giant‘- 1
1.3.1 .l...l.«!t¢(rl$ a.
{:1 .1}: 3%... y

N l
. :3

' ' km

a». «a

ti

 

 

mm
Mathis“ 5““

University

This is to certify that the
dissertation entitled

WOMEN, WORK AND CHILDBEARING

presented by
MARY K. HAMMAN

has been accepted towards fulfillment
of the requirements for the

Industrial Relations
Ph.D. degree in and Human Resources

 

 

iii/T, (EM
Major Professor’s‘fignature

AW \9; 2,008

Date

MSU is an affirmative-action, equal-opportunity employer

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

      
       

  

DATE DUE

 

I“\TE DUE DATE DUE

WOMEN, WORK AND CHILDBEARING
By

Mary K. Hamman

A DISSERTATION

Submitted to
Michigan State University
In partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Industrial Relations and Human Resources

2008

 

Women, Work and Childbearing
BY
Mary K. Hamman

The growing work-family literature provides extensive workplace level evidence
of positive relationships between working-time flexibility policies and practices and work
and family outcomes. Yet, little is known about the role of flexible policies and practices
in the labor market as a whole or with respect to behavioral rather than psychological
outcomes. This dissertation investigates relationships between availability of paid and
unpaid leave and expected and actual childbearing, maternal employment and early
investments in child health through well-baby care.

In Chapter 1, I provide a broad, cohort level analysis of birth expectations. Birth
expectations may foreshadow future fertility and shape current behaviors, including early
career and educational choices. Chapter 1 evaluates these two possibilities. Findings
indicate women do not anticipate their future childbearing very well. Also, the marked
differences in occupational characteristics between mothers and non-mothers, which are
well known in the literature and apparent in my analysis, indicate many women do
eventually sort into more “family friendly occupations”. Yet, my findings imply sorting
occurs after childbearing rather than before.

Chapter 2 extends the findings in Chapter 1 by examining relationships between
availability of paid and unpaid leave in the pre-birth job and other job characteristics and
mothers’ decisions to maintain or quit those jobs. I examine all quits from one year prior
to pregnancy through 1.5 years following the birth and distinguish between quits to leave

the labor force and quits to start a new job. Findings indicate most labor force exits are

concentrated in the three months either side of the birth. Most job changing happens just
before and during early pregnancy and would be missed in a shorter analysis interval.
Women eligible for FMLA leave are less likely to quit their jobs for any reason prior to
the birth and those who work part-time are less likely to change jobs before the birth. I
do find women without paid vacation are more likely to change jobs prior to the birth but
overall the evidence of pre-birth sorting into jobs with paid leave is not very compelling.

Chapter 3 also examines the effect of paid and unpaid leave on mother’s behavior,
but in this chapter my focus shifts from employment decisions to child outcomes. The
American Academy of Pediatrics (AAP) recommends children receive eight well-baby
visits at regular intervals over the first two years of life. I estimate the average baby
receives just over 2. Cost sharing for well-baby care under public and private insurance
is very low. The fact that compliance rates are so low despite the low cost of care
suggests other factors, such as time constraints, may be especially important. In general
my findings imply the type of job a mother holds matters; paid and unpaid leave may
enable mothers in certain types of jobs to take their babies to the doctor but in others they
appear to have no relationship or even a negative relationship with well-baby care use.

In total, the results of this dissertation suggest access to paid leave may help
women to maintain job matches during childbearing years and improve health outcomes
for young children by encouraging mothers to take their children to well-baby care. Yet,
the extent to which paid and unpaid leave influence outcomes of interest may depend on
the context of the job. Furthermore, despite the potential benefits of paid leave, I find
little evidence to suggest women actively seek jobs with paid leave before they have a

child.

Dedication

This dissertation is dedicated to my grandmother, Bea Hoffman, who was a single,
working mother in the 19505.

Acknowledgements

This dissertation was supported in part by the Health Institutions and Policy
Initiative and the Family Research Initiative at Michigan State University.

I thank my committee, Peter Berg, Dale Belman, Mark Roehling, Steven Haider
and Stacy Dickert-Conlin, for their support and guidance. In particular, I thank Peter
Berg, my committee chair, for teaching me to apply my interdisciplinary training to my
research and Dale Belman for his guidance since I first began my doctoral studies.

Successfully completing an interdisciplinary degree requires the support and
cooperation of other departments. I am indebted to the students and faculty of the MSU
Economics Department, who welcomed me in their classrooms and shared their resources
and expertise with me. I especially thank Steven Haider and Stacy Dickert-Conlin for
their invaluable feedback as external members of my committee.

I thank my friends and family. My husband, David Hamman, spent countless
hours talking through ideas and reading drafts for me. I look forward to returning the
favor as he completes his degree. My friend Gregory Derr provided much appreciated
moral support and editing assistance. I thank my parents, Dennis and Cynthia Hoffman,
and brother, Robert Hoffman, for their love and encouragement. Finally, I extend a
special thank you to my Arizona State University “extended family”, Arthur and Candace
Blakemore, Michael and Aileen Ormiston, Stuart and Debra Low, Paul and Marie
Burgess, John McDowell, Roger Faith, Edward Schlee, Jerry Kingston, Lee and Leslie
McPheters, Jose and Nancy Mendez, Bill Boyes, Susan Boyes, Don Schlagenhauf, Betsy
Smith, Gerry Keim and Amy Hillman. They have supported my educational and career

aspirations from the very beginning and are the reason I decided to become a professor.

TABLE OF CONTENTS

 

LIST OF TABLES ................................................................................ vii
LIST OF FIGURES ............................................................................... x
INTRODUCTION .................................................................................. 1
Summary of Dissertation Research ................................................. 1
Trends in Women’s Education, Employment and Childbearing ............... 7
US. Public Policies Related to Employment, Childbearing and
Childrearing 12
Child and Dependent Care Tax Credit and Flexible
Spending Accounts ........................................................... 13
The Pregnancy Discrimination Act ........................................ 14
The Family and Medical Leave Act ....................................... 16
International Policy Comparison .................................................... 19
Employer Flexible Working-Time Policies and Practices .................... 21
What are Flexible Working-Time Policies and Practices? .......... 22
Prevalence of Flexible Working-Time Policies and
Practices ........................................................................ 26
Empirical Evidence of Relationship with Family and
Health Outcomes .............................................................. 29
Figures and Tables ................................................................... 32

CHAPTER 1 EXPECTATIONS OF EXPECTING: WOMEN’S
ACTUAL AND EXPECTED FERTILITY OVER THE PAST HALF

CENTURY ........................................................................................ 37
Introduction .............................................................................. 38
Conceptual Framework ............................................................... 39
Previous Research ..................................................................... 41
Data ....................................................................................... 42
Results .................................................................................... 46
Discussion and Conclusion .......................................................... 53
Figures and Tables .................................................................... 56
Appendix A1 ............................................................................. 64

Description of Data Sources ................................................ 64
Source and Use of Expectations Data ................................... 64
Construction of Actual Cohort Fertility Data ............................ 67
Source and Use of Occupational Characteristics Data .............. 68

CHAPTER 2 JOB CONTINUITY AMONG EXPECTING AND NEW

MOTHERS ........................................................................................ 73
Introduction .............................................................................. 74
Conceptual Framework ............................................................... 75
Previous Studies ....................................................................... 78
Empirical Strategy ..................................................................... 82

vi

TABLE OF CONTENTS

Data ....................................................................................... 83
Results .................................................................................... 85
Discussion and Conclusion .......................................................... 93
Figures and Tables .................................................................... 95

CHAPTER 3 MAKING TIME FOR WELL-BABY CARE: THE
EFFECT OF MATERNAL EMPLOYMENT AND PAID AND

UNPAID LEAVES ............................................................................. 107
Introduction ............................................................................ 108
Background and Previous Studies ............................................... 110
Conceptual Framework ............................................................. 1 15
Empirical Strategy .................................................................... 1 18
Data ...................................................................................... 122
Results .................................................................................. 125
Discussion and Conclusion ........................................................ 137
Figures and Tables .................................................................. 140
Appendix A3 ........................................................................... 154

CONCLUSION ................................................................................. 160
Summary of Findings ................................................................ 160
Implications for Future Research ................................................. 163
Policy Implications ................................................................... 166
Current and Future Public and Workplace Policy Trends .................. 171

REFERENCES ................................................................................. 180

 

LIST OF TABLES

Table I.1 Estimated Demand for and Availability of Part-Time Work,

Paid Leave, Schedule Flexibility and Other Employer Policies ..................... 33
Table I2 Reasons Given for Working Longer Hours than Ideal ..................... 35
Table I3 Barriers to Working From Home ................................................ 36
Table 1.1 Sample Statistics for Chapter 1 ................................................ 61

Table 1.2 Estimated Average Differences in Labor Force Participation
And Occupational Characteristics by Expected and Actual Motherhood

and Timing of First Birth ....................................................................... 62
Table 1.3 Self Reported Influences Affecting Birth Expectations ................... 63
Table A1.1 Selected Occupational Categories and Mean Characteristics ....... 70

Table A1.2 Means, Standard Deviations, and Sample Sizes for
Occupational Characteristics Analysis ..................................................... 72

Table 2.1 Characteristics of Analysis Samples in a Selection of

Previous Return to Work and Related Studies ............................................ 96
Table 2.2 Comparisons to Previous Estimates of Job and Employment

Continuity .......................................................................................... 98
Table 2.3 Potential Omission and Misclassification at Different Intervals

For 21 Month Jobs .............................................................................. 99
Table 2.4 Descriptive Statistics by Interval for Jobs 21 Months Before

Birth ............................................................................................... 100
Table 2.5 Job Continuity Results for 21 Month Jobs at 21 Months before

to 18 Months After Birth ..................................................................... 102
Table 2.6 Changes in Estimated Relationships Before and After Birth .......... 104
Table 2.7 Characteristics of New and Old Jobs Among Job Changers ......... 106
Table 3.1 Estimated Compliance with the AAP Visit Schedule .................... 141
Table 3.2 Replication of Previous Compliance Estimates ........................... 142

viii

LIST OF TABLES (Continued)

