience an early unemployment period during their prime fertile years are expected to have highest rates of childlessne ss. in the labor market, I expect that men who transition to high labor force attachment in the absence of closely following union formation would more likely to remain childless than others. Finally, I expect that men who have substantial decreases in their income during their prime years will be more likely to remain childless

PAGE 43

43 CHAPTER 4 DATA AND METHOD I study the trajectories to childlessness using panel data from National Longitudinal Survey of Youth, 1979 Cohort ( NLSY79 ). NLSY79 is a nationally representative sample of 12 686 men and women who were 14 22 years old when they were first interviewed in 1979. Respondents were interviewed annually from 1979 until 1994, and biennially through 2010. I use all av ailable waves in this analysis. The NLSY79 presents many advantages for studying fertility. Subjects were asked questions about school, family life, social relationships, work and daily activities. The detailed que stions about labor force participation and family life were repeated annually and then biannually, providing an exceptionally high frequency of observations for a nationally representative sample. A large share of respondents has also been fo llowed into middle adulthood, a period when most women subjects reach the bi ological limit for childbirth. I restrict the analysis to the respondents whose childlessness status is known at the age of at least 44 or later. As a result, I exclude the respond ents whose permanent childlessness status is not available (n=4 368). Although it is biologically possible to have a child after this age, especially for men, previous research shows that it is quite unlikely (Keizer et al., 2007; Kirmeyer & Hamilton, 2011) In this sample, 99.2% of fathers and 99.9% of mothers had their first child before the age of 44 while the age of the respondents ranged from 45 to 55 at the date of the last interview. I consider those responden ts who had their first child after the age of 44 as parents (n=27). With these constraints and missing data the analytical sample consists of a total of 6,398 respondents: 3,131 men (48.94%) and 3,267 (51.06 %) women.

PAGE 44

44 I employ a two step analytical approac h to link educational, employment and marital trajectories to permanent childlessness. The first step identifies the most likely clusters of traject ories of education, labor force attachment, and marital status in the sample. The second step involves using trajectory membership as dummy covariates in logistic regression models, controlling for a variety of factors previously identified in the literature as related to age at first birth and permanent childlessness. Dependent Variables This analysis uses two set s of depende nt variables. The first set of variables is measures of education, labor force attachment, and marital union formation of the respondents, and used in estimating the trajectories. The dependent variable of interest remaining childless is u sed in logistic regression models NLSY79 includes t he number of children ever born to the respondent I constructed a dichotomous variable using this they had one or logistic regression models as the dependent variable of interest. I estimated educational trajectories using a variable that captures whether respondent was enrolled at school at each age Approximately 67% of the sample had the same years of education at age 24 and age 44. 12% had one more year of education over the course of 20 years, 9% completed two more years of education and 5% had three or more years of education between two measurem ent points. The mean difference for overall sample was .70 years. Motivated by t his relativel y stable educational attainment after the initial school completion education trajectories can also be considered as showing the timing of school leaving

PAGE 45

45 I consider labor force attachment with reference to a definition of full time work in order to consolidate the many outlying values of hours worked common in studies of labor force participation. For instance, a common concern in reports of hours worked is that self employed individuals or employees of family businesses may report relatively high numbers of hours worked, often reaching as high as 100 hours per week. Using labor force attachment instead of more simple employment status used in other studies a lso allows me to show likely effects of part time employment on the likelihood of remaining childless. The U.S. Department of Labor considers full time employment as working 35 hours or more per week and part time as 1 to 34 hours per week. Rather than rel ying on this rather arbitrary thresh old for defining full time work I used kernel density estimations to distinguish between those who have low, medium and high labor force attachment. Figure 4 1 shows the kernel density graph for each age between 18 and 44. The lines represents two possible cut points to distinguish between low and medium attachment (273 and 396 hours a year), and two possible cut points to distinguish between medium and high attachment (1534 and 1742 hours a year). I rounded these values so that they would be meaningful in terms of weeks worked: 260 (5 hours/week), 416 (8 hours/week), 1560 (30 hours/week), and 1820 (35 hours/week), respectively. Using the latter values, I created four new variables to indicate levels of labor force attach ment for each respondent for each age. I start the analysis with cut points of 260/1560 as indicators of latent employment trajectories and use other operationalizations of labor force attachment in the sensitivity analysis.

PAGE 46

46 Beginning with the 1993 wave N LSY79 allowed respondents to report hours worked on average for each week u p to 168 hours a week an increase from the previous upper limit of 96 hours a week. T he creation of labor force attachment variable accounts for this measurement change successfully as those who worked more than at least 1560 hours at every age (30 hours a week) were grouped together as having high labor force attachment. In e stimating marital trajectories, I used the age at first marriage as an absorbing state variable for which respondents had a value of zero before the age at first marriage and had a value of one after that age The never married were assigned zeros for all ages. Consequently, t he shape of marital trajectory can be read as the average timing of the first marriage for respondents belonging to that parti cular trajectory group. I preferred this operationalization since I already control for the direct effects of incidence and duration of th e first and second marriages in second part of the analysis. While a large share of first births in the U.S. occur s in cohabiting unions (Bumpass & Lu, 2000; Musick, 2002; Musick, 2007) I do not estimate separate cohabitation trajectories due to two main reasons First many cohabita ting unions are short lived (Schoen et al., 2007) which makes it harder to capture the formation or dissolution of the union due to data un availability Second, for a considerable number of cohabitating couples, cohabitation is just a step towards marriage (Mu sick, 2007) As such, some of the effects of cohabitation would be expected to be picked up by the eventual marital union formation. Considering these factors, I include the number of co residential partnerships in the logistic regression models to acco unt for the direct effects of cohabitation on remaining childless instead of estimating trajectories of cohabitation.

PAGE 47

47 Trajectory Estimation In the first stage of the analysis, I estimate the clustering in trajectories u sing the variables I just described I define trajectories as common states (i.e., roles) a group of respondents hold at a certain age between ages 18 and 44. Consider two simple array s denoting two s to first marriage at age 24 and age 26 The array s would look like th e SSSSSSMM SSSSSSSS What is common to both arrays is the ir first 6 and the last 19 values. T hese two trajectories differ in terms of three values and i f we assume a sample is composed of only these two individuals, we can conclude that two trajectories of transition to first marriage sufficiently summarize the data. Now consider a third respondent who never made the transition to first marriage. The array The third respondent shares only the first 6 values with the first respondent and 8 values with the second one. While this introduces more variance to the array structure, t his hypothetical sample can still be summarize d using three trajectories repr esenting three different array s C onsidering that the transition to first marriage can occur at any age between 18 and 44 in the analytical sample, 27 separate trajectories of transition to first marriage would be needed to exactly fit the data. However t ransition to first marriage is only one of the trajectories I examine in this analysis. In my operationalization of trajectories, the labor force attachment is represented by a three category ordinal variable (low/medium/high) and schooling and marital sta tus by binary variables. This means that each respondent could follow one of the 27x2 27 x3 27 (2.76 10 22 ) possible patterns While that is theoretically the case, a relaxing attribute of life course analysis is that transitions are not random. For instance leaving school mostly occurs early in young

PAGE 48

48 adulthood and many individuals never go back to school, which decreases the number of likely patterns in a major way. There have been attempts at defining certain patterns over the life course in terms of prece dence of a state change in related to another ( for instance, see Cooksey & Rindfuss, 2001) This type of pattern analysis o ver developmental trajectories which aim at repre senting transitions amo ng different states over time requires the researcher to decide on which patterns are analytically relevant. For example, with two binary variables, there are four patterns that can empirically emerge from the data. As the number of variables increases, there will be fewer people who follow any of such patterns. The researcher may try to allocate individuals into most frequent patterns using a predetermined rule, but this allocation would not be statistically testable. In addition, th is type of pattern analysis ignores the time dimension of the data, that is, the age at which transitions occur. W orking with longitudinal data covering a long lifespan represents a challenge due to large number of repeated observations, all correlated with each other. Since these repeated measures cannot be considered independent from each other, more standard models such as OLS regression cannot be applied to this kind of data and more elaborate models are needed. In this context, eithe r structural equation models or growth mixture models can be used (Nagin, 2005) In this context, Longitudinal Latent Class Analysis (LLCA ) empirically identifies clusters of individuals who follow similar patterns of status (Goodman, 1974) It presumes that variation between individuals can be represented by using classes for a finite numb er of distinct subgroups. LLCA models can be conside red in terms of either

PAGE 49

49 structural equation models or growth mixture models, depending on interpretation (Muthn, 2004) I n both interpretations, observed variation in state differences over time is assumed to be explained by a latent variable. That is, had data fitted a LLCA model perfectly each respondent would be a member of one and only one C latent class and that observed measurements are independent of one another conditional on latent class mem bership (i.e., local independence). Put differently the latent variable is assumed to explain why the observed items are related to one another. If that assumption holds, then each trajectory group includes individuals empirically similar regarding the characteristics concerned. Whereas growth mixture models assume a specific shape for the trajectory, LLCA does not hold such an assumption a characteristic which allows presentation of complex or irregular behavior. More specifically growth models force data to fit to a smooth change as a function of time represented by the latent intercept, slope and quadratic terms On the contrary, t wo separate sets of parameters are estimated in LLCA. The first one, conditional probabilities, corresponds to the probability of an individual being in a certain category of a measurement given class me mbership, e.g., probability of having high labor force participation at a certain age given his or her membership to a certain trajectory. The second one, posterior class probabilities, s pecify the relative prevalence of each class with the sample. Let represent the latent variable and be the number of latent classes. Moreover, let repres ent one of the observed variables, where and be the number of levels of As such, a particular latent class can be denoted by an index

PAGE 50

50 and a particular value of by Finally, t he vectors Y and y refers to a complete data array T he probability of obtaining array pattern y, is a weighted average of the class specific probabilities ; which is: Given the local independence assumption, each observed variables are assumed to be mutually independent, which leads to: Time dimension can be considered as a different operationalization of repeated observat ions of manifest variable Each trajectory array extends over If we index observed variable by time (t), then each measurement point can be denoted by For instance, labor force attachment at age 18 can be represented by an observed v ariable with three (j) ordered categories where I estimate all models using Mplus, Version 6.12 which estimates thresholds instead of conditional item and class probabilities (Muthn & Muthn, 2010) That is, each threshold, is estimated separately to determine the item and cla ss probabilities in or above category j and indexed by time (t) and class (k): In this conceptualization of trajectories, the timing of transitions in each trajectory is a side outcome that can be exploited for the purposes of this analysis. For instance, increasing probabilities of an outcome such as transition to first marriage signal that

PAGE 51

51 members of that specific trajectories are completing their transition to first marriage. Similar interpretations can be made regarding other trajectories. Combined trajectories related to school leaving behavior, labor force attachment and transit ion to first marriage form the pathways, which show the interrelated behavior in each of these trajectories. More than one selec tion criteria have been proposed to decide on how many trajectories are enough and/or necessary to empirically describe data. Bayesian Information Criterion (BIC) is the most widely used measure, where a smaller value corresponds to a better fit with a lar ge log likelihood value and not too many parameters (Raftery, 1995) E ntropy represents a measure of classification of the model using individual class probabilities. In other words, e ntropy measures how distinctively individuals are distributed to each trajectory in the model (Muthn, 2004) However, there is yet an agreement as to which one is the best or proper (Nylund et al., 2007; Pickles & Croudace, 2010) Muthn (2004) suggests that several tests and substantive theory should be used to decide on the num ber of classes to represent data. In addition, groups with very small percentages (e.g., below 3 5%) may be less informative especially when coupled with low frequencies. As a result, I count on all three measures discussed above (whenever possible) in add ition to group percentages, and substantive theory to decide on which model is more suitable for the purposes of this research. Since the concern in this thesis is the effect of timing, similar trajectories which can successfully differentiate between an e arly and later transition to labor force would be preferable. Regression Models and Control Variables The second stage of the analysis estimates the risk of childless ness conditional upon the estimated trajectories controlling for a broad range of factors related to fertility.