Table 3.3 Sample Means and Standard Deviations for Selected Variables 143
Table 3.4 Base Model Results by Maternal Employment ........................... 144

Table 3.5 Sensitivity of Estimated Cross-Wage Elasticities to Measure
of Wages ........................................................................................ 145

Table 3.6 Estimated Effects of Maternal Employment and Paid Time Off ...... 146

Table 3.7 Results for Paid and Unpaid Leave and Other Job Attributes ........ 147
Table 3.8 Results for Paid and Unpaid Leave and Other Job Attributes

by Work Schedule and Wage Level ...................................................... 149
Table 3.9 Results for Paid and Unpaid Leave and Other Job Attributes

by Occupation and Work Schedule ........................................................ 151
Table 3.10 Probability of Weekend, Weekday (Except Friday) and Friday

Visits by Employment Status, Paid Leave and Occupation ........................ 153
Table A3.1 Equation for Predicted Wages for Non—Employed Mothers ......... 155
Table A3.2 Equation for Predicted Out-Of-Pocket Cost ............................. 156

Table A3.3 Means and Standard Deviations of Independent Variables
by Maternal Employment .................................................................... 157

Table C1 Comparison of Flexible Working-Time Policies at the Working
Mother Magazine “100 Best Companies” and the Average Company ........... 179

 

LIST OF FIGURES

Figure |.1 Classification of Flexible Working Time Policies and Practices ........ 32

Figure 1.1 Data Coverage Across Secular Trends in Family Size

and Birth Timing ................................................................................. 56
Figure 1.2 Expected and Actual Motherhood Across Cohorts ....................... 57
Figure 1.3 Expected and Actual Timing of First Births Across Cohorts ........... 58
Figure 1.4 Expected and Actual Family Size at Age 20 to 23 ........................ 59
Figure 1.5 Expected and Actual Family Size by Education .......................... 60

Figure 2.1 Monthly Hazard Rates for Job Changes and Labor Force
Exits for Jobs Held 1 Year before Pregnancy ............................................ 95

Figure 3.1 Correspondence between AAP Recommended Care Intervals
and Distribution of Actual Visits ............................................................ 140

Introduction

Summary of Dissertation Research

Across the labor supply, fertility, and child development Iiteratures, researchers
acknowledge the interrelationships between women’s employment, childbearing and
Childrearing decisions. Women who have more children also tend to spend less time in
paid employment. Those who have access to affordable childcare are less likely to work
outside the home than those who do not. Maternal employment may improve child
outcomes by increasing household income but it also reduces her time available to spend
with children. Numerous studies examine these tradeoffs. Yet less is known about the
role of governmental and workplace policies that may help women to combine work and
non-work roles.

The growing work-famin/work-life literature provides extensive case study and
empirical evidence of positive relationships between work outcomes, such as job
satisfaction, employee retention, and family outcomes and the availability of flexible
working-time policies and practices such as paid time off, part time work and flexible
schedules. Most of these studies are at the individual workplace level; less is known
about the role of these policies and practices in the labor market as a whole or with
respect behavioral rather than psychological outcomes.

The fact that the work-family literature presents workplace level evidence of the
importance of flexible working-time policies and practices and finds linkages between
flexible working-time policies and practices and psychological outcomes suggests these
policies and practices may influence employment decisions in the labor market as a

whole and objectively observable behaviors, as well. This dissertation investigates

 

relationships between availability of paid and unpaid time off, including FMLA leave,
and expected and actual childbearing, maternal employment and early investments in
child health through well-baby care.

In Chapter 1, I provide a broad, cohort level analysis of birth expectations. Birth
expectations may foreshadow future fertility and shape current behaviors, including early
career and educational choices. Chapter 1 evaluates these two possibilities. Substantial
changes in childbearing behavior, and fertility timing in particular, occurred across
cohorts born in the early 19405 through the 19705. I examine the trends in actual and
expected childbearing across these cohorts to determine whether or not secular changes in
childbearing behavior were anticipated. Although individual cohorts’ expectations of
completed family size approximate their actual completed family size, I find the trend in
expectations across cohorts does not anticipate trends in actual childbearing.
Furthermore, cohort expectations did not foreshadow the increases in childlessness and
single child families or the extent of fertility delay.

To examine the relationship between expectations of future childbearing and
current behavior, I hypothesize birth expectations influence early occupational choices.
To test this hypothesis, I compare the characteristics of chosen occupations among
childless women by expectation of future fertility to otherwise similar women who have
already had children. Occupational characteristics considered include mean weekly
wages and hours worked, percentage of part-time workers and percent of weeks per year
worked part time, average days worked per week, availability of flextime and the
concentration of women and mothers with young children in the occupation. In total, I

find the largest differences in occupational characteristics exist between women who

have children and those who have not; I find little evidence of systematic differences in
occupational characteristics by expectations of future childbearing.

In all, Chapter 1 concludes women, at the cohort level at least, are not very
forward looking with respect to their future childbearing behavior. Furthermore the
marked differences in occupational characteristics between mothers and non-mothers,
which are well known in the literature and apparent in my analysis, indicate many women
who become mothers do eventually sort into more “family friendly” occupations. Yet,
my findings among childless women imply the bulk of this sorting occurs close to and
after childbearing.

Chapter 2 extends the findings for occupational sorting in Chapter 1 by examining
relationships between the characteristics of mothers’ pre—pregnancy jobs and their
decisions to maintain or quit those jobs from one year prior to pregnancy through 1.5
years following the birth. While the premise of this study is similar to many existing
analyses in the return to work literature, my work differs in four fundamental ways. First,
most empirical studies in the return-to-work literature and related literatures rely on data
for women who gave birth in the 19705 and 1980s and many rely on the NLSY, which is
a powerful longitudinal data set but is not nationally representative. I provide updated
estimates of job continuity (exits from the initial job) using a nationally representative
sample of women who gave birth between 1993 and 2005. Second, while most studies
examine the decision to exit the labor force only, I separately examine the determinants
of labor force exits and job changes. Third, I include all quits which occur in the year
preceding pregnancy through 1.5 years after the birth. Although most labor force exits

are concentrated in the three months periods either side of the birth, most job changing

happens just before and during early pregnancy and would be missed in a shorter analysis
interval. I also provide a detailed discussion highlighting the importance of interval
studied and sample designation in general in the return-to-work literature and related
studies. Fourth, I focus on the relationships between job characteristics and employer
paid leave policies which may influence not only leave taking behavior associated with
the birth itself but also the ongoing arrangement of work and non-work time.

My findings suggest women who are eligible for FMLA leave are less likely to
quit their jobs for any reason prior to the birth and those who work part-time are less
likely to change jobs before the birth. I do find women without paid vacation are more
likely to change jobs prior to the birth but overall the evidence of pre-birth sorting into
jobs with more family friendly policies is not very compelling. After the birth, household
income and demographic variables explain more of the variance in quit behavior than
working time policies and practices and other job attributes.

Chapter 3 also examines the effect of paid and unpaid leave on mother’s behavior,
but in this chapter my focus shifts from employment decisions to child outcomes. The
American Academy of Pediatrics recommends children receive well-baby visits at regular
intervals over the first two years of life. The relevant AAP policy for survey respondents
recommended children receive at least eight visits between birth and age 18 months,
however, I estimate the average baby received only 2.25. While as with any form of
healthcare, cost and access are important determinants of demand, federal and state
policies have made well-baby care virtually free for all publicly insured infants.

Furthermore, cost sharing under private insurance is also very low. The fact that

compliance rates are so low despite the low cost of care suggests other factors, such as
time constraints, may be especially important.

The goal of Chapter 3 is to determine why compliance is so low. Findings
suggest out of pocket cost of care and household income have the expected relationships
with receipt of care and contrary to previous studies, I find the price elasticity of demand
is relatively high. My estimates imply a 1 percent increase in the out-of-pocket cost of
well-baby care (around $1 at the mean) is related to a 10.6 percent reduction in the
likelihood of receiving a given recommended visit. For paid and unpaid leave, I find
large differences in well-baby care use between low wage women who work long hours
who are eligible for FMLA leave and those who are not. For higher wage women and
those working a more conventional firll-time schedule (fewer than 40 hours per week)
FMLA eligibility does not seem to matter. Furthermore, the relationship between well-
baby care and paid leave varies by occupation and is even significantly negative in
certain occupations. Thus, my findings imply the type of job a mother holds matters;
paid and unpaid leave may enable mothers in certain types of jobs to take their babies to
the doctor but in others they appear to have no relationship or may even serve to ration
work absences and reduce the amount of care received.

In total, this dissertation contributes to the understanding of complex relationships
between childbearing and women’s employment by highlighting the potential importance
of flexible working time policies and practices and empirically investigating relationships
between paid and unpaid leave, job continuity among new and expecting mothers and
babies’ receipt of medical care. Although paid and unpaid leave can potentially help

women to combine paid work and care for a new baby, I find little evidence to suggest

women actively seek jobs with paid leave. Furthermore, the fact that women do not
appear to anticipate future fertility well at the cohort level implies the ability to do so at
the individual level is limited as well. This may explain why I find little evidence of
early sorting into more “family friendly” occupations. In Chapter 2 I identify cluster of
job changing activity around 1 year prior to the birth but I do not find compelling
evidence to suggest these job changes are motivated by search for paid leave. However, I
do find women who will be eligible for F MLA leave when they give birth are more likely
to keep their jobs prior to and during pregnancy. Together, the findings in Chapters 1 and
2 indicate women’s employment behavior around childbearing may have little to do with
the availability of paid leave. This could be because other aspects of the job, and wages
in particular, are more important. Or it could be because women who currently do not
have paid leave are unlikely to obtain it in any other jobs available to them and, vice
versa, among those who currently have paid leave most other jobs available to them will
also offer paid leave. More generally, their employment behavior may be more
reactionary than forward looking. These propositions are more thoroughly considered in
the conclusion.

Before delving into the specific analyses presented in Chapters 1, 2 and 3, I want
to make some key points regarding the trends in childbearing, the web of public and
employer policies and practices that may influence women’s employment and
childbearing behavior in the US. and existing empirical evidence of their effects on
employment behavior and health outcomes. Throughout this discussion, I stress the
contrast between the institutions which govern work and family life in the US. with the

institutional context in other developed countries. Compared to other developed nations,

in the US. the proportion of flexible working time policies which are legislated at the
federal level rather than determined at the workplace level is very low. In this sense, the
US. is a special case and some of the findings of my analyses might not apply beyond
the US. context. However, since other developed countries (EU. countries in particular)
have more governmental policies that intervene in the arrangement of working time and
issues of work-life balance arguably attract more attention in current public policy
debates, a sizeable proportion of the studies that are similar to my research and much of
the audience interested in the questions 1 pose are from substantially different policy
regimes. The international comparisons provided here are intended to position my
research in the broader international context of work-family research.
Trends in Women’s Education, Employment and Childbearing

Most of the analyses in Chapter 1 focus on the cohorts for which I have
expectations data, which include women born in the 19405 through the early 19705. But
the trends presented in that chapter are part of a longer series of changes in childbearing
throughout the twentieth century. The latter half of the twentieth century ushered in a
period of substantial economic and social change for women and families. While
somewhat exaggerated in popular accounts of the 19405 and 19505, the dominant family
form at that time was a nuclear family headed by a primary male earner who earned a
“family wage”. Mothers tended to stay at home, especially when children were young
(Rindfuss et al 1996). Childbearing and sex were socially unacceptable outside of
marriage, men and women were expected to marry and have children as soon as
financially feasible, women might work outside the home but not when children were

young, and parents were expected keep the marriage in tact at all costs for the sake of the

children (Rindfuss et al 1996). Since the 19505, the social and economic lives of men
and women in developed nations have changed drastically.

Between 1969 and 1994, women’s real wages increased by 31 percent while
men’s wages rose by only 3 percent (Blau 1998). This reduction in men’s relative
earnings led to the end of the “family wage” (Oppenheimer 1994). In the 19705, divorce
rates more than doubled, average age at first marriage increased and premarital
cohabitation became more socially acceptable (Cherlin 1992; Bumpass and Sweet 1989).
The increase in earnings power and labor market opportunity and the growing instability
of marriage as a social institution coincide with a 23 percentage point increase in female
labor force participation between 1970 and 1995 (Blau 1998).

This increase in women’s labor force participation began among older women
who had finished childbearing. By 1970, however, the labor force participation rate
among women with children under age 6 was 30.3 percent up from 18.6 percent in 1960
and by 1995 had reached 63.5 percent (Blau 1998). Whereas previously women would
take years out of the labor force to care for young children, as of 1987 more than half of
all mothers returned to work before their child’s second birthday (Klerman and Leibowitz
1990). By the 19905, 40 percent of mothers with 1 month old children were employed
(although 60 percent of these women were on leave) and 60 percent of women who
worked full time prior to the birth returned to their pre-birth employers (Klerman and
Leibowitz 1999).

Intrigued by these striking trends in women’s behavior, many researchers have
attempted to explain the causes. Some have argued that as men’s real wages were falling,

women entered the labor force to help maintain family income. However, the data do not

support this contention. The greatest increases in labor force participation were among
the wives of high earning, well educated men (Blau 1998). Therefore, it seems women
respond more to changes in their own wages than to any cross wage effect (Mincer 1980;
Juhn and Murphy 1997). Given that average educational attainment across women was
rising, the observed changes in wages and participation could be due to compositional
changes in human capital across cohorts. However, comparing cross cohort and within
cohort trends by educational attainment, this story does not hold either. There were real
wage and participation gains within educational groups and within cohorts (Blau 1998).
For example, between the graduating classes of the mid 19505 and those of the 19705, the
primary returns to a college education shifted from leverage in the marriage market to
leverage in the labor market (Goldin 1997).

Whatever the causal ordering of events, the trends outlined above have helped to
shrink the gender wage and experience gaps and break down gendered occupational
barriers. Yet despite this progress, few women have attained both a career and a family
(Goldin 1997). After the post war baby boom, which lead to unprecedented high rates of
marriage at young ages and early family formation, the long term trend towards smaller
families and later childbearing reemerged. Delayed fertility, among other factors,
allowed women to take advantage of labor market opportunities available to them in the
latter half of the twentieth century. From 1973 to 1988, the proportion of women who
had their first child between the ages of 20 and 24 declined by 7 percent (Rindfuss et al
1996). At the same time, the proportion who had their first child between ages 25 and 2'9
remained roughly constant but between ages 30 and 34 and ages 35 and 39, first birth

rates rose by 33 percent and 26 percent respectively (Rindfirss et al 1996). These changes

in the timing of births coincide with large increases in labor force participation among
younger women. However, fertility delay has also been linked to higher rates of
childlessness and longer intervals between births, which in turn lead to forgone fertility
(Morgan 1982). Indeed average completed family size declined from a high of 3 to 4
children during the baby boom, dipped below 2 in the mid 19705 and then stabilized at 2
by the 19905 (Van Horn 1988; Brewster and Rindfuss 2000).

Also, although women on average have come closer to economic equality, the
distribution of earnings, labor force participation rates and patterns of family formation
among the most and least educated workers have become increasingly disparate
(Rindfuss et al 1996). From 1979 to 1999, real earnings increased by nearly 22 percent
among college educated women and 3 percent among women with some college but
earnings stagnated among women with only a high school education and fell by 15
percent among female high school dropouts (Bowler 1999). Although women’s overall
labor force participation rate increased by 23 percentage points from 1970 to 1995, the
labor force participation rate among female high school dropouts increased by only 4
percentage points and the incidence of female headship among dropouts increased by
12.2 percentage points as compared to 6.5 percentage points among all women (Blau
1998). Yet, since there has been an upward shift in the educational distribution, the
average education of mothers has increased. The proportion of all babies born to college
educated mothers has increased from 8.6 percent in 1970 to 25.9 percent in 2002 whereas
the proportion born to mothers with less than a high school education fell from 30.8
percent to 21.5 percent (National Center for Health Statistics 2007). Nonetheless, more

educated women still bear fewer children on average than less educated women; the ratio

 

of average births per college educated women to average births per high school drop out
fell from 0.97 among women born in 1925 to 0.69 among women born in 1953 (National
Center for Health Statistics 2007).

Trends in fertility timing also differ by educational attainment. College educated
women under age thirty had the greatest decline in birthrates from 1975 to 1995 and were
more likely than less educated women to remain childless (Martin 2000). Yet some of
the decline in childbearing during their twenties was made up in their thirties (Martin
2000). However, women with less than a college education, who make up the majority of
those who postpone first births, do not appear to be catching up in their thirties (Martin
2000)

The patterns of delay and decline are not unique to the US. Japan, the EU.
countries and Australia all experienced their own baby boom after WWII, but generally
not to the extent that the US. did, and the booms were then followed by busts of varying
degrees (Brewster and Rindfuss 2000). Mediterranean countries have settled into some
of the lowest childbearing rates; average completed family size in the mid 19905 hovered
just above 1 (OECD 2007). Interestingly, women’s labor force participation is lower in
Mediterranean countries whereas in Nordic countries where participation rates are quite
high, women have sustained childbearing rates more comparable to the US. Although
the relationship between labor force participation and childbearing is negative at the
individual level, from 1970 to 1990 the correlation between country level female labor
force participation rates and fertility rates changed from negative to positive (Brewster
and Rindfuss 2000). But in the other high participation/high fertility regimes, part-time

work among women is much more common, presumably because part-time work often

.. A

has the same benefit status as full-time work in those countries (Hoem 1995; Stier et al
2001; Brewster and Rindfuss 2000). Societal norms also play an important role in
determining cross-country work and childbearing patterns. For example, in Germany
maternal care is strongly preferred to other child care and therefore labor force
participation rates among mothers of young children are very low and the market for
outside child care is quite thin (Brewster and Rindfuss 2000).

Many factors, including social norms, wealth, and the cost of raising children,
contribute to the differences in fertility between the US. and other developed nations.
The following sections examine public policies in the US. and other countries that likely
influence childbearing decisions, and more precisely, the coordination of paid
employment, childbearing and Childrearing.

U.S. Public Policies Related to Employment, Childbearing and Childrearing

Across developed countries there is a wide variety of policies targeted at families
with young children. These policies can be classified into three broad types: transfer
payments (i.e. tax credits and subsidies), paid and unpaid leaves, and other regulations of
the employment relationship and working time in particular. In the US. most policy
intervention is in the form of transfer payments. Arguably any policy that transfers
income from persons with children to persons with children potentially affects
employment, childbearing and Childrearing decisions. However, I restrict my attention to
those that are likely to have more direct effects on the coordination of work and non-
work time. The following sections review the set of US. policies that relate to the

arrangement of work and non-work time and contrast the evolution of the US policy

approach with that of other developed nations.l This discussion underscores the limited
governmental intervention in the coordination of employment, childbearing and
Childrearing and in and working time issues more broadly.

Child and Dependent Care Tax Credit and Flexible Spending Accounts

Enacted in 1976, the Child and Dependent Care Tax Credit (CDCTC) reduces tax
liability by up to 35 percent of child care expenditures (less for higher income families)
for individuals who incurred child care expenses while working or looking for work.
Flexible Spending Accounts (FSAs) allow employees to set aside a pre-tax portion of
their income to pay for child care expenses. FSAs, unlike the CDCTC, are only available
to workers whose employers offer them as part of their benefit plan. The CDCTC is the
largest policy that addresses child care costs in terms of benefits claimed (Maag 2007)
but FSAs are likely to be more important for higher income families. Still, in 2005 only
27 percent of employers with ten or more employees offered a dependent care FSA and
only 14 percent of eligible employees used the benefit (EBRI 2007).