PAGE 52

52 I use logistic regression to assess the effects of membership to a specific trajectory on the lik elihood of remaining childless. Including membershi p indicators for trajectories into the base model allow s me to test whether trajectory membership actually explains some of the variance in remaining childless, and if so, in what ways. At this stage, I include a set of control variables, including work hi story in the form of average hours worked between the ages 18 and 44 family background variables suggested by literature: maternal and paternal education, number of siblings, and youth religious attendance and affiliation To control for the direct effects of education on the likelihood of remaining childless, I calculated completed educational attainment at or after age 44, using number of years of schooling. I next categorized years of education at age 44 in a more informative way. I consider those who have less than 12 years of education as high school dropout, those who have 12 years of education as high school graduates, those who have 13 to 15 years of education as some college educated, those who have 16 years of education as college graduates and those who have more than 16 years of education in the category of higher education. To account for the effects of lifetime labor force participation apart from the effects of timing of labor force attachment, I created a cumulative measure of yearly ho urs worked. I divided this cumulative measure by a full time equivalent (35x52 =1820 ) to better reflect average work experience of respondents compared to full time employment. I use three measures to control for the effects of marital history. These measur es include t he length of time in months the respondent spent in a marital union, separately

PAGE 53

53 for first and second marriages and total number of spouses and coresidential partners who were living with the respondent at the time of interview Considering the fact that hourly wages differ, it is necessary to account for the effect of overall income over the life course. I account for income with a measure of individuals' re lative standing within the NLSY 79 income distribution. Assuming that spousal income may contribute as much as personal income for the decisions related to having a first child, I see it more relevant to include a measure of family income at ages when respondents were married and individual earnings when they were single. I also account for ch anges in income over time. While income has been shown to display a nonlinear rel ationship with total fertility (Fieder et al., 2011; U.S. Census Bureau, 2010) and a posit ive relationship with age at first birth during early adulthood, sudden shifts in income may alter these relationships. An unanticipated loss of income among childless men and women in the middle of the income distribution may result in postponing childbir th plans, resulting in a decrease instead of the increase in the probability of childbirth suggested by the negative income coefficients commonly estimated in fertility models. Consider an individual with low earnings during their 20s due to getting higher education (which could decrease their likelihood of finding a partner or having children). Their returns on education would not be realized until their early 30s when they graduate and find a stable job. To get the simple average of income during these tw o periods would conceal the upward or downward trends in income in addition to the effects of timing. Instead, I calculated the average income of each respondent before and after the age of 30. I then assigned respondents to three categories (low, medium,

PAGE 54

54 high) depending on whether their income fell within the bottom, middl e or top 33 percentiles of the income distribution of the overall sample for the given time period. Using these two variables, I constructed a 3x3 mobility matrix with nine categories where the first part shows the initial relative income position of the respondent among all other respondents and the second part shows whether the respondent increased, decreased or remained relatively the same over time. For the res pondent I gave as an example at the beginning of the paragraph, the assigned value would be low/high (1 3). I use dummies to control for different initial and change situations in income in logistic regression models. Additional controls added to the model include race, family background and religios ity. Race is a key variable in NLSY79 coded as (1) Hispanic, (2) Black, and (3) Non Black Non Hispanic White. I constructed dummy variables to include in the logistic regression models. Famil y background is meas ured by maternal and paternal education in addition to the number of siblings the respondent reported. Religiosity is measured by religious attendance in 1982. Missing Va lues In trajectory estimation part of the analysis, m issing variables are dealt with M AR (missing at random) assumption using a full information maximum likelihood (FIML) estimator with robust standard errors. There is an emerging consensus among statisticians and family scholars that FIML provides less biased estimates than other methods o f dealing with missing data, e.g., listwise deletion or multiple imputations (Acock, 2005; Johnson & Young, 2011; Schafer & Graham, 2002) As a result, I did not use imputed values for any of the measures of school enrollment, labor force

PAGE 55

55 att achment, or marital history. C ovariance coverage value of analytical data was above .60 in all models estimated, which is way above than minimum required value of .1. NLSY79 provides information about weekly labor status and hours worked on average for eac h week for each respondent from one year before the initial wave began until the week of the last interview (1978 2010). For the periods for which respondent was not available for interview, data were collected retrospectively. As a result, there are few m issi ng data for employment history. In calculating the cumulative measure of hours worked, I imputed missing values with the ir mean value of nonmissing values if missing data was less than 65% of data for the year (i. e. complete data for 35 weeks) for respondents who had missing data for their weekly data I chose this method because missing values were generally within a long array of nonmissing data with similar values. For example, the respondent reported that they had worked 35 hours a week for 39 w eeks, then there was a missing value, and then they reported 35 hours a week for another 12 weeks. Many respondents in this initial sample had missing data for paternal (n=1254) and maternal (n=540) education since it was asked directly only in the first i nterview (1979). However, NLSY79 asked detailed questions about educational attainment of household members at each wave. If respondents were still living with their mothers or fathers (that was especially the case for younger respondents), I was able to r etrieve education when there were many waves in which respondents were still living with their mother and/or father. Data belonging to 347 mothers and 287 fathers were ret rieved. Finally, NLSY79 provides the number of siblings of the respondent.

PAGE 56

56 T here was a considerable amount of missing data for the family income variable. Percent of missing values for the family income variable provided by NLSY79 ranged between 14% and 25 % depending on the survey wave. Put differently, 19% of person period data before age 30 and approximately 35% of person period data after age 30 had missing values. However, cross tabulation showed that there was no directional relationship between number of missing variables and the categorical measure of income. In other words, higher number of missing variables was not systematically associated with higher or lower levels of income over time. As a result, I decided to impute the average values of non mi ssing values for missing values (i.e., I calculated the mean value for the non missing values). Finally, t he frequency of religious attendance was measured in 1979 and 1982, and religious affiliation was measured in 1979. Since a considerable number of res pondents were quite young in 1979, I use the 1982 question. If the respondent had a missing variable for that year, then I retrieve religiosity from the initial survey in 1979. As a result, no respondents have missing values for religious attendance. 35 re spondents had missing values for religious affiliation in 1979. There were a couple of differences in characteristics between the initial and the final sample. T test results show that the final sample is less likely to be White, have more years of educati on, worked more hours, spent more time in 1st marriage, and have higher maternal education and fewer siblings (not shown). Although this reduction may raise concern with regards to representativeness of the analytical sample and the external validity of th e study it should be noted that the characteristics of the sample still remain similar to those used in other studies, especially in terms of rates of

PAGE 57

57 childlessness, and employment, marital and educational outcomes. I discuss these similariti es in Results section

PAGE 58

58 Figure 4 1 Kernel density estimations of cumulative hours worked by age for men and w omen between ages 18 44

PAGE 59

59 CHAPTER 5 RESULTS Correlates of Childlessness Childless NLSY79 cohort members display many of the characteristics of chi ldlessness previously reported. Table 5 1 shows descriptive statistics by sex and parenthood status at age 44. The first thing to note is that t he proportion of the analytic sample that is childless is comparable to estimates for wome n based on National Vital Statistics (Kirmeyer & Hamilton, 2011) nationally representative surveys such as the National Survey of Family Growth for men and women (Martinez et al., 2006) and findings from studies of childlessness for women (Abma & Martinez, 2006) That is, 16.53% and 22.17%, for women and men, respectively White and Black women are more likely to be childless but these differences are not significant based on t test results (Table 5 3 ). The only racial difference is among Hispanics as Hispanics are disproportionately represented among mothers than among the childless. This result partially confirms p revious research which suggest s that White women are more likely to remain childless than Black and Hispanic women (Abma & Martinez, 2006; Livingston & Cohn, 2010) Note that all groups include higher proportion of Blacks and Hispanics due to problems explained in sample selection. Childless women are more educated at age 44 than mothers as suggested by the literature (Abma & Martinez, 2006; Keizer et al., 2007) Childless women are more likely to have college degrees and higher education than mot hers while mothers are more likely to be high school drop outs and high school graduates than the childless (Table 3) Logistic model including only an indicator of college degree (16 years of education

PAGE 60

60 and above) shows that those who have college degree a nd above are 2.16 times more likely to remain childless than those who have lower than college education. Childlessness is associated with higher lifetime cumulative labor force participation among women. For instance, c hildless women are more likely to have worked for more hours between the ages 18 and 44 Specifically, childless women worked for 12.7% more hours equivalent of full time on average compared to mothers a difference which corresponds to a n approximate mean difference of 9 4 00 hours. In add ition, childless women worked for 177 weeks more than mothers (not shown) These provisional findings are in line with the literature which suggests that the transition to motherhood increases the probability of exiting labor market and decreases the proba bility of return for women leading to lower labor force attachment over the life course (Brewster & Rindfuss, 2000; Gangl & Ziefle, 2009) T he duration of marriage is an other important indicator of remaining childless for women. Table 5 2 shows that c hildless women spent less time in both first and second marriages than mothers as suggested by the literature (Keizer et al., 2007) Note that 14.5% of all women in the analytical sample were never married which is comparable to national cohort est imates (U.S. C ensus Bureau, 2011) C hildless women were more likely to be never married with a percentage of 38.2 while more than 90 % of mothers were ever married a finding similar to what literature indicates (Abma & Mart inez, 2006) Mo thers had also significantly higher total number of spouses a nd coresidential partners than the childless. In addition to occurrence and duration of marriage timing of marriage is strongly associated with remaining childless. Age at first marriage was the

PAGE 61

61 lowest among ever married mothers with an average of 22.94 while it was 27.38 on average among ever married childless women Mothers have lower average income than childless women during 20s, but this gap disappears during 30s. S pecifically, w hile childless women have approximately 1500$ higher average family income than mothers during their 20s, the reverse is the case after 30 with mothers having 4317 $ higher average family income than childless women S ince I calculate the average income in terms of family income, this result may simply imply that mothers are getting married during 30s (and their income now includes In addition to crude measures of average family income, I created mobility tables to reflect the effect of change in income over time. Table 5 4 shows the mobility table for women. Those who are in the lower and higher strata during their 20s are more likely to remain in the same position during their 30s while middle class women show more social mobility. In terms of rates of childlessness, the most striking finding is the case of women who went from high strata down to low strata. Althoug h they are a small minority in the overall women sample with a prevalence of 6.75%, they are the most likely group to remain childless. The second most likely to remain is another declining group in terms of income who were in the high strata during 20s an d got down to middle strata during their 30s. These findings provide provisional evidence for the likely effects of social mobility on the likelihood of remaining childless. Finally, family background is associated wi th childless status among women (Table 5 2 ) Childless women have significantly higher paternal and maternal education than mothers with difference s of 1.2 years and .92 years, respectively. Childless women