Across states, the average annual cost of full time center based child care for a
four-year-old ranges from $3,380 to $10,787 and full time infant care is even more
expensive (National Association of Child Care Resource and Referral Agencies 2007).
FSAs are currently limited to no more than $5,000 per year. In 2005, the average
contribution among employees with dependent care FSAs was $2,630, which for an
employee making $35,000 a year would have yielded a net tax savings of $770 (EBRI

2007). Parents filing in 2007 are eligible for up to $2,100 through the CDCTC on their

 

‘ My focus in this section is on the evolution of the institutional framework in place in the US. today. For
a detailed discussion of the particulars of current policy initiatives in the US. and other countries see the
conclusion to this dissertation.

federal return but the credit is not refundable. 28 states provide child care tax credits for
state tax liability, and in 13 of the 28 states at least some of the credit is refundable.

Labor supply elasticity estimates with respect to child care costs vary from near
zero (Michalopoulos et a1. 1992; Hotz and Kilbum 1992) to -0.78 (Averett et al 1997).
Policy simulations conducted by Averett et al (1997) suggest eliminating the CDCTC
would reduce hours of work by 15 percent. Despite empirical evidence of the positive
effects of the CDCTC on labor supply, it is slated to be phased out because of its
interaction with the Alternative Minimum Tax (AMT) (Maag 2007). Currently there is a
patch in place that allows families to claim CDCTC against both regular and AMT
liability but if that patch is not extended, there will be a discrete drop in the assistance
provided under the CDCTC in fiscal year 2008 (Maag 2007).

To claim benefits under the CDCTC, state versions of the CDCTC and FSAs,
parents must incur child care expenses for center-based care or for in-home providers
who are willing to report their income. Low income families, such as those on or
transitioning from Temporary Assistance to Needy Families (TANF) may qualify for
direct subsidies of child care costs to enable parents to work or attend school, but again in
some states the provider must be licensed. Most child care centers are licensed but they
tend to be more costly than home-based care, and certainly more expensive than relative
care. Furthermore, many center—based care facilities are ill equipped or charge additional
fees to serve children under two.

The Pregnancy Discrimination Act of 1978
Prior to 1963, dismissal of pregnant women was legal and not uncommon. The

Equal Pay Act of 1963 and the Civil Rights Act of 1964 made discrimination on the basis

of sex illegal. However, equal treatment policy has a disparate impact on female workers
because the physical burdens of childbirth are not shared equally by men and women
(Guthrie and Roth 1999). Under equal treatment, the courts upheld differential treatment
for male and female workers in general and differential health insurance coverage for
pregnancy related conditions in particular. Two notable class action suits (Geduldig v.
Aiello 417 U.S. 484 [1974] and General Electric v. Gilbert 429 U.S. 125 [1976]) were
brought against employers who failed cover pregnancy related conditions under disability
policies but the Supreme Court found in favor of the employer because the exclusion of
pregnancy conditions from disability coverage is differential treatment by pregnancy
status, not sex per se (Williams 1984). The Pregnancy Discrimination Act (PDA) was a
direct response to the debate inspired by these controversial rulings.

The Pregnancy Discrimination Act (PDA) of 1978 amended the Civil Rights Act
of 1964 to require pregnancy and childbirth related disability be treated in the same
manner as any other disability under policies upheld by employers with 15 or more
employees (Guthrie and Roth 1999). At the time the PDA was enacted, an estimated 60
percent of female workers were covered by disability policies that granted an average of
six weeks of leave and approximately 50 percent had employer provided insurance
(through their own or their husband’s employer) which treated pregnancy and childbirth
related health expenses differently than other similar health expenses (Gruber 1994;
Trzcinski and Alpert 1994).

The FDA did not qualitatively alter the equal treatment approach to pregnancy. In
fact, by equating pregnancy with other disabilities, it reinforced this perspective. Under

the PDA, after the immediate postpartum period and physical recovery, new mothers are

to be treated as any other employee. Provision of leave or other benefits beyond that due
to other disabilities was at the discretion of employers. In fact, a Montana law which set
aside separate requirements for “reasonable leave of absence” for pregnancy was found to
be unconstitutional.2 Furthermore, since new fathers do not experience any physical
disability at the birth of a child, they receive no mandated access to leaves at the onset of
new care—giving roles under the PDA.

The treatment of pregnancy as a physical condition without regard to the
emotional transition and assumption of new care-giving responsibilities has lead to a
system in which expectant mothers (and to some extent fathers) bank leave time available
to all employees to obtain a leave for childbearing. As of 1998, one third of full-time
private sector female employees of medium and large establishments were offered unpaid
maternity leave and only 2 percent were offered paid maternity leave (Ruhm 2004).
Instead, women use paid vacation, sick and personal leave days, to obtain paid leave at
the time of childbirth. 74.9 percent of women with children under six-years-old have
paid vacation through their employer with an average of 10.3 days per year and 67.8
percent have paid time off for personal illness.3 Until the Family and Medical Leave Act
of 1993, women without paid leave whose short-term disability time had expired
presumably faced a stark choice: return to work or lose the job.

The Family and Medical Leave Act
The Family and Medical Leave Act (FMLA) of 1993 was specifically targeted at

the needs of working families at the time of birth or adoption or for their own or a family

 

2 Miller-W011! Company, Inc. v. Commissioner of Labor & Industry 515 F. Supp. 1264 (D. Mont. I981),
vacated, 685 F.2d 1088 (9‘h Cir. I982)
3 Author’s tabulations of the National Study of the Changing Workforce. See Table I. 1.

l6

 

members’ illness. Unlike the PDA, the F MLA distinguishes between those with family
responsibilities and those without and thus extends beyond addressing the physical
burden associated with childbirth itself to the time demands new parents face.
Furthermore, the FMLA mandates a benefit of up to twelve weeks of leave regardless of
other employer policies whereas the PDA simply required existing employer disability
policies be extended to pregnancy related conditions.

FMLA leave may be used at the birth of a newborn, to care for a new baby or
newly placed adopted or foster child, to care for an immediate family member with a
serious health condition or to take medical leave for one’s own serious health condition.
The definition of “serious” health condition is ambiguous and likely depends on the
employers’ interpretation of the law and employee’s ability to provide medical
documentation. In the strictest sense, FMLA leave does not cover well-baby care visits.
However, 6 of the 8 recommended visits occur within the first year of life and those that
fall in the same calendar year as the birth should be covered under FMLA. Importantly,
F MLA need not be used for one lump sum leave. Employees might negotiate to use
FMLA leave to achieve shorter work days following the birth of their new baby or to take
part of a day off to take a child to the doctor. However, there is no remuneration required
during FMLA leave and it applies only to full time employees with at least one year of
tenure who work for employers with at least 50 employees within a 75 mile radius of the
employee’s workplace.

Given the coverage limitations of FMLA, the public debate inspired during the
1992 presidential campaign and after the enactment of the FMLA was probably more

effectual than the provisions of the law itself (Ruhm 1997). The lack of remuneration

requirement means, for many workers, using F MLA leave to extend parental leave
beyond any paid maternity or sick leave they already possess is cost prohibitive.
Employees who have accumulated paid time off (sick time, vacation days) may apply it
towards the twelve weeks of FMLA leave to receive some compensation and, in fact,
employers may require employees to use all accumulated paid time off (sick days and
vacation days) to access FMLA leave (Guthrie and Roth 1999; Ruhm 1997). Yet, the

F MLA specifically states that “nothing in this title shall require an employer to provide
paid sick leave or paid medical leave in any situation in which such employer would not
normally provide any such paid leave” (Guthrie and Roth 1999; Family and Medical
Leave Act of 1993; Sec. 102[d][2][B])

Immediately following the enactment of the FMLA, the Commission on Family
and Medical Leave ordered two surveys to measure the effects of the law, one employee
survey and one establishment survey, and the results were presented to the Commission
in a 1996 report. The report found the F MLA had indeed increased the provision of
family and medical leave among covered establishments and there was little evidence of
any adverse business effects (Commission on Family and Medical Leave 1996).
However, they also found only 46.5 percent of private sector employees worked for
covered employers and were eligible for leave and among those covered, 41.9 percent
were not familiar with the law. Further, many employees stated they needed more leave
but did not take the full twelve weeks because they could not afford it. Other studies
have found similar positive results for expansion in coverage of leave policies but
confirm that changes in leave use were limited, which suggests there are important

financial and other barriers to use by employees (Walfogel 2001). Follow up

 

establishment and employee surveys conducted in 2000 found non-covered
establishments were more likely to offer benefits similar to covered establishments than
they had been in 1996, however the gap between covered and non-covered was still
substantial (Cantor et al 2001). Thus, although FMLA provisions appear to have
increased the incidence of leave benefits and set a minimum standard for leave, there is
still substantial diversity in benefits across employers and employees, both due to
variation in F MLA coverage and to employers’ voluntary provision of more generous
policies.
International Policy Comparison

The fact that the U.S. intervenes in ongoing work and family arrangements
primarily through the tax code and on the side of child care arrangements rather than
hours of work is in stark contrast to the approach of other OECD countries. In general,
the U.S. policy regime is reflective of a decidedly different view of the normative role of
government in work and childbearing and care decisions. European public policies
strongly reflect the assumption that children are public goods. Furthermore, they hold
that the time parents spend caring for children should not only be heavily publicly
subsidized but should also be facilitated by enabling institutions which directly affect the
organization of work and non-work time, and in some cases augmented by public child
care programs (Gomick et al 1997). Many European countries mandate paid leave
benefits for both new mothers and fathers and leaves may be paid at rates as high as 85
percent of wages and extend to as much as a year with unpaid leaves available beyond

that (Brandth and Kvande 2002, 2001; Perrons 1999; Ruhm 1998). Indeed throughout the

A

EU, eligible and actual maternity leave lengths are generally measured in months and
even years rather than in weeks as in the U.S.