PAGE 62

62 also have fewer siblings with an average difference of .72 Religious attendance at yout h does not differ between childless women and mothers while mothers are more likely to be Catholics than childless women and the reverse is the case for Protestants. Specifically, mothers have a 9% higher frequency of reporting Ca tholic affiliation during youth compared to childless women. There are considerable contrasts for the contemporaneous correlates of childlessness among men compared to women While racial proportional differences among men are similar to that of women, only White men are significantly disproportionately represented among the childless. C ontrary to women, t here are not s completed education at age 44 while the college educated are disproportionately represented among the ch ildless men. Although these results may imply that education does not seem to have much direct as it does for women indirect effects of education through earnings potential and prospects of marriage are still hi ghly likely for men A striking difference between men and women is that t he association of labor force attachment with remaining childless is reversed among men compared to women. Fathers are significantly more likely to work for more hours. During ages b etween 18 and 44 fathers worked approximately 3.5% more than the childless men on average. Put differently fathers seem to have higher labor force attachment overall. The lower average hours worked among childless men illustrate s an important difference between the childless men and women. While education differences between fathers and childless men do not show the sharp differences evident among women, childless men's

PAGE 63

63 lower level of labor force participation is suggestive of th e distinct origins of men and women's childlessness. Similar to differences among women, f athers are also more likely to spend more time in first and second marriages, and have a greater total number of spouses and partners. However, t hese differences seem to be even larger than the difference between mother s and the childless women. For instance, more than half of childless men (52%) were never married by age 44 the largest share among all groups Even ever married childless men have the highest me a n age at first marriage with an average of almost 30. These findings stress the likely higher importance of marital trajectory in terms of incidence, duration and timing, among men compared to women Not only do fathers have higher average family income during their 20s, the size of the difference also gets bigger over time The average family income difference between fathers and childless men increases from 450$ up to 13585$ between the two periods While a considerable share of this increase might be reflect highe r likelihood of getting married, a large proportion of the increase between two periods may be due to the fact that f atherhood is associated with an spent in paid work (Gla uber & Gozjolko, 2011; Knoester & Eggebeen, 2006) The mobility table for men in Table 5 5 provides a different perspective for the likely effects of income mobility on the likelihood of remaining childless among men. While both the prevalence of social mobility among men and the contemporaneous association between income mobility and the likelihood of remaining childless is similar from middle to low strata with a childle ssness rate of 34.45%. This finding can be

PAGE 64

64 considered in the general trend that those who declined in the relative income position seem to be more likely to remain childless than those who did not. This finding provides provisio nal evidence for the hypothesis that men who experienced substantial decreases in their income during their prime years are more likely to remain childless. Childless men have higher maternal education (but not paternal) than fathers in addition to fewer s iblings. Moreover the difference in maternal education among women is bigger than the difference among men. telling more about gender and f amily attitudes in the family wh ile these men were growing up. Finally, fathers show higher levels of religious attendance than childless men when they were young Note that the signifi cant difference among men indicates its relative importance of religios ity compared to women for which the difference was nonsignificant However, there are no religious affiliation differences between fathers and childless men. The descriptive statistics clearly show two likely pathways into remaining childless among men con sidering that education might be associated with lower likelihood of unemployment, especially for these cohorts. First, college educated and full time employed men are delaying union formation and having children. Second, chronically unemployed men are finding it hard to establish themselves, find partners and have children. Overall, these initial findings suggest two things. First, as I discussed whenever previous literature was available for comparison the sample I use is comparable to other studies of childlessness in terms of educational, employment and marital

PAGE 65

65 characteristics although not so much for r acial differences. S econd, education and childless than it is for men. Related to the latter point, lifetime cumulative labor force participation has diverging effects on remaining childless for men and women which support my argument about the necessity of analyzing the two separately. Estimated differences between par ents and childless individuals in the NLSY79 cohort from logistic regression models with a full set of controls verify many of the characteristics of the childless previously reported. Table 5 6 reports estimates from stratified logistic regressions including a set of controls common in models of fertility outcomes, separately for women and men. shows that white women are almost two times more likely to be childless than Black women controlling for educational attainment, em ployment and marital history, and family background. The same is true for Hispanics however, the odds ratio of that difference is smaller than it is for Whites. High school dropout, high school graduates and women who have some college education have low er likelihood of remaining childless compared to those who have higher education (more than 16 years). T here is a steady linear relationship between completed education and remaining childless for women as evident from the change in the magnitude of the co efficients. Note that this is in contrast to the effects of education but mostly nonsignificant association. In terms of employment history, higher cumulative h ours worked on average increase the odds of remaining childless for women. Specifically, women who always

PAGE 66

66 worked full time over their fertile years (between ages 18 44) are approximately 6.8 times more likely to remain childless than women who never worked during this period. The time s pent in first and second marriage decreases the odds of remaining childless significantly and the effects of first and second marriage do not seem to differ in terms of odds ratios. In addition, controlling for marital history and all other covariates in t he model, the total number of spouses and coresidential partners is still statistically significant and reduces the odds of remaining childless for women. Since there are few women in the sample who had three or more marriages, this finding s uggests that b oth marriages and coresidential partnership have independent diminishing effects on the odds of remaining childless. Descriptive statistics showed that a ver age family income before age 30 was higher among childless women compared to mothers but I suggested that this may be that independent of race, education, marital and employment history, and family background, family income during 20s has a positive effect on the likelihood of remaining childless for women. Although family background seemed to be associated with remaining childless when looking at the descriptive statis tics, the significance levels are much lower when controlled for other factors. In addition, the effect of maternal education disappears. Every year of paternal education increases the odds of remaining childless by 4.3% for women. In addition, each siblin g significantly reduces the odds of remaining childless by 6.4%. Finally, those who were raised as Protestants when they were growing up are more likely to remain childless than their Catholic counterparts.

PAGE 67

67 in Table 5 6 shows that the differenc es for men and women in the race estimates are quite striking Even after controlling for socioeconomic status, marital history and family background, White and Hispanic men are likely to remain childless than Black men. Specifically, White men are more th an three times more likely to be childless than Black men controlling for other covariates in the model. Hispanics are also more likely than Blacks to remain childless, but similar to women, they have lower odds of remaining childless compared to the diffe rence between Whites and Blacks. This finding is not inconsistent with the possibility of two pathways to childlessness for men since controls for men's employment trajectories are not included in this model. As a result, this model cannot account for diff erent timing of transitions among men which may explain some of the racial differences. The effect of education is not as big as it is for women. The only significant difference is between the higher educated and college graduates, which is quite the oppos education increases the odds of remaining childless by 79% among men controlling for race, employment and marital history, and family background. The descriptive statistics had s hown a positive effect from education for men, although it was weaker than for women. Logistic regression results here show a somewhat significant effect, and it is not only negati ve, but it is very large. This finding is consistent with the negative effec ts of socioeconomic status for men's childlessness. Employment history does not seem to have any effect on likelihood of remaining childless among men controlling for other covariates in the model, similar to what Parr (20 10) found among Australian men. In this context, average family income during any

PAGE 68

6 8 period does not have any effect on the likelihood of remaining childless for men unlike women. These non significant result s may be suggesting that there is nonrandom selection into marriage whereby men who have low labor force participation and lower income may be less likely to enter marriage. In other words, men who cannot find stable jobs might be having trouble finding suitable partners too. I checked this probability by entering cumulative employment, income, and marital history variables (which measures duration spent in first and second marriages), to the model one by one. The results provide some s upport for this argument. With out controls for marital history and income measures the cumulative hour s worked significantly decreased the odds of remaining childless for men by 64%. Including income measures turned the effect of employment to nonsignificant, explaining all the effect. Income measure after age 30 was itself significant, decreasing the odds of remaining childless almost by half. Finally, as seen in the final model, including marital history again turned the effect of income measures to nonsignificant. Overall, these nested models suggest that the effect of employment and income for men ldlessness is most likely being mediated by their ability find a suitable partner using their resources. Spending time in first and second marriages reduces the odds of remaining childless among men and the magnitudes of these effects are similar to those among women. However, total number of spouses and partners decreases the odds of remaining childless among m en more than it does for women. childless except rel igious background and attendance, which are all non significant. Contrary to women, paternal education reduces the odds of remaining childless among

PAGE 69

69 men. On the other hand, every year increase in maternal education increases the odds of remaining childless childless approximately 7%. Trajectories I now estimate models of trajectories of education, employment, and transition into f irst marriage for women and men I s with a discussion of characteristics of each trajectory. I continue with detailed description of certain trajectories, especially the ones that are at high risk for remaining childless Finally, I recount the formal model selection process which involved deciding on the number of trajectories necessary to fully explain transition differences between individuals I Figure 5 1 shows the sample prevalence and conditional probabili ties for latent trajectories for women and Table 5 7 shows descriptive statistics trajectories. Overall, nine trajectories fit the data best identifying differences in the analytical sample regarding timing of transitions. I leave formal discuss ion of model selection to later after discussion of trajectory characteristics. I assigned individuals to different trajectories using highest predicted probability Considering that entropy was very high for the model with 9 trajectories for women (discussed below), the assignment was very clear cut. I facilitate the discussion by considering trajectory graphs and descriptive statistics together. The distinctions I make here regarding timing of transitions in each trajectory help me in the second part of the analysis where I test whether differences in timing influence the likelihood of remaining childless. I consider each trajectory independently as well as in rela t ion to each other considering that labor force attachment trajectories

PAGE 70

70 are highly dependent on school leaving trajectories in the sense t hat the latter mostly determine the timing of the former. The first thing to note is that not all trajectories have the same prevalence in the sample as would be expected. The biggest group is trajectory #1 with 17.05%. It is followed by trajectories #3 and #5 with percentages of 15.18 and 13.96, respectively. The smallest group identified is trajectory #9 with a prevalenc e of 5.97%. The most strik ing finding reflected in Table 5 7 is the natural clustering of trajectories in terms of likelihood of remaining childl ess. Trajectories #6 and #7 have much higher rates of childlessness than all other tr ajectories while trajector ies #2, #4, and #5 include almost no childless individuals. Since no consideration is given to the probability of childlessness in the estimation of models this finding provides convincing evidence for my hypothesis that there are in fact pathways to chil dlessness. Looking at educational trajectories, it is easily seen that trajectories # 2, #3, #4, #5 and #9 are characterized by early school leaving which occurs between the ages 20 and 22. On the other hand, trajectories #1, #6, #7, and #8 show later school leaving which starts around age 25 and extends up to age 29. Descriptive statistics agree with this distinction. Early school leavers are mostly high school graduates and those who have some college education. On the other hand, later school leavers (#1, #6, #7, and #8) are evenly distributed between some college education, college graduation, and higher education. There are striking educational features of some trajectories. For example, trajectory #8 is formed by women of whom almost %25 have highe r education. Contrarily, trajectories #4 and #9 have a high school dropout rate of approximately 24%.

PAGE 71

71 Transition to labor force can be considered in relation to school leaving behavior or independently. In relation to school leaving behavior, there are thr ee types of trajectories with simultaneous, delayed, or no transition. Simultaneous transition to labor market occurs in tandem with declining probability of school enrollment and trajectories # 1 #2, #6, #7, and #8 show this behavior. Delayed transition i s seen in trajectories #3 and #5, which is characterized by later increasing probability of high labor force attachment. Finally, trajectories #4 and #9 show almost no transition to labor force and remain mostly marginally attached or unattached to the lab or market. In addition to initial transitions, labor force attachment varies over time among trajectories. For instance, although trajectories #2 and #8 make earlier transitions, individuals who belong to these trajectories soon leave the labor market in t andem with their increasing likelihood of marriage. Distinct differences for lifetime labor force participation among trajectories are also evident in Table 5 7 with cumulative hours worked range from 37.3% to 96.5%. In terms of marital trajectories, t hes e nine trajectories are grouped into four categories of transition Trajectories #2, #3, #4, and #5 are characterized by an early transition to first marriage While there are difference s in the exact timing of transition to first marriage among these trajectories, the transition is completed before age 27 for all of them Descriptive statistics show that the mean age at first marriage for these trajectories range from 19.23 to 20.23 T he latter four of these trajectories are also early school leavers. Second group trajectories #1 and #8, includes those who made a later transition to first marriage between ages 24 and 26 This is also evident in their mean age s at first marriage, which are 24.40 and 25.77, respectively. In addition to later

PAGE 72

72 transition t o first marriage, these traject ories are later school leavers. Trajectory #7 is unique in terms of timing of marriage. While this group marries even later than trajectories #1 and #8, they complete their transition to marriage around age 3 8 with a mean age of first marriage of 31.76. In this respect, t hey differ from the last group. The last group includes trajectories which show almost no marital transition. Descriptive statistics confirm this clustering: 83.7% of trajectory #6 and 79.5% of trajectory #9 were never married. Even those who married in these trajectories had quite high average at first married, 41.24 and 38.69, respectively. It should also be noted that these two trajectories have no occurrence of second marriage. Different timings of school leaving, labor force participation and marriage seem to overlap among trajectories allowing comparisons for the effects of unique timing of each trajectory. While trajectories #2, #3, and #5 have all early school leaving and marriage transitions, their tra nsition to la bor force differs considerably. In addition, the only difference between trajectories #1 and #7 in terms of transitions is their transition to first marriage with the latter transitioning much later than the former. I use these comparisons in the second part of the analysis to test whether these timing differences matter in terms of likelihood of remaining childless. D ifferences between trajectories with regards to racial composition and family background also merit some discussion. For instanc e, trajectories #2 and #8 are overwhelmingly White while trajectory #9 is disproportionately Black. Considering that approximately %17 of the women is Hispanic, trajectories #4 and #5 consists of more Hispanics disproportionate to the sample. Family backgr ound between trajectories differs in unique ways. Trajectory #8 have unmatched levels parental education and

PAGE 73

73 have the lowest average number of siblings. Trajectories #6 and #9 are much more likely to be Protestant compared to other trajectories both of whi ch have high percentages of childlessness. The most notable among these trajectories in terms of remaining childlessness is trajectory #6. 56.7% of individuals who belong to this trajectory remained childless at the end of their fertile years. Compared to other trajectories, these individuals have definitely higher education. In fact, 18.7% of women who belong to that trajectory have higher (more than 16 years) education. Overall, it can be said that this group have a later school leaving pattern coupled wi th a fast transition into the labor market (see graph s ). This is also evident in their average cumulative labor force participation. These women were almost always full time employed. In terms of marital history, these women almost never spent time in firs t marriage and actually never spent time in second marriage. Compared to other trajectories, they also had the lowest number of spouses or partners. These women have also comparably low family income, which probably indicate their never transition to marri more educated than many of the other trajectory groups and they have the second highest percentage of being Protestant (lower than only trajectory #9). In summary, these are highly educated, never married women who made their school to employment transition very fast and early, and worked full time during their prime fertile years. Considering that these are all the risk factors identified in the literature regarding remaining childless, it is no surprise that th ey have such a high likelihood of remaining childless.

PAGE 74

74 The second interesting group is trajectory #7 with their 27.7% childlessness rate. Although that is not as high as trajectory #6, it is still much higher than other trajectories. They have comparable r acial distribution to all women in the sample (see Table 1). These women have even higher overall education than trajectory #6. For instance, 24.4% of women who belong to that trajectory have more than 16 years of education. Visual inspection of the trajec tories in Figure 5 1 reveals that an important difference between these groups lies with the timing of events. T rajectory #7 made transition into first marriage during their 30s and spent almost 10 years in their first marriages on average. Their labor for ce participation seems to decline after they got married (see graph) and this is also evident in their average cumulative labor force participation (89.5%) which is lower than trajectory #6. Overall, this group can be identified as later school leaving, fa st transition into the labor market, and later transition into the first marriage. The discussion regarding characteristics of trajectories clearly shows that these t rajectories identify substantively meaningful trajectories that allow comparing the e ffect of timing of transitions. The purpose of employing this novel method was to find an optimal number of trajectories that would best represent the differences in timing of transitions in the sample. T he first decision in estimating the trajectories is const raining the model by the nu mber of possible trajectories. This requires the analyst to make informed decisions and judgment. Table 5 8 shows the summary of fit statistics for trajectories of school enrollment, labor force attachment and transition into the first marital union for women. Notice that parameters are quiet high because is an estimated threshold parameter for n 1 category for each trajectory for each age As the number of

PAGE 75

75 trajectories increase, BIC and AIC statistics decrease. This indicates that more trajectories are needed. I estimated up to 10 trajectories for women yet the attempt to extract 10 trajectories was unsuccessfu l due to convergence issues. Entropy did not help in deciding on which number of classes should have been chosen because all the models had excellent discriminatory power between individuals as entropy for all models are over .95. In other words, posterior probabilities for class membership for almost all of the individuals were either 1 (belonging to that particular class) or 0 (not belonging). This discrimination power is very important since I use the class membership as a covariate in the logistic model s. From the model with 7 trajectories to the model with 8 trajectories one of the trajectories split into two trajectories, one with steep transition into high attachment during 30s and one with persistently low attachment over the life course. Other traj ectories were not influenced from the increase of trajectories judging from trajectory characteristics (descriptive statistics by trajectories not shown ). From the model with 8 trajectories to the one with 9 trajectories, there was not an obvious split. H owever, there was now another trajectory with early high attachment coupled with transition into marriage, low attachment during the late 20s and early 30s and increasing high attachment during late 30s and early 40s. As this is a unique trajectory among o thers, I decided to choose 9 trajectories to describe the transitions. I now proceed by testing the effects of trajectory membership controlling for all other factors in the base model I have previously established. Note that these trajectories are formed by school enrollment, levels of labor force attachment and absorbing state marital status and they inform us about the overall transitions for those

PAGE 76

76 who belong to a certain trajectory. Since I am controlling for the direct effects of race, education, emplo yment, marital history and family background, any significant effect of covariates which indicate belonging to a trajectory compared to others can be read as the li kelihood of remaining childless compared to others. Table 5 10 shows baseline logistic regression model from Table 5 6 in addition to models c ontrolling for class membership for women I initially chose trajectory #5 as the reference category as it is the least likely to group to remain childless and allow for comparison of transitions. Note that trajectory #5 is characterized by early school leaving, early transition to marriage and later transition to labor force. The results clearly show that trajectory membership is an important predictor of remaining childless even after controlling for the direct effects of race, education, employment, marital history income and family background The most obvious finding is that compared to trajectory #5, all other trajectories are more likely to remain childless. More importantly, t he magnitude of effects of class membership differ s immensely between trajectories This finding confirms that timing of transitions matter a nd the effects of transitions in different trajectories differ on their effects on remaining childless. In line with findings from descriptive statistics, trajectories #6 and #7 have the highest likelihood of remaining childless For instance, belonging to trajectory #6 increases the odds of remaining childless almost 13 times compared to trajectory #5. This finding confirms my hypothesis that there are in fact pathways to child lessness. While magnitudes of effects in terms of odds ratios differ between tra jectories #6 and

PAGE 77

77 #7, joint tests of equality shows that this difference is not significant 2 (1, N = 3267) = 0.59, p = .44. Put differently although membership to these trajectories increases the likelihood of remaining childless for women separately, the ir effects are not different. Unique characteristics of trajectories allow comparisons between timing of transitions controlling for some other transition. In discussion characteristics of different trajectories, I have previously delineated these pairs of transitions I now test whether distinct comparisons lead to unique likelihood of childlessness using joint tests of equality, i.e., Wald test. First, while both trajectories #3 and #5 left school early and made earlier transitions to first marriage, their transition to labor f orce differ in terms of timing. As results from Model 1 show, those who belong to trajectory #3 are 7 .3 times more likely to remain childless compared to those who belong to trajectory #5 Note that this is independent of the act ual time spent working since I control for cumulative labor force participation in the logis tic model. A similar timing difference exists between trajectories #2 and #5 in terms of transition to the labor force B oth of these trajectories are characterized by early school leaving and early marriage, but trajectory #2 transition ed to labor market upon school lea ving while trajectory #5 remained marginally attached. Model 1 shows that belonging to trajectory #2 increases the likelihood of remaining childless more than three times compared to trajectory #5. Both of t hese finding s support my hypothesis that earlier transition to labor force participation among women le ad to higher li kelihood of remaining childless independent of timing of school leaving and marriage, race, completed education, cumulative labor

PAGE 78

78 force participation, family income, marital history, and fam ily background. It is important to note that these findings are still significant against a very rigorous test since some of these controls are either not included in other studies of childlessness and age at first birth or not as detailed as they are here. W hile both trajectories #1 and #7 is characterize d by early school leaving and early transition to labor force the former trajectory made an earlier transition to first marriage while the latter made that transition much later Jointly estimating coefficients of these trajectories did not reveal significant differences 2 (1, N = 3267) = 1.60, p = .20. This result indicates that although they both lead to higher likelihood of childlessness compared to trajectory #5, the magnitude of effect does not differ. This nonsignificance can be explained by the direct effect of duration of first marriage which is controlled in the model. S ince those who marry later would be expected to spend less time in the first marriage, all the effects of transition to first marriage might be being explained by the durati on variable Yet it still shows that timing of first marriage does not have an independent effect controlling for occurrence and duration of first marriage. Finally, although both trajectories #7 and #8 were later school leavers with later transitions to f irst marriage and simultaneous transition to labor force, trajectory #7 did not exit the labor market while #8 did after the initial high attachment The joint test result indicates a significant difference between the coefficients of membership to these t rajectories 2 (1, N = 3267) = 14.05, p < .001. This means that those women who leave labor market at some point during their 20s are less likely to remain childless, independent of their school leaving behavior, initial labor force attachment, and

PAGE 79

79 transition to first marriage, and all other controls in the model. This is not surprising, though, considering that most women leave labor market to have children. In terms of model fit, BIC shows that trajectory membership variables did not improve the mo del compared to the number of variables included. However, pseudo R squared shows that trajectory membership explains 2.1% of the variance of remaining childless. Including trajectory membership as covariates attenuates the effects of marital history and e mployment history variables while it enhances the effects of race, completed education, and family background. The first effect is expected since trajectories carrying information about duration of marriage and employment, explain away some of the effects of marital history and employment. However, the second effect is rather unexpected. This means that there is a suppression effect of trajectory membership operating on both likelihood of childlessness and race, completed education, and family background. For instance, the effects of race covariates increased after the inclusion of trajectory membership. Furthermore, the effect of paternal education as well as the effect of being raised in other religion became significant. I next include income mobility co variates instead of coarse measures of average family income in the model (Table 5 10 Model 3 ) None of the effects I previously found change regarding the effects of timing even though the effects of trajectory membership attenuate. R esults also reveal t hat there are certain income mobility differences in pathways to childlessness among women. For instance, those who are always low in the income distribution are less likely to remain childless compared to those who are always in the high high group. Although high medium group was more likely to remain