However other developed countries have very low birth rates whereas the U.S.
has maintained above replacement level fertility rates. The 2007 estimated total fertility
rate in the EU. is 1.5 births per woman, ranging from 1.21 in Lithuania to 1.99 in the
Netherlands, whereas the U.S. estimate is 2.09 (The World Fact Book 2008). Even in
Nordic countries, which have maintained higher birth rates than the rest of Europe and
are notorious for their generous maternity leaves and working time policies, the current
total fertility rates range from 1.66 in Sweden to 1.78 in Norway (The World Fact Book
2008). Yet, the history of policies in the EU. suggests the differences in policy regimes
cannot be entirely explained by differences in fertility rates. Although Nordic countries
have used working time policy to achieve equality since the early 20th century, most
industrialized nations were not active in intervening until labor shortages following
WWII necessitated female entry into the labor force.

While the laissez faire approach to employment relations has been credited with
maintaining the competitiveness of the U.S. economy and promoting innovation and
growth, it has also contributed to growing economic inequality. In the case of workplace
flexibility, a worker’s ability to negotiate flexible working-time arrangements with her
employer will depend on her strategic position in the production process and overall
economy. Core jobs will come with good working time benefits, periphery jobs will
offer little flexibility to meet the needs of the employee and may require a lot of flexible
accommodation of the employer’s needs (e. g. through acceptance of layoffs, overtime

and seasonal and temporary positions) (Kalleberg 2003; Grimshaw et al 2001; Kalleberg

20

 

2000). Although the FMLA grants a subset of the workforce the right to a set amount of
time off, the very same workers who are disadvantaged in bargaining for that time absent
the law are likely to be those who cannot afford to take an unpaid leave. Furthermore,
workers on the fringes of the labor market who are unable to find continuous full—time
employment are not covered by the FMLA.

One final crucial difference between determination of working time in the U.S.
and other developed nations is the tying of health insurance benefits to full-time
employment. In the U.S. health insurance coverage for an employee and his or her
dependents is often only offered to core (full time, permanent) employees. As the cost of
healthcare soars the purchase of individual coverage is cost prohibitive and Americans
who might otherwise prefer to work a reduced schedule and could afford to do so may
find themselves holding full time positions to obtain health insurance benefits. Indeed,
Buchmueller and Valletta (1999) find women who have health insurance coverage
through their husbands’ employers work 15 percent fewer hours than those who do not.
The tying of health insurance to full time employment in essence places a large penalty
on reduced work schedules. This peculiarity of the U.S. system provides one explanation
for the fact that many European women seek accommodation of care giving
responsibilities through part time work but relatively few American women do so.

Employer Flexible Working-Time Policies and Practices

Although the U.S. does reduce parents’ opportunity cost of work through child
care tax credits and does engage in some, albeit limited, intervention in leaving taking at
the time of a birth, the ongoing arrangement of work and non work time in the U.S. is

largely left to the discretion of employers, employees and households. As is generally the

21

 

case, workers’ “discretion” is greatly limited by their labor market options and power to
negotiate individually or collectively with their employers (Berg et al 2003). Since 1988,
the incidence of unpaid maternity leave has increased sharply, especially after the
enactment of the FLMA, but at the same time the incidence of other types of paid leave
appears to be declining (Ruhm 2004). This trend assessed alongside relatively high labor
force participation rates for both men and women suggests the time crunch for working
families may be worsening.

What are Flexible Working-Time Policies and Practices?

In Chapters 2 and 3, I focus on employer provided paid leave and eligibility for
FMLA leave because these policies are observable in my data. The web of flexible
working-time policies and practices that may influence job continuity or the ability to
take a child to the doctor is much broader. Figure 1.1 categorizes various flexible
working-time policies and practices by the way in which they affect the organization of
working time and activities and the magnitude of their influence. Flexible working-time
policies and practices affect decisions regarding work and non-work time and activities
through the temporal and spatial boundaries of work. By temporal boundary I mean the
division between work and non-work time and by spatial boundary 1 mean the division
between work and non-work space. For example, a flexi-time or flexible schedule policy
generally allows employees to select their own start and end times within a set range;
under this policy, the duration of work and nature of the job generally do not change but
the boundary of work and non-work time is more flexible. Both on—site child care and

telework influence the boundaries between work and non-work space. The former takes

22

 

a portion of work space and dedicates it to non—work activities whereas the latter does
just the opposite.

Policies are further classified by the degree to which they alter the temporal or
spatial boundary. F lexitime, overtime and compensatory time or “comp-time”, which
allows workers to bank overtime hours as future time off rather than accept overtime
wages, are classified as the most flexible working time policies because they allow for
day to day and week to week variation in the work schedule. Maternity leave on the
other hand is less flexible because it generally provides a fixed allotment of time off at
the time a child is born. Similarly compressed work week and other paid leave alter the
boundary of work and non-work time but generally operate under a fixed arrangement.
Job sharing and gradual return also tend to be more flexible because arrangements are
often tailored to the needs of the individual rather than dictated by a universal policy.

Among policies which operate at the spatial boundary, telework generally
provides a more flexible boundary than on-site childcare because the employee is usually
telecommuting from his or her home or a space in which they are not visually supervised.
Thus he or she may switch between care-giving or other non-work activities and work at
his or her own discretion. The opportunity to see one’s child during the day even at an
on-site facility may be restricted to breaks and lunch time.

The fundamental nature of the job a person holds and the informal support he or
she may receive from supervisors and coworkers may also alter the arrangement of work
and non-work time and space. For example, a professor has substantial control over her
hours of work and, depending on the culture of the department, can often choose fieely to

work a substantial portion of hours at home because of the nature of the job. These job

23

 

characteristics afford her control over working time that may make orchestrating work
and family roles easier. In other professions, an employee may have an especially
understanding supervisor who informally allows him to come in late or leave early when
child care arrangements fall through.

It is important to note that more flexible policies are not necessarily “better”
policies. Many flexible arrangements require continual negotiation and, depending on
the balance struck, can tend to favor employer needs for flexibility more than employees.
The use of paid vacation time, sick time or even banked compensatory time is generally
governed by formal and informal rules and doled out by the day or negotiated to obtain a
leave of fixed length. Although approval may be required before a leave begins, once in
place a paid leave or reduced hours arrangement is less likely to be altered to
accommodate the employer. In a study of paternity quota and time account usage among
Scandinavian men, Brandth and Kvande (2001) show men were far more likely to use the
paternity quota and other “compulsory benefits”, presumably because the quotas are
subject to less negotiation.

As stated previously, of this broad set of policies, I am only able to include paid
vacation and sick leave and eligibility for FMLA leave in my empirical analyses in
Chapters 2 and 3. This is a common drawback among existing nationally representative
surveys. Furthermore, large scale surveys are rarely able to capture informal flexibility.
Even when surveys do include information about formal and informal employer working-
time policies and practices, the wording of the questions often inhibits desired inference.
For example, many surveys which ask about employer policies and practices structure

questions as follows: “Do you have flexible work hours that allow you to vary or make

24 A

changes in the time you begin and end work?” (May Current Population Survey 2004).
Questions like this confound availability and use. Respondents could answer no either
because they do not have the option to work a flexible schedule or because they were
offered a flexible schedule and refused it. Assuming the intent of studying workplace
policies and practices is to infer the likely effect on behavior if the policy or practice were
extended beyond the current covered population, questions of this sort will lead to
overestimates of the true effect. With respect to paid time off, the question “Do you have
paid vacation leave?” is also problematic. An employee who receives an allotment of
days each year but has already used them all may respond no. Furthermore, survey data
cannot measure the availability of informal flexibility among employees who never
attempt to use it. Sometimes they may know of others in their department or workgroup
who have made informal arrangements for greater flexibility but this knowledge is tied to
their colleagues attempt to use flexibility and it does not necessarily mean the individual
surveyed had the opportunity to obtain similar flexibility and declined it.

There are ways to structure questions about formal flexibility that measure
availability separately from use. For example, Berg and Kossek’s (2006) battery of
workplace flexibility questions includes separate questions for availability and use and
possible responses for availability are “yes”, “no” and “don’t know”. This structure
allows the researcher to observe offer and use separately, to identify persons who enjoy a
given form of flexibility informally (without a policy in place) and persons who probably
have not tried to obtain the benefit or accommodation in question and therefore don’t
know if it would be available to them. Also, their survey includes a probe for reasons

why an employee does not use policies available to him or her. This sort of information

25

helps the researcher to infer the potential cost (either real monetary costs as associated
with unpaid leave or sacrifice of future opportunities or intangible penalties imposed by
superiors and peers) to the employee of using a policy or practice.

The National Study of the Changing Workforce (N SCW) is a publicly available
survey that also has many questions design to measure availability apart from use. Yet
relative to similar publicly available data sets the NCWS is small and it is cross-sectional.
Thus the costs of using it for many detailed analyses of labor market behavior, including
those in this dissertation, outweigh the benefits.

Prevalence of Flexible Working-Time Policies and Practices

The distribution of flexible working-time policies and practices throughout the
labor force is very unequal. Only 27 percent of the workforce reported they had schedule
flexibility in the 1997 Current Population Survey and availability varied from 12 percent
of the workforce to over 50 percent of the workforce across occupations with lower
incidence among female employees (Golden 2001 ). Table 1 uses NCWS data to
tabulates the availability and excess demand for part time work, paid leave and other
workplace policies and practices among workers in general, men and women, mothers
and fathers and parents with children under five-years-old.

The first row of Table 1 shows part-time work is relatively scarce but
disproportionately concentrated among mothers with young children. Only about 20
percent of all employees work part-time, but 30 percent of mothers with young children
and only 6.89 percent of fathers with young children do. Among those working full-time,
about 15 percent would prefer to work part-time and among full-time mothers with young

children over 20 percent would prefer to work part-time. However, the ratio of mother’s

26

ideal to actual hours of work is generally closer to 1 than for fathers and the average ratio
among mothers with young children is actually above 1. Thus it seems mothers have a
higher demand for part time schedules and are more likely to obtain schedules that satisfy
their preferences than fathers. The majority of fathers (64.70 percent) would prefer
shorter hours.