PAGE 80

80 childless compared to high high group, an explanation for this result is less readily obvious and merit s further future analysis Figure 5 2 shows the sample prevalence and conditional probabilities for latent trajectories for men and Tab le 5 11 shows descriptive statistics by trajectories. Overall, eight trajectories fit the data best identifying differences in the analytical sample regarding timing of transitions. Again, I leave formal discussion of model selection to later after discussion of trajectory characteristics. I used an assignment strategy for men similar to women (details above). bigger with a prevalence of 22.58% ( trajectory #5). These men also have the lowest likelihood of remaining childless. Only 6% of these men remained childless by age 44. M even denser clustering of trajectories in terms of lik elihood of remaining childless Trajectories #3 and #8 have much higher rates of childlessness than all other trajectories. However, even trajectories where men are least likely to likely trajector ies. The first group includes trajectories #5, #6, #7, and #8, which complete their school leaving around age 21. This behavior can also be observed in the descriptive statist ics. Most of men who belong to these trajectories are either high school dropout or graduates or have some college education Very few of these men are college graduates and almost none of them have higher education. The second group, trajectories #1 and #3, left school around age 24 with some college education and

PAGE 81

81 college degrees. Finally, third group, trajectories #2 and #4, is composed of men who continue their education until early 30s. For instance, trajectory #2 includes men of whom 38% have higher e ducation. In sum, men are more diverse in their school leaving behavior compared to women. in tandem with school leaving : synchronized, delayed, and no transition to labor market. T he synchronized category include s trajectories #2, #3, #4, #5, and #6. These men experience much unemp loyment throughout their lives. Descriptive statistics also confirm this hours worked equivalent of full time ranging from 94.7% to 99.4%. Delayed category is trajectories #1 and #7 with a transition to labor market much later than school leaving. They never fu lly make transition to labor market and remain marginally attached to the labor market. Finally, trajectory #8 is composed of men who never actually make transition to labor market and remain unattached to the labor market throughout. Men show transition to first marriage similar to women with four categories of timing. These categories include early, normative, later and no transition. Early transition to marriage is seen in trajectories #5 and #7 with an average age at first marriage around 21. The secon d category, normative, is trajectories #2 and #6 and these men have an average age at first marriage of 25 to 26. The third category includes men who make transition to first marriage around their early 30s, trajectories #1 and #4. It is interesting to see that none of these trajectories includes men who never made transition. On the contrary, the final category includes those who never made

PAGE 82

82 transition to first marriage. T rajectories #3 and #8 include men who have a non marriage rate of 92.8% and 86.9%, res pectively. Considering these trajectories together reveals important insights to the interplay between employment and marital trajectories for men. Men who transition to first marriage with an increasing likelihood of high labor force attachment seem to be the least likely to remain childless (trajectories #2 and #5). The exception is trajectory #7 who never transition to labo r market but still get married, and have a low likelihood of remaining childless. The trajectory with highest frequency (22%) is traj ectory #5 with the lowest childlessness rate (6%). The trajectory is characterized by high school (62.9%) and some college (21.2%) education. Most of these men left school during their late teens (between 18 and 20) and they started working immediately. B y age 24, most of them had completed their transition into the labor market and working full time (high attachment). In the meantime, they had successfully transitioned into their first marriages as can be seen from the graph This trajectory can be called both qualitative and quantitative terms, since it is the most prevalent trajectory and it is compatible with the social expectations for men. Men who followed this path almost never left the labor force and spent the highest amount of time in marriage compared to other trajectories. Since social norms related to marriage, fatherhood, and employment (Townsend, 2002) it is no wonder that these men have also the lowest childlessness rate. The most notable trajectory in terms of childlessness is trajectory #3 with a childlessness rate of 63%. Considering that childlessness is more prevalent among men

PAGE 83

83 than women, the highest risk trajectory for men has also higher childlessness rate than highest risk trajectory for women (which was trajectory #6 with 56.7%). A casual look comparing the highest risk trajectories for men and women reveals that these are very similar, if not exact, trajectories formed by high school gr aduates and those who have some college education. In addition, both sexes made the transition from school to employment very fast and almost never left the labor force. However, between these two trajectories (#6 for women and #3 for men), women have some what higher education than men. This may be reflecting the differential educational requirements for women to attain the same level of employment with men for these cohorts. Finally, although both women and men in highest risk trajectories made transition to marriage quite late, women on average spent approximately 5 months more in their first marriage. Although men and women are similar in terms of their highest risk trajectory for childlessness, men have two higher risk trajectories unlikely women who hav e only one higher risk trajectory. Higher risk trajectory for women (#7) was characterized by college to employment with high labor force attachment and no union formation. The first higher risk trajectory for men (#4) has very similar characteristics. In fact, it can be seen that these two trajectories have almost identical childless rates (27.7% for women and 27.3% for men). On the other hand, a unique pathway to childlessness for men is trajectory #8. 44.1% of men who belong to that trajectory remained c hildless. These men have the highest rate of high school dropout and 55.9% of these men are high school graduates. Due to this lower educational attainment compared to other trajectories, these men have hard time finding jobs and remain lowly attached to t he labor market over the course of their life. This can also be seen from their average full

PAGE 84

84 time hours of employment which indicates that their worked approximately half of their available time (53.5%), the lowest among all other trajectories. They also d id not make for women. This is a clear example of gendered pathways of c hildlessness where stable Table 5 9 shows the summary of fit statistics for trajectories of school enrollment, labor force attachment and transition into the first marital union for men. BIC and AIC statistics for these models are much lower than for women indicating that LLCA model s fit me trajectories. I estimated up to 9 trajectories for men. The attempt to extract 9 trajectories o ut of these data was unsuccessful due to convergence issues even after random, automated starts over 1000. As BIC difference suggested, all models for men had even better discriminatory power among individuals. Entropy is always over .97. However, the mode l with 8 trajectories included one class with frequency less than 5% in the sample. From the model with 7 trajectories to the model with 8 trajectories, there emerged a new trajectory with highly volatile labor force attachment over the life course and a t ransition into the first marriage during 30s. Although a similar labor force attachment trajectory was present in the model with 7 trajectories, they never made transition into a marital union. Since this can allow me to compare the attenuating effect of m arital union when coupled with a volatile labor force attachment, I decided to keep this trajectory and to continue with 8 trajectories. Other trajectories were not drastically influenced

PAGE 85

85 from the increase of trajectories judging from trajectory characteri stics (descriptive statistics, not shown ). I next proceed by testing the effects of trajectory membership controlling for all other factors in the base model I have previously established for men. Table 5 12 shows the baseline model from Table 5 6 in addit ion to controls for trajectory membership I previously identified. I initially used membership to trajectory #2 as the reference category because it allowed many of the comparisons I would like to make. Note that trajectory #2 is characterized by l atest sc hool leaving later transition to marriage and e arly attachment to the labor force R esults show that, c ompared to trajectory #2, all other trajectories are m ore likely to remain childless. This provides some evidence that there are different effects of timing in education, labor force attachment and marital trajectories that influence ng childless at age 44. I now compare trajectories based on their different timings in specific trajectories. Later school leaving would be expe cted to lead to higher likelihood of childlessness for men. I test whether this is the case by comparing the trajectories # 2 #5 as both trajectories have similar marital transition and labor force attachment trajectories while they only differ by their sch ool leaving behavior. However, Table 5 12 (Model 2) shows that membership in these trajectories does not affect the likelihood of remaining childless differently for men. As a result, it can be concluded that timing of school leaving does not have an indep lihood of childlessness. One explanation would be that all the effects of timing of school leaving are expressed in the form of transition to first marriage

PAGE 86

86 B oth trajectories #2 and #4 show latest school leaving and early transi tion to the labor market but they differ in terms of their transition to first marriage. Whereas trajectory #2 transition to first marriage later around age 25, men in trajectory #4 trans ition to marriage around age 33. This allows me to test whether timin g of first marriage matters for the likelihood of remaining childless. As Model 2 shows, that is actually the case. Since the model controls for the direct effects of completed education, lifetime cumulative labor force participation, race, marital history family background and timing of school leaving and labor force transition, it is safe to conclude that later transition to marriage leads men to remain childless. More specifically, belonging to trajectory #4 increases the odds of remaining childless 1.6 times compared to trajectory #2. This is in contrast a l ater transition to first marriage did not matter for the likelihood of remaining childless. The same comparison can be made between trajectories #5 and #6. Both of these trajectories are early school leavers with early transition to labor market. However, while trajectory #5 is characterized by early transition to marriage, men who belong to trajectory #6 marry later than the former trajectory. I tested wheth er the effects of trajectory membership to these two trajectories differ by a joint test of equality. The Wald test result reveals that the effects differ from each other, 2 (1, N = 3131) = 4.15, p < .05. Again, this result provides some evidence that timi ng of marriage is critical in kelihood of remaining childless in contrast to women. I tested whether timing of transition to labor market influences the likelihood of remaining childless by comparison three pairs of trajectories, #5 #7 and #1 #3. The first pair is early school leaving and transition to first marriage trajectories. They differ only

PAGE 87

87 by the timing of their transition to labor force. The second pair, on the other hand, experiences later school leaving and later or no marriage, and they differ by the timing of labor force transitions too Joint tests of equality for each pair revealed nonsignificant results, concluding remaining childless independent of race, comple ted education, cumulative labor force participation, marita l history average family income, family background and timing of transition to first marriag e, and school leaving behavior. This finding does not confirm my hypothesis that men who experience an e arly unemployment period during their prime fertile years would more likely to be childless Considering that the effects of timing of transition to the labor market may be expressed I excluded marital history variab les (duration and number of spouses and coresidential partners) from the model and rerun the te st for the first pair since the second pair was different in terms of their marital history too T here was still no difference between the effects of class membe rship for the first p air which do es not provide evidence for the proposed explanation. I conclude that timing of transition to the labor market does not have any independent effect on tion and other covariates controlled in the model. Finally, i ncluding trajectory covariates do es not change the significant effects of any other controls in the base model. In addition, they do not influence model fit statistics (BIC or r squared) in any m ajor way. One interesting finding is that controlling for timing of transitions through trajectory membership turns the odds of cumulative labor market participation to negative, yet it still remains nonsignificant.

PAGE 88

88 I expected that men who have substantial decreases in their income during their prime years will be more likely to remain childless. I tested this hypothesis by looking at i nstead of using crude me asures of ave rage family income Model 3 shows that men who were in high category during their 20s and in low category during their 30s have the highest likelihood of remaining childless among others. More importantly, adding measures of income mobility turns the almos t all coefficients of trajectory membership to nonsignificant. This finding can mean two things. First, income dynamics explain the effects of timing in a major way. In other words, effects of timing in marriage and labor force attachment are expressed thr ough family income for men. Second, it may also mean that using that many dummy variables affects the discriminatory power of the model, as increased BIC statistics also show. Overall, these results suggest that men and women differ in terms of the effects of to first marriage does not have any effect on their likelihood of remaining childless, it ansition to labor market does not have any influence on their likelihood of remaining childless market increases their likelihood of remaining childless even after all the controls in the model. Finally, although income mobility over the life course does not influence higher likelihood of remaining childless compared to other men.