Over 41 percent of employees, and 64.29 percent of mothers with young children,
claim they could arrange part time work. Table 2 provides tabulations of a multi-punch
question about barriers to part-time work, which are suggestive of the implicit and
explicit costs of part-time schedules. The vast majority of firll-time who would prefer
part time work (82.32 percent overall and 88.16 percent of mothers with young children)
cited reductions in income among the reasons why they do not work part-time. Income
reduction was a more common reason among mothers than fathers, perhaps because
families where mothers are working full-time are more likely to have lower other sources
of income or be single parent families. Fathers, on the other hand, were more likely to
cite effects on personal and organizational success and achievement as barriers to part-
tirne work than mothers. This was also true of male relative to female employees overall.
In general, Table 2 suggests current household income constraints are the most common
barrier to part-time work among all employees and adverse affects on long term career
opportunities and earnings grth are relatively more common concerns among men than
women.

Returning to Table 1, men and women also differ in the availability and demand
for working from home and the availability of paid leaves. Overall, only 9.90 percent of

all employees claim they do some work from home. Importantly, this figure represents

27

 

employees who have an arrangement to complete a portion of their paid hours of work
while at home; this figure does not include the 23 percent of employees who regularly
take work home with them but are not paid for the extra time. Men and fathers are
slightly more likely than women and mothers to work from home but more women who
do not work from home would like to than men. Among young parents, the proportion of
mothers who would like to work at home is nearly 20 percentage points higher than the
proportion of fathers who would like to work at home.

Mothers are also less likely than fathers to have access to paid vacation leave and
those who do have vacation leave receive fewer days on average than fathers. Yet
fathers, and those with young children in particular, are more likely to be unable to use all
of their vacation time and more mothers than fathers have access personal days.
Somewhat surprisingly, mothers are substantially less likely to have paid holidays than
fathers. This could reflect their relatively higher concentration in service occupations.

Most differences in the availability of flextime, reduced or compressed schedules
and childcare assistance across genders and by parental status are less pronounced.
However, father’s flextime policies appear to be more flexible on average than mothers,
meaning they are more able to vary start and end times of work daily rather than have
their chosen schedule remain fixed. Also, mothers are far more likely than childless men
and women and fathers to be able to arrange a part year work schedule. This could be
because they count maternity leave arrangements as achieving part year work. Cash or
in-kind assistance with childcare appears to be relatively equally distributed across
parents and non-parents. This finding indicates the survey did elicit the availability of

policies rather than use.

28

 

In summary, the patterns in Table 1 reflects the uneven distribution of workplace
policies and practices across the workforce and highlights differences between men and
women in general, and mothers an fathers in particular. The only policies which come
close to universal coverage are paid vacation and paid holidays. Yet there are still
marked differences in availability between mothers and fathers. Thus, for U.S. workers,
the options available for managing work and non-work time and combining childbearing
and rearing activities with paid employment largely depend on the job. Chapter 3
provides further evidence of the importance of job context. I find the relationship
between paid leave and compliance with recommended well-baby care differs
substantially across mothers’ occupations. But again, the common theme of high
variability across workplaces and jobs holds.

Empirical Evidence of Relationships with Family and Health Outcomes

The empirical literature on working time policies and practices is vast and crosses
disciplines including psychology, sociology, management and economics. In the
economics literature, maternity leave has received the most research attention. Empirical
studies have found leave provision increases the instance of leave taking and duration of
leaves (Berger and Waldfogel 2004; Waldfogel 1999) and may have some positive
employment effects. However, employment effects are at least partially offset by wage
reductions (Ruhm 1998; Gruber 1994).

For work schedules, in a convenience sample of 324 women who were employed
during their first trimester, Glass and Riley (1997) find mandatory overtime decreases the
likelihood of return to the pre-birth employer. They also find length of leave, flexibility,

hours worked at home and perceived supervisor and coworker support all increase the

29

 

likelihood of return. Part-time work is related to lower levels of work-to-family conflict
and mothers who work part-time report more success in balancing the demands of their
work and family life (Hill et al 2004). However, part-time work may come at a cost.
Comfort et al (2003) find only 17 percent of employees working part-time received a
promotion during their tenure with their current employer. Also, they find only 2 percent
of male and 5 percent of female part-time workers are managers. Errnisch and Wright
(1993) find evidence of lower wages for part-time work even after accounting for self-
section into full-time and part-time jobs.

For FMLA leave, as stated previous, empirical findings show evidence of
expanded leave coverage but little change in leave taking (e. g. Waldfogel 2001). This
may be because many employees are not familiar with their rights under the law
(Waldfogel 2001). Two years after FMLA was passed, only 63 percent and salaried and
50 percent of hourly employees said they had heard of the law (Budd and Brey 2003).

A small number of existing studies examine relationships between flexible
working—time policies and practices and child health outcomes. In a study of breast
feeding behavior among working mothers, Jacknowitz (2004) finds provision of on site
child care and ability to work eight hours at home increases the likelihood of
breastfeeding at six months by 59 percent and 21 percent respectively. 78.7 percent of
employees who took F MLA leave said it had a positive effect on their ability to care for
family members and, among those who reported positive health effect, 93.5 percent
indicated FMLA leave made it easier for them to comply with doctors’ instructions

(Waldfogel 2001). These finding invites similar analysis on other child outcomes.

30

In total, flexible working-time policies and practices appear to improve family
outcomes and enable mothers to manage work and care-giving roles. Across the work-
family literature, most evidence for the effects of working-time policies and practices is
at the workplace level. While national surveys cannot match the rich detail of many of
these primary data studies, the growing evidence of the importance of working-time
policies and practices invites large scale analyses to develop nationally representative and
policy relevant estimates of relationships. Although the richness of the workplace
context is missed in large scale surveys, the need for nationally representative estimates
based on more robust statistical inference to verify findings in smaller scale studies
justifies their use. This dissertation is aimed at meeting that need. The following three
chapters contain the specific analyses I have introduced and provided background for
here. I conclude with a discussion of policy implications of my findings in the context of
current policy initiatives and highlight my contributions to existing literature and

directions for future research.

31

Figures and Tables

Figure |.1 Classification of Flexible Working Time Policies and Practices

 

 

 

On-Site Childcare

 

Work Boundary
Fixed Flexible
Temporal l , ’ I
F l l 1
Reduced Work Hours Sick Days Overtime
Maternity Leave Vacation Days Flextime
Compressed Work Week Job Sharing Compensatory Time
Gradual Return
Flexible
Fixed
Spatial l l l l
| l T |
Telework

 

Notes: The policies and practices I am able to study in Chapters 1, 2 and 3 are in bold italics. I am only

able to consider fiextime in Chapter 1.

32

 

HmEm ... 9:330... 0032... .2 man ><m=mc=z< o. .um: .230 <<o_._x_ .uma rom<o. mozmac.m use 0.30.. m30_o<ma po_.o.mm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

...08. 2.0.... 7.0320 10:58 3090; 3902 K0902.
0...... Aka 0...... Amfim
.x. 0.. 550.888 .58 (<0; .00: .230 3.3.x. 5.8.x. ”3.3.x. «2.x. 3.8.x. 9.3.x. ”5.8.x.
8.3. 8.3. A... .. A. be. A. .wN. AN. .A. ANA...
e\o 0A. .288 €0~E=m T... .230 $30 $9.... 3.1.9. 8 3.3.x. . .. .uo\o 3.3.x. . ..m~..\.. 3.3.x. .o..Ro\.. N . .u ...\o
.29.: .02.. .25... A93. A93. A905. 2.8 A. . . m. AN. .. 0.3.
”2.0 0A. Emu. $0...» 20.5 8 >025. £0} :05 oh» Poo ob; oh. 93 9.3 .bA
A98. A99... 803 A98. A98. 8.0... 8.2..
e\.. in. Eon. «<0... :95 macs. >25. 93.8 .88 mchwfie 8.3.x. mmhufi: mnhofi. 3.8.x. mmbéx. mmbwox.
3m: .. 8.03 A. .3. A78. A. be. A. .3. Auras. Awmm.
..\.. 0.. 2.0% €0~E=m 3... 23a :50 00:... >203ma 8 2.3.x. 3.8.x. 3.. ...\o magmas. 3.3.x. 3.3.x. 3.8.x.
«<01» .02.. .230 Area. C .m .. ATE. 2...». A. .3. A93. APoo.
as. .388 550.853 03.0.. .0»: .230 $04.83 moan 3.8.x. 3.8.x. wmbfix. 3.3.x. 3.3.x. 3.3.x. warm .o\o
wazoma em 2... .230 A. .No. ANS. C .3. Awum. A. .2. A93. Awba.
..\.. m<2 €04.» m8... 10:8 0.8.x. . ..oco\o «Owes. 5.3.x. cows; 3.3.x. 9.3.x.
8.3. 8.00. 8.3. A. bu. A93. AN. .A. A. .3.
.x. Do 20. $0} #03 10:8 9: $050 5.8 ...0 3.8.x. A. .m~..\o 3.3.x. 2.3.x. 3.3.x. 2.3.x. m... . q..\..
A. bu. A. .3. 2.3. A. .3. Arab. 3.3. Awhw.
as. <52. m3. pea <83? 23m 3.3.x. 8.3.. 3.8.x. 3.3.x. Eases. 3.3.x. 3.8.x.
8.8. A. .3. A. .No. A. .m .. 2.3. ANS. Am. .3
>8an 2:309. 0m <wom~.0= 08a .90... True .wbm 3.3 .mb. SUN .9...
Ac. .0. Ash... 8b.... 8...... 8.3. A99... 88 ..
..\.. 9.283 (S... v»... 5.8.8: <55 m3. 20. >20 8 . 2.0.x. $.me 3.3.x. 3.8.x. 3on0 3.3.x. Mass.
:8 >.. $83.0: 8...”... 2.5. 8...... Arum. Aron. ANS. A. .3.
.x. <5... B... .0»... 10....me 3.3.x. 3.3.x. 3.8.x. mm. 3.x. 3.9.x. 3.2.x. 3.8.x.
8.0... A. .3. A. .N .. A. .8. A. .3. Qt... .. Aura.
..\o 2.03.2. .0»... .230 O»... .804 .0283... 5.5% 3.51%.. 3.2.x. 3.8.x. 3.3.x. 0 . 80.x. Samoa 3.3.x.
A. .om. A. .3. A. .3. A. .8. A. .3. Awbm. Ammo.
..\o 03.8% <53 3.0 .23... 012.9 330:». 5:03 3.8.x. . ..A...\.. 3.3.x. 5.3.x. 3.3.x. 5.3.x. NNUQX.
wow 20. 2.052. 5.05: 8.3. A..~o. A. . .0. A. .3. A. .3. AWN... Aw... ..
..\o 2.022. .030 .230 0.8m 8 00.... m3 max 035 300$ Em is 3.8.x. 3. .N..\.. 3.3.x. the...
Am... 038 .883 092. .138. A. 8.. ANA... APE. 0.00. Ohm.
.x. 0.. .288 $3. .00... 2:8 03. 8 03¢ 84 m3.» . .tSAX. is in 3.3.x. 5.5.x. 50...... 3.3.x.
0.1.0 we 20. 2.0.22. m8: .. Ammo. ANUN. A23. A93. Awbo.