PAGE 89

89 Sensitivity Analysis I next estimated all LLCA trajectory models using different operationalizations of labor force attachment (see Method section for details) to see if results are sensitive to cut points that determine low/medium/high attachment. Table 5 13 (women) and Table 5 14 (men) shows membership correspondence between different models based on distinct operationalizat ions of labor force attachment. Both tables for trajectories show that different operationalizations of labor force attachment mostly agree on the classification of individuals to trajectories, and this is especially the case for the highest and higher risk trajectories for childlessness. For instance, o ut of all respondents who were in the highest risk group (trajecto ry #6) in the first wave of models I have estimated with cut points of 416/1560, 97.3% were still in the highest risk group in all other estimations with different operationalizations. As a result, it can be said that this trajectory is insensitive to the specification of labor force attachment. Similar correspondence was achieved for those respondents who were in the higher risk group (trajectory #7) in the first wave of models I have estimated with cut points of 416/1560. Table 5 14 shows similar results Only trajectories where part time labor force participation was influential were affected by reallocation and re specification (#1 and #7) and this was only the case when both cut points are changed. Overall, these results show that labor force attachment trajectories I used in the analysis are quite indifferent to specifications of low/medium/hig h attachment.

PAGE 90

90 Table 5 1 Descriptive s tatistics by sex and parenthood status at age 44, National Longitudinal Survey of Youth 1979 2010

PAGE 91

91 Table 5 2 T tests for differences on selected variables between parents and the childless by sex

PAGE 92

92 Table 5 3 Proportion t est s for differences on selected percentages between parents and the childless by sex

PAGE 93

93 Table 5 4 Mobility table with rates of childlessness based on average family i ncome before and after 30 for w omen Table 5 5 Mobility table with rates of childlessness based on average family income before and after 30 for m en

PAGE 94

94 Table 5 6 Odds ratios a nd standard errors of odds ratios from logistic regression models of remaining childless at or afte r age 44 on selected covariates by sex

PAGE 95

95 Figure 5 2 Estimated population prevalence and conditional role probabilities for latent trajectories for w omen, National Longitudinal Survey of Youth 1979 2010

PAGE 96

96 Table 5 7 Descriptive s Nation al Longitudinal Survey of Youth 1979 2010

PAGE 97

97 Table 5 8 Summary of fit statistics of school enrollment, labor force attachment and transition into first marital union trajectory models by number of t rajectories for w omen Table 5 9 Summary of fit statistics of school e nrollmen t, labor force attachment and transition into first marital union trajectory models by number of t rajectories for m en

PAGE 98

98 Table 5 10 Results from logistic regression m odels of remaining childless at or after age 44 on trajectory covariates for w omen

PAGE 99

99 Figure 5 3 Estimated population prevalence and conditional role probabilities for latent trajectories for m en, National Longitudinal Survey of Youth 1979 2010

PAGE 100

100 Table 5 11 Descriptive s tatistics by school attendance, labor force attachment and transition into the en, Nationa l Longitudinal Survey of Youth 1979 2010

PAGE 101

101 Table 5 12 Results from logistic regression m odels of remaining childless at or after age 4 4 on trajectory covariates for m en

PAGE 102

102 Table 5 13 Cross tabulation of trajectory membership between models based on different operationalizations of labor force attachment for w omen

PAGE 103

103 Table 5 14 Cross tabulation of trajectory membership between models based on different operationali zations of labor force attachment for m en

PAGE 104

104 CHAPTER 6 DISCUSSION The purpose of this thesis was to assess the effects of timing of school to work transition and the first marital union on the likelihood of remaining childless. It is methodologicall y challenging to account for the effects of incidence, duration, and timing of life events simultaneously I suggested that Longitudinal Latent Class Analysis is a tractable method to account for heterogeneity in terms of timing of transitions over the lif e course. I estimated trajectories of education, labor force attachment and marital status to represent consequential effects of differences and similarities between different timing patterns of transitions on the likelihood of remaining childless for wome n and men, separately. Since trajectories show simultaneous transitions in different life domains (i.e. education, employment, marriage), they can be considered as biographical representations defined by the timings of multiple different states. This thesi s confirmed some of the findings other studies either presented or speculated about. First, relationship history is clearly associated with fertility. Specifically, duration of first and second marriages significantly decreases the likelihood of remaining childless for both men and women after controlling for race, completed education, employment history, average income, and family background. In addition, incidence of coresidential partnership seems to be associated with lower likelihood of Second, cumulative effects of employment are not as consequential for men as they are for women. Women who worked more on average over their fertile years, i.e., betwe en the ages 18 and 44, are more likely to remain childless. However, cumulative

PAGE 105

105 marriage and I found some support for this proposition. The effect of cumulative average hours worked was highly significant unless the time spent in first and second marriages are controlled for. Third, family background is influential on the likelihood of remaining childless for both men and women, but the effects are somewhat larger for men. In addition, family background does not affect men and women in the same way. For instance, while he reverse is true for men. The number of siblings has the same negative effect on the likelihood of remaining childless for men and women. Finally, religious background does not seem to have much effect on the likelihood of remaining childless after contr olling for race, completed education, marital and employment history, and family background. The only significant difference was between Catholic and Protestant women where the latter was more likely than the former to remain childless. I found that timing of transitions in different spheres of life have different bearings for men and women. More specifically, timing of transition to the labor force was highly independent effect s for men. Men seem to be able to compensate for the time they spent outside of labor force later in life while women who transition to high labor attachment directly after school and remain attached are more likely to remain childless than their counterpa rts. This finding provides some support for the argument that

PAGE 106

106 childbearing, especially the first one, is more costly for women in terms of time and economic returns than it is for men. Contrarily, the timing of first marriage was not significantly associat ed with to remaining childless. Women may be recuperating the time they spent outside marriage during early years of life after age 30 while for men, that does not s eem to be the case. Overall, these findings suggest that there are distinct origins of childlessness in terms of employment and family building behavior. Not only incidence but also timing of transitions is important for fertility outcomes. Additionally, e arly life decisions (e.g., when to get married) have long lasting effects on whether someone will remain childless at the end of their fertile years. The timing of transitions is important due to two main reasons. First, marriage eligibility and desire is highly time dependent. The highly educated are most likely to delay marriage (Goldstein & Kenney, 2001) probably due to later school leaving patterns as the results of this thesis showed. In addition, desire to marry decreases by age (Mahay & Lewin, 2007) Finally, those who marry later seem to have lower quality marriages, which can in turn increase the likelihood of remaining childless if the spouses have not had children by that time (Glenn et al., 2010) The shapes of marriage trajectories partially confirmed this result. Women and men who were not married until 30s were less likely to transition into the first marriage. In return, they were also more likely to remain childless.

PAGE 107

107 Second, unemployment can have lifelong scarring effects, especially when it occurs during ages when u nion formation is most likely, i.e., 20s and early 30s. Since employment and potential earnings are consequential for marriage and fertility eligibility (Kravdal, 2002; Meron et al., 2002; Neumark, 2002; zcan et al., 2010) spells of short or long unemployment during these ages may lead to higher likelihood of remaining childless. In addition, e conomic downturn during these fertile years can lead individuals to postpone having children (Sobotka et al., 2011) or getting married (Lichter et al., 2002) Finally, these lifelong effects can be even more consequential for men. As the to work transitions are generally intertwined with their e mployment patterns. Those men who transition to high labor force attachment but delay union formation in the meantime are less likely to ever transition to the first marriage. This was not as strongly the case for women as it is for men. What do pathways t o childlessness inform us about? Studies which focus on certain outcomes of childlessness at later life, such as happiness or civic engagement, generally do not take into account for the critical periods of union formation and employment patterns (Bures et al., 2009; Dykstra & Wagner, 2007; Keizer et al., 2010) As a result, many of those outcomes that seem to be associated with childlessness may be related to factors that lead to childlessness in the first place. In addition, the method I used here provides interlocked trajectories many other methodological approaches such as regression or event history modeling cannot. There are four main limitation s in this study. The first limitation is related to methodological assumptions of LLCA modeling. Since there is no one formal fit statistic to decide on the number of trajectories, it heavily depends on the judgment of the

PAGE 108

108 researcher. However, I used all a vailable information (BIC, AIC, frequencies, substantive theory) to decide on the number of trajectories I extracted. In addition, since the method I used is data driven, trajectories would not be exactly identifiable using another data. Scholars who use o ther longitudinal data would still be able to identify similar pathways to childlessness. Second, trajectories I used do not represent changes in behavior in union formation in cohabitations although I control for the direct effects of the number of partners and spouses. One mitigating fact is that the cohorts studied here were born between years 1957 1964 and cohabitation was not as common as it is presently when these respondents were in their prime fertile years. However, I suggest that future stud ies that focus on more recent cohorts can incorporate trajectories of cohabitation by extending marital union to a three category variable to indicate single, married or cohabiting. Third, spousal characteristics (except family income) are not controlled f or in logistic models and there is some evidence that they might be influential in terms of explaining fertility behavior through the presence of stepchildren (Stewart, 2002) attitu des towards the timing of the birth of first child (Corijn et al., 1996; Jansen & Liefbroer, 2006) and religious homogamy (Krishnan, 1993) However, the analysis showed that childlessness is mostly due to a lack of partnership and this limitation applies to cases of within union childlessness. To avoid this pitfall, further research shou ld focus on within union childlessness by looking at the marriages that ended without childbearing and see whether those marriages differ in terms of certain

PAGE 109

109 characteristics (e.g., family background, racial and religious homogamy, and spells of unemploymen t) from marriages that do produce children. Finally, since the analytical sample consists of individuals who were born around the same years, it is hard to distinguish whether the effects I found in this thesis are period or cohort driven. We know that chi ldlessness have become more accepted in the United States as a life style. This is apparent in the results of studies which focus on attitudes toward childlessness. This can also be seen in media products revolving around voluntary childlessness. Books suc (Walker, 2011) (Defago, 2005) an ( Foster, 2011) are only a couple of recent publications that directly address this increasing phenomena in popular media. Starting with the early 2000s, there has been increasing attention to a movement iduals who demand equality between parents and the childfree. They are highly critical of pronatalistic practices at the federal level such as parental leave and tax breaks for children. This movement finds its voice through websites (e.g., http://thechildfreelife.com/ http://www.childfree.net/ ), forums and social networks. For instance, there were about 1900 members of a group called of this writing. What do these new developments imply for the future of childlessness in the United States? As my analysis has shown, there were actual pathways that lead to remaining childlessness for cohorts born during 1950s and 1960s. For these cohorts to work transition with family building behavior, and this was especially the case for

PAGE 110

110 women. In addition, the success in those transitions was dependent on social class in terms of education and family background. The most important point of my research is that it shows the emergence of this stratification in family forms occurring over a lifetime of experiences in the labor force. However, as attitudes in society are cha nging, more recent cohorts may have more options regarding their fertility. As a result, I expect to see that these pathways become blurry.

PAGE 111

111 LIST OF REFERENCES Abma, J. C., & Martinez, G. M. (2006). Child lessness Among Older Women in the Journal of Marriage and Family 68 1045 1056. Acock, A. C. (2005). Working With Missing Values. Journal of Marriage and Family 67 (4), 1012 1028. Alkema, L., Raftery, A. E., Gerland, P., Clark, S. J., Pelletier, F., Buettner, T., & Heilig, G. K. (2011). Probabilistic Projections of the Total Fertility Rate for All Countries. Demography 48 (3), 815 839. Amato, P. R., Landale, N. S., Havasevich Brooks, T. C., Booth, A., Eggebeen, D. J., Formation Pathways. Journal of Marriage and Family 70 (5), 1271 1286. Bachu, A. (1996). Fertility of American Men U.S. Bureau of the Census. Baizan, P., Aassve, A., & Billari, F. C. (2003). Cohabitation, Marriage, and First Birth: The Interrelationship of Family Formation Events in Spain. European Journal of Population/Revue europeenne de demographie 19 (2), 147 169. Becker, G. S. (1981). A Treatise on the Family Cambridge: Harvard Univ. Press. Becker, G. S., & Barro, R. J. (1986). Altruism and the Economic Theory of Fertility. Population and Development Review 12 69 76. Bianchi, S. M. (2000). Maternal Employment and Time with Children: Dramatic Change or Surprising Continuity? Demography 37 (4), 401 414. Bianchi, S. M. (2011). Family Change and Time Allocation in American Families. The ANNALS of the American Academy of Political and Social Science 638 (1), 21 44. Bianchi, S. M., Milkie, M. A., Sayer, L. C., & Robinson, J. P. (2000). Is Anyone Doing the Housework? Trends in the Gender Division of Household Labor. Social Forces 79 (1), 191 228. Blau, D. M., & Robins, P. K. (1989). Fertility, Employment, A nd Child Care Costs. Demography 26 (2), 287 299. Blossfeld, H. P., & Huinink, J. (1991). Human Capital Investments or Norms of Role Formation. American Journal of Sociology 97 (1), 1 43 168. Bongaarts, J. (2002). The End of the Fertility Transition in the Developed World. Population and Development Review 28 (3), 419 443.

PAGE 112

112 Bongaarts, J., & Feeney, G. (1998). On the Quantum and Tempo of Fertility. Population and Development Review 24 (2) 271 291. Brand, J. E., & Davis, D. (2011). The Impact of College Education on Fertility: Evidence for Heterogeneous Effects. Demography 48 (3), 863 887. Industrialized Nations Annual Review of Sociology 26 271 296. Demographic and Institutional Change. Demography 27 (4), 483 498. Bumpass, L., & Lu, H. Family Contexts in the United States. Population Studies 54 (1), 29 41. Bureau of Labor Statistics. (2010, September). Number of Jobs, Labor Market Experience, and Earnings Growth: R esults from a National Longitudinal Survey News Release. Bureau of Labor Statistics. (2011). Women in the Labor Force: A Databook (2011 Edition) ( No. Report 1034). U.S. Department of Labor. Bures, R. M., Koropeckyj Cox, T., & Loree, M. (2009). Childlessne ss, Parenthood, and Depressive Symptoms Among Middle Aged and Older Adults. Journal of Family Issues 30 (5), 670 687. Carlson, M., McLanahan, S., & England, P. (2004). Union Formation in Fragile Families. Demography 41 (2), 237 261. Carmichael, G. A., & Wh ittaker, A. (2007). Choice and circumstance: Qualitative insights into contemporary childlessness in Australia. European Journal of Population 23 111 143. CDC/NCHS National Vital Statistics. (2012). NVSS National Marriage and Divorce Rate Trends. Cherl in, A. J. (2010). Demographic Trends in the United States: A Review of Research in the 2000s. Journal of Marriage and Family 72 403 419. Clark, R. L., Naohiro, O., Kondo, M., & Matsukura, R. (2010). Population Decline, Labor Force Stability, and the Fu ture of the Japanese Economy. European Journal of Population 26 207 227. Cooksey, E. C., & Rindfuss, R. R. (2001). Patterns of Work and Schooling in Young Adulthood. Sociological Forum 16 (4), 731 755.

PAGE 113

113 Corijn, M., Liefbroer, A. C., & de Jong Gierveld, J. (1996). It Takes Two to Tango, the First Child. Journal of Marriage and the Family 58 (1), 117. Crenshaw, E., & Rabison, K. (2010). Socio demographic Determinants of Economi c Growth: Age Structure, Preindustrial Heritage and Sociolinguistic Integration. Social Forces 88 (5), 2217 2240. Davis, K. (1963). The Theory of Change and Response in Modern Demographic History. Population Index 29 (4), 345 366. Defago, N. (2005). Childfree and loving it! London: Fusion. DeOllos, I. Y., & Kapinus, C. A. (2002). Aging Childless Individuals and Couples: Suggestions for New Directions in Research. Sociological Inquiry 72 (1), 72 80. P., & Rohwer, G. (1999). D Patterns over the Family Life Course: A Comparison of the United States and Germany. Journal of Marriage and Family 61 (1), 133 146. Dye, J. L. (2010). Fertility of American Women: 2008 ( No. P20 563). U.S. Census Bureau. Dyks tra, P. A. (2006). Off the Beaten Track: Childlessness and Social Integration in Late Life. Research on Aging 28 (6), 749 767. Dykstra, P. A., & Wagner, M. (2007). Pathways to Childlessness and Late Life Outcomes. Journal of Family Issues 28 (11), 1487 151 7. Elder, G. H., Jr. (1994). Time, Human Agency, and Social Change: Perspectives on the Life Course. Social Psychology Quarterly 57 (1), 4 15. Elder, G. H., JR. (1995). The Life Course Paradigm: Social Change and Individual Development. Examining Lives in Context: Perspectives on the Ecology of Human Development (Vol. APA Science, pp. 101 139). Washington: APA Press. Esping Andersen, G. (1996). Welfare states without work: the impasse of labour shedding and familialism in Continental European social policy. Welfare States in Transition: National Adaptations in Global Economies (pp. 66 87). London: SAGE Publications. Esping Andersen, G. (1999). Social Foundations of Post Industrial Economies New York: Oxford Univ. Press. Fieder, M., Huber, S., & Bookstein F. L. (2011). Socioeconomic Status, Marital Status and Childlessness in Men and Women: An Analysis of Census Data from Six Countries. Journal of Biosocial Science 43 619 635.

PAGE 114

114 Foster, K. (2011). decision not to have children [United States]: s.n. Frejka, T., & Westoff, C. F. (2007). Religion, Religiousness and Fertility in the US and in Europe. European Journal of Population / Revue europenne de Dmographie 24 (1), 5 31. Gangl, M., & Ziefle, A. Careers: An Empirical Assessment of the Wage Penalty for Motherhood in Britain, Germany, and the United States. Demography 46 (2), 341 369. Glauber, R., & Gozjolko, K. L. (2011). Do Traditional Fathers Always Work More? Gender Ideology, Race, and Parenthood. Journal of Marriage and Family 73 (5), 1133 1148. Glenn, N. D., Uecker, J. E., & Love, R. W. B. (2010). Later first marriage and marital success. Social Science Research 39 (5), 787 800. Goldin, C., & Katz, L. F. (2000). Career and Marriage in the Age of the Pill. The American Economic Review 90 (2), 461 465. Goldstein, J. R., & Kenney, C. T. (2001). Marriage Delayed or Marriage Forgone? New Cohort Forecasts of First Marriage for U.S. Women. American Sociological Review 66 (4), 506. Gonzlez, M. J., & Jurado Economies: A Comparison of France, West Germany, Italy and Spain, 1994 2001. European Journal of Population 22 317 352. Goodman, L. A. (1974). Exploratory latent structure analysis using both identifiable and unidentifiable models. Biometrika 61 (2), 215 231. Guzzo, K. B. (2006). The Relationship Between Life Course Events and Union Formation. Social Science Research 35 384 408. Hayford, S. R. (2009). The Evolution of Fertility Expectations Over the Life Course. Demography 46 (4), 765 783. Heaton, T. B., Jacobson, C. K., & Holland, K. (1999). Persistence and Change in Decisions to Remain Childless. Journal of Marriage and Family 61 (2), 531 539. Houseknecht, S. K. (1987). Voluntary Childlessness. Handbook of Marriage and the Family New York: Plenum Press. Iacovou, M., & Tavares, L. P. (2011). Yearning, Learning, and Conceding: Reasons Men and Women Change Their Childbearing Intentions. Populatio n and Development Review 37 (1), 89 123.

PAGE 115

115 Inglehart, R. F. (2008). Changing Values among Western Publics from 1970 to 2006. West European Politics 31 (1 2), 130 146. Labor. Journal of Family Issues 27 (11), 1487 1511. Johnson, D. R., & Young, R. (2011). Toward Best Practices in Analyzing Datasets with Missing Data: Comparisons and Recommendations. Journal of Marriage and Family 73 (5), 926 945. Kaa, D. J. van de. (1987). ( No. 41 (1)). Population Bulletin. Washington: The Population Reference Bureau. Kaa, D. J. van de. (2001). Postmodern Fertility Preferences: From Changing Value Orientation to New Behavior. Population and Development Review 27 290 331. Keizer, R., Dykstra, P. A., & Jansen, M. D. (2007). Pathways Into Childlessness: Evidence of Gendered Life Course Dynamics. Journal of Biosocial Science 40 863 878. Keizer, R., Dykstra, P. A., & Poortman, A. R. (2010). Life Outcomes of Childless Men and Fathers. European Sociological Review 26 (1), 1 15. Kirmeyer, S., & Hamilton, B. (2011). Transitions Between Childlessness and First Birth: Three Generations of U.S. Women ( No. 153). National Center for He alth Statistics. Kneale, D., & Joshi, H. (2008). Postponement and childlessness Evidence from two British cohorts. Demographic Research 19 1935 1968. Knoester, C., & Eggebeen, D. J. (2006). The Effects of the Transition to Parenthood and Subsequent Chi Being and Social Participation. Journal of Family Issues 27 (11), 1532 1560. Kohler, H. P., Billari, F. C., & Ortega, J. A. (2002). The Emergence of Lowest Low Fertility in Europe During the 1990s. Population and Development Review 28 ( 4), 641 680. Koropeckyj Cox, T., & Call, V. R. A. (2007). Characteristics of Older Childless Persons and Parents Cross National Comparisons. Journal of Family Issues 28 (10), 1362 1414. Koropeckyj Cox, T., & Pendell, G. (2007). Attitudes About Childlessnes s in the United Journal of Family Issues 28 (8), 1054 1082. Kravdal, (2002). The impact of individual and aggregate unemployment on fertility in Norway. Demographic Research 6 (10), 263 2 94.

PAGE 116

116 Krishnan, V. (1993). Religious Homogamy and Voluntary Childlessness in Canada. Sociological Perspectives 36 (1), 83 93. Landale, N. S., Schoen, R., & Daniels, K. (2010). Early Family Formation Among White, Black, and Mexican American Women. Journal of Family Issues 31 (4), 445 474. Lichter, D. T., McLaughlin, D. K., & Ribar, D. C. (2002). Economic Restructuring and the Retreat from Marriage. Social Science Research 31 (2), 230 256. Lino, M. (2008). Expenditures on children by families, 2007 Washington, DC: U.S. Department of Agriculture, Center for Nutrition Policy and Promotion. Livingston, G., & Cohn, D. (2010). Childlessness Up Among All Women; Down Among Women with Advanced Degrees Pew Research Center. Lundberg, S., & Supply and Wages. Review of Economics and Statistics 84 (2), 251 268. Macmillan, R., & Copher, R. (2005). Families in the Life Course: Interdependency of Roles, Role Configurations, and Pat hways. Journal of Marriage and Family 67 (4), 858 879. Mahay, J., & Lewin, A. C. (2007). Age and the Desire to Marry. Journal of Family Issues 28 (5), 706 723. Marsiglio, W., & Hinojosa, R. (2007). Social and psychological influences on male procreative ex periences. In F. R. Kandeel, R. S. Swerdloff, & J. L. Pryor (Eds.), Male reproductive dysfunction: Pathophysiology and treatment (pp. 129 140). NY: Informa Healthcare USA, Inc. Martin, J. A., Hamilton, B. E., Sutton, P. D., Ventura, S. J., Mathews, T. J., & Osterman, M. J. K. (2010). Births: Final Data for 2008 National Vital Statistics Reports. U.S. Department of Health and Human Services. Martin, S. P. (2000). Diverging Fertility Among U.S. Women Who Delay Childbearing Past Age 30. Demography 37 (4), 523 533. Martinez, G. M., Chandra, A., Abma, J. C., Jones, J., & Mosher, W. D. (2006). Fertility, Contraception, and Fatherhood: Data on Men and Women From Cycle 6 (2002) of the National Survey of Family Growth National Center for Health Statistics. May, K. A. (1982). Factors Contributing to First Fatherhood: An Exploratory Study. Family Relations 31 (3), 353 361. McDonald, P. (2000). Gender Equity, Social Institutions, and the Future of Fertility. Journal of Population Research 1 7 (1), 1 16.