 

 

33

400.0 I 903000 00303. .2 0:0 ><0._00._..< o. 00... 4.30 <<o:r .um.0 .0020. 93000.0 0:0 050. m30_0<0q 02.0.00

 

 

 

 

 

 

 

 

 

 

 

60:03.00.
$ 00: 0:080 m8: 0:0 m:0 .2303 A20... 0 . ..0$ 305$ A00~$ 0 . 00$ 500$ 093$ 0A.00$
A900. A. .AN. 2.8. A. .00. A. .5. A00 .. Awbn.
$ $20. 0.0.3.30 00: <5 ms: 0:0 00:0 .2303 00:0. 00.0.$ 0A.00$ 00.00$ 00.00$ 00.0.$ 083$ 085$
A. N0. A700. A500. 00.. .. A~.00. A93. A000.
$ :30 0.0.. 93000.30 008:0 ><0..00.0 095$ 0. 05$ 598$ 0 . 00$ 300$ 00.0..$ 593$
C .0. A. .08 A. .00. A. .00. A. .05. A000. A00...
$ mg. .033 5.8.0. .00 .0538 .m .....00. C30 0.0.. 00. E$ 300$ w . 00$ 00.00$ 093$ k005$ 093$
mo..00=..:m 00:0:3 ><0..00.0 A. .00. A~.0 .. ANS. $.50. A000. A00... A005.
$ 00:.0 33:00 8 (<qu 00... <03 300$ N0.50$ 300$ 850$ 0......0$ 300$ N0..0$
A900. A. .05. 2.00. A. .00. 2.5. A900. A0.~0
$ 2.05.00 8 $01. 00303300 $012.00.. 300$ 00.....$ A . 00$ 0~.N0$ 095$ 00.00$ 00.00$
:03 A. .00. 2.00. A. .55. A. .00. Aw.0.. A000.
$ m30.000q 03.03 w0..<.00 8 10.0 2:0 0....0 00.0 .0..0$ .0.00$ 500$ 500$ 300$ 500$ ~900$
A958 A..N0. A9... A. .0.. 2.00. AN.00. A000.
$ m30.00.0~ 0000083000303 0:3..0 0....0 00.0 .90q$ .9..0$ . ...0$ 000$ . ..0A$ .90~$ .90 .$
A908 8.8. A950. A. .00. A. .0N. A000. AN.00.
$ 0030.00.01 08300 0.30. 030:0.0. >mw.m.0:00 .noq .0.00$ .0.0N$ . ..0A$ 005$ . ..0.$ 500$ 5.0.$
0....0 00.0 A000. A000. A955. 2.3. 200. A900. A. .50.
$ 00.2 .u..0-...0x >083. 00. 03.0003 9.0038 300$ 093$ 00b0$ 300$ ”300$ 300$ 095$
A900. A53. AS... 2.5. A..00. A000. A000.

 

 

 

 

 

 

 

m0500. 2000:... 9:00. 00 30 0:0:m...m €013,030 NO0N.
0000300 .: 0033:0303.

 

 

M0300 £05.33 000 000:00

8 08300 03:30.03 0.. 00053.0: 00330.03. $030000 0.403 03

34

436 _.N mmmmoam 02m: 3.. <<o_.x.:m rose 9 1053.5: .amm.

 

 

 

 

 

 

 

 

 

 

.38. Kw.» 313.0 33.8; 2.02.03 3:33 7.9.83
0:... A33 0...... Aug
2:: 2.2.8.. :95. 20:... am.» roasm .oa 8.3.x. 8.3.x. u . .wafi. 8.3.x. w . .wwfi. 3.3.x. 3.3.x.
2.2.. A. .3. 2.3. A. .3. 2...... 0.43 3.3
2.3 wronoq :95 20:3 22 an 8 mannowmmc. tbmfi. 3.3.x. “6.3.x. 3.3.x. 3.3.x. 3.3.x. 3.3.x.
2...». Chm. Cum. 3.3. 2...: $8. Show
2.»:me mxuaoa rosmoq :95 .A. 5.x. 3.3.x. .N... .o\.. 6.3.x. 5.3.x. 5.8.x. 3.8.x.
8.me Chm. 2.93 A. 03 2:8. 3.3. 308
3.82:.» was note—.83 8 go...» roams. =95 QNQX. 9.3.x. 98.x. 9. .e\.. 9.3.x. 93.x. Wong
80$ 8...: 8.3. 8.8. 8.3. 2.8. 2.3.
.Lozmfi :95 Won—Eda .10.. >0...o<o=.c:. mafia...“ mmuofi. A933» 3.3.x. 3.3.x. 3.8.x. APEX.
2...: 2.3. 2.3. 3.8. 2.3. 3. .8 3.3.
rozmfi 20:2 Won—5Q. men Gawaigaozm. mcnnnmm Am. . ix. 3.3.x. 3.3.x. 3.3.x. A _ Luck. muruofi. 3.3.x.
2.3. A. .3. Co .v 3.03 2...... 3g. 3. . d
roams, :95 ”3.59. .d .93 Cu 3.3.x. Aobmfio 3.3.x. 3.3.x. 3.3.x. u . .38 3.5.x.
2.3. A. .3. :03 3.3. 2...... 3N3 E. .3
.383» Waggon. .m go...» 9532 Ion; 3.3.x. woken} 3.3.x. 3... .o\.. $.39; umbofi. mm. 3.x.
8.x... :93 2.3. 2.0 3 2:3. $.va Rho.

 

 

 

 

 

 

 

 

9::an 2.2.95. man; 0?.5 0.58.5 5.2.598 meow.
antenna. .: 33:538.

$36.0 21%.? a.» 26:2.

208” $55303 30 3.18.538 8 m 35:658.. 3:820?

 

8 Reina 2:33am 0% 32.2.0: 383083. 92:.qu 2.33 30

35

Table 1.3 Barriers to Working From Home

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Total Male Female
lmpractical 60.86% 64.06% 57.41%
(1 .59) (2.40) (2.04)
Not Allowed 19.96% 17.81% 22.28%
(1.24) (1.82) (1.67)
Not as Productive 1.27% 1.42% 1.10
(0.34) (0_.52) (0.45)
No Space 1.22% 0.75% 1.73%
(0.37) (0.47) Q53
No Equipment 8.15% 7.27% 9.08%
(0.97) (1.55) (1.14)
Employer Won’t Provide Tools 1.10% 1.27% 0.93%
(0.31) (0.50) (0.35)
Co-worker Jealousy 0.24% 0.35% 0.12%
(0.14) (0.25) (0.12)
Threatens .lob Advancement 0.45% 0.52% 0.37%
(0.19) (0.30) (0.22)
Not Taken Seriously 0.47% 0.50% 0.43%
(0.20) (0.29) (Q6)
Would Interfere with Family Life 1.75% 1.56% 1.96%
(0.46) (0.66) (0.63)

 

Source: National Study of the Changing Workforce 2002. Sample weights are applied to provide estimates
of population parameters. Standard errors are reported in parentheses.

Note: Tabulations are of responses to a multi-punch question.

36

 

CHAPTER 1

Expectations of Expecting: Women’s Actual and Expected Fertility Over the
Past Half Century

37

Introduction

As prefaced in the introduction to this dissertation, women’s employment,
education and childbearing decisions changed dramatically over the twentieth century,
and in particular during the last half. From the 19503 through the 19905 family sizes
shrunk from over three children to stabilize atjust under two, and first births after age 25
and even 35 became more common.4 These changes in fertility coincide with
unprecedented increases in women’s labor force participation and educational attainment
(O’Neill and Polachek 1993), gains in relative wages (Blau and Kahn 1997), growing
labor force attachment (Blau and Kahn 2007) and declining occupational segregation
(Jacobs 1989). They accompany other changes in the family formation as well, including
later marriages (Bianchi and Spain 1996), increases in assortative mating (Mare 1991),
higher divorce rates and cohabitation rates (Bumpass and Sweet 1989; Martin and
Bumpass 1989) and the introduction of oral contraceptives (Bailey 2006; Goldin and
Katz 2002, 2000). The actual fertility trends are well known and often referred to in
studies of other trends over this time period. But, did successive cohorts of women
anticipate these patterns in their fertility? Furthermore, did their plans for future
childbearing shape concurrent changes in employment and other behaviors? Cohort birth
expectations provide insight for investigating each of these questions.

This paper uses cohort level birth expectations and actual fertility data to examine
the secular trends in relationships between actual and expected fertility and early
occupational choices. Specifically, I compare expectations of ever having children,

completed family size, and timing of first birth among women born in the 1936 through

 

4 . . .
Author’s calculation usmg June Current Population Surveys 1971 through 1998.

38

1972 cohorts to the patterns in actual fertility that evolved for their cohort. Additionally,
I examine the relationships between birth expectations, actual childbearing, and the
characteristics of women’s chosen occupations to see if women who expect to have
children appear to choose more family friendly occupations early in their careers.