PAGE 117

117 McDonald, P., & Kippen, R. (2001). Labor Supply Prospects in 16 Developed Countries, 2000 2050. Population and Development Review 27 (1), 1 32. Meron, M., Widmer, I., & Shapiro, D. (2002). Unemployment Leads Women to Postpone the Birth of Their First Child. Population (English Edition, 2002 ) 57 (2), 301 330. Miller, A. R. (2009). The effects of motherhood timing on career path. Journal of Population Economics 24 (3), 1071 1100. Morgan, S. P. (1996). Characteristic Features of Modern American Fertility. Population and Development Review 22 19 63. Morgan, S. P., & Rackin, H. (2010). The Correspondence Between Fertility Intentions and Behavior in the United States. Population and Development Review 36 (1), 91 118. Morgan, S. P., & Taylor, M. G. (2006). Low Fertility at the Turn of the Twenty First Century. Annual Review of Sociology 32 375 399. Murphy, M. J., & Wang, D. (2001). Family Level Continuities in Childbearing in Low Fer tility Societies. European Journal of Population 17 (1), 75 96. Murphy, M., & Knudsen, L. B. (2002). The intergenerational transmission of fertility in contemporary Denmark: The effects of number of siblings (full and half), birth order, and whether male o r female. Population Studies 56 (3), 235 248. Musick, K. (2002). Planned and Unplanned Childbearing Among Unmarried Women. Journal of Marriage and Family 64 915 929. Musick, K. (2007). Cohabitation, nonmarital childbearing, and the marriage process. Demo graphic Research 16 249 286. Muthn, B. O. (2004). Latent variable analysis: Growth mixture modeling and related techniques for longitudinal data. In D. Kaplan (Ed.), Handbook of quantitative methodology for the social sciences (pp. 345 368). Newbury Par k, CA: SAGE Publications. Muthn. Nagin, D. S. (2005). Group Based Modeling of Development Cambridge, MA: Harvard Univ. Press. National Center for Health Statistics. (200 8). Adoption experiences of women and men, and demand for children to adopt by women 18 44 years of age in the United States, 2002. Hyattsville, Md.: U.S. Dept. of Health and Human Services, Centers for Disease Control and Prevention, National Center for H ealth Statistics.

PAGE 118

118 Neumark, D. (2002). Youth Labor Markets in the United States: Shopping Around vs. Staying Put. Review of Economics and Statistics 84 (3), 462 482. Notestein, F. W. (1945). Population the long view. In T. W. Schultz (Ed.), Food for t he World (pp. 37 57). Chicago: Cambridge University Press. Nylund, K. L., Asparouhov, T., & Muthn, B. O. (2007). Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Structural Equation M odeling 14 (4), 535 569. OECD Family Database. (2011, October 21). PF3.1: Public spending on childcare and early education. OECD Social Policy Division Directorate of Employment, Labour and Social Affairs. Pathways to Adulthood and Their Adolescent Precursors. Journal of Marriage and Family 72 (5), 1436 1453. Cross National Contexts of Gender. Journal of Marriage and Family 65 (2), 275 286. zcan, B., Mayer, K. U., & Luedicke, J. (2010). The impact of unemployment on the transition to parenthood. Demographic Research 23 807 846. Park, K. (2005). Choosing Chi the Voluntarily Childless. Sociological Inquiry 75 (3), 372 402. Parr, N. (2005). Family background, schooling and childlessness in Australia. Journal of Biosocial Science 37 (2), 229 243. Parr, N. (201 0). Childlessness Among Men in Australia. Population Research and Policy Review 29 (3), 319 338. Employment Rates from 1960 to 2005. American Sociological Review 73 (3), 497 517. Pickles, A., & Croudace, T. (2010). Latent mixture models for multivariate and longitudinal outcomes. Statistical Methods in Medical Research 19 271 289. Quesnel Valle, A., & Morgan, S. P. (2003). Missing the Target? Correspondence of Fertility Intentio ns and Behavior in the U.S. Population Research and Policy Review 22 497 525. Raftery, A. E. (1995). Bayesian Model Selection in Social Research. Sociological Methodology 25 111 163.

PAGE 119

119 Raley, R. K., & Bumpass, L. (2003). The topography of the divorce pla teau. Demographic Research 8 245 260. Rendall, M., Aracil, E., Bagavos, C., Couet, C., DeRose, A., DiGiulio, P., Lappegard, T., et al. (2010). Increasingly heterogeneous ages at first birth by education in Southern European and Anglo American family poli cy regimes: A seven country comparison by birth cohort. Population Studies 64 (3), 209 227. Riddell, W. C., & Song, X. (2011). The impact of education on unemployment incidence and re employment success: Evidence from the U.S. labour market. Labour Economics 18 (4), 453 463. Rindfuss, R. R., Guzzo, K. B., & Morgan, S. P. (2003). The changing institutional context of low fertility. Population Research and Policy Review 22 411 438. Rindfuss, R. R., Morgan, S. P., & Offutt, K. (1996). Education and th e Changing Age Pattern of American Fertility: 1963 1989. Demography 33 (3), 277 290. Rindfuss, R. R., Morgan, S. P., & Swicegood, G. (1998). First Births in America: Changes in the Timing of Parenthood University of California Press. Rindfuss, R. R., Swic egood, C. G., & Rosenfeld, R. A. (1987). Disorder in the Life Course: How Common and Does It Matter? American Sociological Review 52 (6), 785 801. Rowland, D. T. (1998). Consequences of Childlessness in Later Life. Australasian Journal on Ageing 17 (1), 24 28. Rowland, D. T. (2007). Historical Trends in Childlessness. Journal of Family Issues 28 (10), 1311 1337. Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods 7 (2), 147 177. Schoen, R., Landale, N. S., & Daniels, K. (2007). Family Transitions In Young Adulthood. Demography 44 (4), 807 820. Schoen, R., Landale, N. S., Daniels, K., & Cheng, Y. H. A. (2009). Social Background Differences in Early Family Behavior. Journal of Marriage and Family 71 (2) 384 395. Labor Force Exit. Journal of Marriage and Family 73 (1), 250 263. Shanahan, M. J. (2000). Pathways to Adulthood in Changing Societies: Variability and Mechanisms in Life Course Perspective. Annu. Rev. Sociol. 26 (1), 667 692.

PAGE 120

120 Sobotka, T. (2004). Is Lowest Low Fertility in Europe Explained by the Postponement of Childbearing? Population and Development Review 30 (2), 195 220. Sobotka, T. (2010). Shifting Parenthood to Advanced Reproductive Ages: Trends, Causes and Consequences. A Young Generation Under Pressure? (pp. 129 154). Springer. Sobotka, T., Skirbekk, V., & Philipov, D. (2011). Economic Recession and Fertility in the Developed World. Population and Developmen t Review 37 267 306. Social Security and Medicare Trustees. (2011). The Trustees summary of the 2011 Social Security and Medicare Trustees Reports Somers, M. D. (1993). A Comparison of Voluntarily Childfree Adults and Parents. Journal of Marriage and Family 55 (3), 643 650. Staff, J., & Mortimer, J. T. (2011). Explaining the Motherhood Wage Penalty During the Early Occupational Career. Demography 49 (1), 1 21. Stewart, S. D. (2002). The Effect of Stepchildren on Childbearing Intentions and Births. Demo graphy 39 (1), 181 197. Thornton, A., Axinn, W. G., & Teachman, J. D. (1995). The Influence of School Enrollment and Accumulation on Cohabitation and Marriage in Early Adulthood. American Sociological Review 60 (5), 762 774. Thornton, A., & Young DeMarco, L. (2001). Four Decades of Trends in Attitudes toward Family Issues in the United States: The 1960s through the 1990s. Journal of Marriage and Family 63 (4), 1009 1037. Toossi, M. (2002). A Century of Change: the U.S. Labor Force, 1950 2050 (pp. 15 28). Bu reau of Labor Statistics. Townsend, N. W. (2002). lives Philadelphia: Temple University Press. U.S. Census Bureau. (2010). Fertility of American Women: 2008 ( No. P20 563). Current Population Repo rts. Washington, DC: U.S. Census Bureau. U.S. Census Bureau. (2011). Table SF1. Percent Childless and Births per 1,000 Women in the Last Year: Selected Years, 1976 to 2010. U.S. Census Bureau. (2011). Number, Timing, and Duration of Marriages and Divorces : 2009 ( No. P70 125). Current Population Reports. Washington, DC: U.S. Census Bureau.

PAGE 121

121 U.S. Department of Education. (2010). Enrollment rates of 18 to 24 year olds in degree granting institutions, by type of institution and sex and race/ethnicity of stude nt: 1967 through 2009. Uecker, J. E., & Stokes, C. E. (2008). Early Marriage in the United States. Journal of Marriage and Family 70 (4), 835 846. Uhlenberg, P. (1974). Cohort Variations in Family Life Cycle Experiences of U. S. Females. Journal of Marriag e and Family 36 (2), 284 292. United Nations. (2011). World Population Prospects: The 2010 Revision, CD ROM Edition. United Nations, Department of Economic and Social Affairs, Population Division. Vincent, G. K., & Velkoff, V. A. (2010). THE NEXT FOUR DECADES: The Older Population in the United States: 2010 to 2050 U.S. Census Bureau. Walker, E. L. (2011). choice or by chance Austin, TX: Greenleaf Book Group Press. Wenger, G. C., Dykstr a, P. A., Melkas, T., & Knipscheer, K. C. P. M. (2007). Social Embeddedness and Late Support Networks. Journal of Family Issues 28 (11), 1419 1456. Wildeman, C., & Percheski, C. (2009). Associations of Childhood Religious Attendance, Family Structure, and Nonmarital Fertility Across Cohorts. Journal of Marriage and Family 71 (5), 1294 1308. World Bank. (2012). World Development Indicators: United States. Xie, Y., Raymo, J. M., Goyette, K., & Arland Thorn ton. (2003). Economic Potential and Entry into Marriage and Cohabitation. Demography 40 (2), 351 367. Zhang, L. (2008). Religious affiliation, religiosity, and male and female fertility. Demographic Research 18 233 262.

PAGE 122

122 BIOGRAPHICAL SKETCH Ozcan Tunalilar graduated from Koc University in 2010 with Bachelor of Art s in business administration and s ociology. He worked as an undergraduate assistant in a project called E arly Childhood E cologies in Turkey (TECGE) dur ing his studies. He currently serves as teaching assistant at the Department of Sociology and Criminology and Law at the University of Florida He is continuing his graduate studies to date.