Similar to previous studies, I find cohort expectations correspond reasonably well
with their own completed family size. However, nearly all cohorts underestimated the
likelihoods of delaying their first birth, having only one child, or remaining childless.
Thus, I conclude cross-cohort fertility patterns were generally not anticipated.
Furthermore, I find most of the differences between the jobs mothers and childless
women hold are due to sorting at or after the first birth rather than early in their careers.
Conceptual Framework

There are (at least) three reasons why birth expectations are of interest for
research and policy. First, their relationship to future childbearing may inform our
understanding of family formation. Second, expectations of future childbearing may
influence current and future behavior including marriage, educational investment, and
employment. Third, fertility expectations may be used to generate forecasts of future
fertility. Neither of these relationships depends on the other; birth expectations may
influence other current behaviors or convey useful information about family formation
even if they do not predict future fertility. In this study, I am interested in the first two
applications of expectations data.

The “other current behavior” of interest in this study is choice of occupation.
When selecting occupations, women who expect to become mothers may trade off wages

for job attributes that will help them to combine paid employment with future or current

39

care-giving roles (Altonji and Blank 1999; Fuchs 1989; Filer 1985; Mincer and Polachek
1974). These job attributes may include reduced work schedule, paid time off work,
flextime or job sharing policies, on-site childcare and any other policies or practices that
may reduce the costs of combining paid employment and Childrearing activities.

There are certainly other reasons why men and women or women who expect to
become mothers and those who do not would differ in their tastes for “family friendly”
job characteristics. Research has shown evidence of mediating mechanisms (e.g. Hakim
2002; Cable and Judge 1994; Judge and Bretz 1992; Forgionne and Peeters 1982). Yet
differences in the incidence of childbearing and Childrearing responsibilities have
consistently been found to explain a large portion of the observed wage differentials and
occupational segregation by gender. Estimates of the “family gap” in pay find a 10 to 15
percent difference in wages between women who have children and otherwise similar
women who do not (Waldfogel 1997; Korenman and Neumark 1992; Fuchs 1989) and as
the gender wage gap declined over the 19803, the family gap increased (Waldfogel 1998).
Analyzing the trends in expected childbearing and early occupational choices over this
time period could inform our understanding of the trends in gender and family wage
differentials.

One key challenge. faced when comparing trends in expected and actual fertility
across cohorts, and fertility timing in particular, is that marital behavior and birth spacing
may change across cohorts as well. The correspondence between expected and actual
future childbearing will likely depend in part on current marital status and past
childbearing. This may muddle comparisons of expected and actual fertility across

cohorts at a given age. If for example, women are more uncertain about future

40

childbearing before they are married, we might find the earlier cohorts’ expectations at
age 20 better approximated their actual fertility than later cohorts simply because later
cohorts married later. Ideally, this possibility would be examined directly by comparing
expected and actual fertility for cohorts as a whole and for the married and unmarried
proportions of each cohort separately and the findings could inform our understanding of
relationships between marriage and family formation.

Unfortunately expectations data were not collected among unmarried women in
most surveys until the 19903. Thus, in keeping with previous studies, I restrict the
majority of my analysis to married women. Additionally, I restrict my analysis to
childless married women. Conceptually, the transition to motherhood is a pivotal event
and marks the assumption of new time-intensive responsibilities. Empirically, the
differences in educational attainment, labor force participation, wages and other
economic outcomes are greatest between women who have had children and those who
have not rather than by parity.

Previous Research

Most studies of expectations data seek to relate birth expectations to future

childbearing. In general, there are three ways to compare expected and actual

childbearing. First, average expectations in a cross-section of women can be compared to

the period measures of fertility such as the total fertility rate.5 Examples of this type of
comparison include Fischoff et al. (2000), Moore (1981) and Peterson (1995). The

second type compares individual level expectations to future individual level fertility

 

5 The total fertility rate is an estimate of the total number of children a women will have. It is derived by
summing the current age specific fertility rates over the remainder of her fertile lifecycle. This is referred
to as a period measure because it uses age specific fertility rates at a given time. It hinges on the
assumptions that women who have not yet completed childbearing will experience the current age specific
fertility rates over the rest of her fertile lifecycle and will live at least to the end of their reproductive life.

41

using longitudinal data. Examples of this type include Trent and Crowder (1997, Westoff
and Ryder (1977) and Freedman et a1. (1980). Third, studies may compare past cohort
level expectations to actual cohort fertility. By cohort level expectations and cohort
fertility, 1 mean average expectations within a cohort are compared to average actual
fertility among members of that cohort surveyed at some point after expectations were
surveyed. The follow-up measure of actual fertility is generally not based on the same
sample of women. Cohort level measures have been used in Westoff (1981), Hagewen
and Morgan (2005), and O’Connell and Rogers (1983) and are the preferred measure for
this analysis.

Technically, expectations and actual fertility can be compared at the individual
level across cohorts as opposed to comparing past cohort average expectations to actual
cohort average completed childbearing. Conceptually, the former comparison would
address questions about individuals’ abilities to anticipate their own fertility rather than
whether or not the secular trends in cohort fertility were anticipated. Practically, few data
sets contain a long enough time series of data for a broad enough band of cohorts to
compare average individual level correspondence between expected and actual
childbearing across cohorts.

Results among studies that compare expected to actual fertility using any of the
measures described here have been mixed but generally, studies that compare cross-
sectional expectations to period fertility measures or average cohort expectations to
average cohort fertility find a closer correspondence than individual level comparisons.
For example, Freedman et al (1977) examined the correspondence between childless

couples’ expected family size in 1962 and the same couples’ actual family size in 1977

42

and found 61.3 percent of couples who were childless in 1962 had expectations that
exceeded their achieved 1977 fertility. Yet, Westoff (1981) find individual differences
appear to average out at the cohort level. Much of the existing evidence is based on
narrow bands of cohorts (e. g. Hagewen and Morgan 2005) or average across single birth
year cohorts (O’Connell and Rogers 1983). In both cases, the ability to make
comparisons across cohorts is limited.

The existing empirical evidence of women sorting into more family friendly
occupations or career tracks based on their expectations of future childbearing is mixed.
Averett and Whittington (2001) hypothesize women who expect to have children may
select into jobs with maternity leave benefits but they find no evidence of selection.
However, Blakemore and Low (1984) use fertility expectations to examine college major
choices and find women with higher expected fertility choose majors that lead to jobs
with less skill atrophy. Moreover, they find fertility expectations explain a large share of
the differences in distribution of male and female college majors. Baber and Monaghan
(1988) also find evidence of selection in their study of occupational choices among 250
college women. Although nearly all expected to have children, those pursuing less
conventionally female careers were less child oriented. Related studies, Mott and Mott
(1984), Powers and Salvo (1982), Presser and Baldwin (1980), Waite and Stolzenberg
(1976, 1977), provide evidence of a negative relationship between childbearing
expectations and women’s likelihood of working at all.

Data
The actual and expected fertility data come from the 1971 through 2004 June

Supplements to the Current Population Survey (CPS). The CPS is a nationally

43

representative household survey. In the June Supplement women in the household ages
15 to 44 are asked to answer questions about their fertility histories and, in selected years,
about their future fertility expectations. From these data it is possible to construct various
measures of average cohort expected and completed fertility. These measures are
explained here. As stated previously, all expectations data are for women who have ever
been married and are currently childless.

Expectations questions in the CPS collect information about expected family size
and expected timing of the first/next birth. I use both types of information to create three
measures of cohort birth expectations. Women who are eligible for the expectations
supplement (married and of childbearing age) are first asked whether or not they expect
to have any (more) children. Possible responses are “yes”, “no” and “uncertain”. Since I
consider childless women only, answers to this question indicate whether or not a woman
expects to eventually become a mother or is uncertain about having children. I refer to
these expectations as “expected motherhood”.

In survey years 1973 through 1983, women who say they do expect to have
children are also asked how many they expect to have in the next five years. Among
childless women this question refers to the timing of first births. Therefore, I refer to
these expectations as “expected first birth timing”. Responses to this question are
combined with responses to the first question to identify women who expect to have
children but not in the next five years. I refer to this group as women who “expect to
postpone” childbearing.

Expected number of children is asked of those who respond “yes” they expect to

have children; obviously those who respond “no” expect not to have any children. Those

44

who respond uncertain are not asked further questions about expected number of
children. Clearly their expected number of children is not zero but, without more
information, I cannot determine the underlying expectation. To arrive at expected family
size estimates from these data, I assume those who respond “uncertain” have underlying
expectations equal to the mean of those within their cohort and marriage age quartile who
respond “never” (i.e. those who expect 0 children) and those who respond “yes”. Clearly
this assumption is unsatisfactory, but given the lack of a superior alternative or further
information in this or other data sources, I opt for this measure because it is transparent
and easily replicable. I have experimented with other similar assumptions and the
general conclusions of the analysis do not change. Throughout the analysis I refer to
these expectations as “expected family size”.

Average cohort actual fertility measures are created using retrospective birth and
marriage histories. Since the expectations data is limited to ever married women only,
the average cohort actual fertility measures are based on retrospective birth histories for
only those women who were married at the time expectations were surveyed. For
example, expectations among married women in the 1950 cohort at age 22 would be
compared to actual fertility among women born in 1950 who had married at or before age
22.

Figure 1.1 displays the trends in cohort completed family size and timing of first
birth over the twentieth century and highlights the cohorts I am able to analyze. The
broken lines indicate the range of cohorts I compare at older ages (age 27 to 35) and the
solid lines illustrate the cohorts I can compare in their twenties. Since expected the

expected timing of births survey questions were not fielded before 1975 or after 1983, the

45

expected and actual timing analysis depicts a narrower range of cohorts than is shown in
Figure 1.1.

To analyze occupational characteristics I combine information from the March,
May and June CPS surveys. Wage and hours information come from the March Annual
Social and Economic Study Supplement. Other occupational characteristics data are
obtained from the May Workplace Topics Supplement. In both cases I obtain mean
occupational characteristics by year, detailed occupational code and educational
attainment and match the occupational characteristics to observations in the expected and
actual fertility data. For more information about the occupational characteristics data
refer to Appendix Al.

The data used to calculate occupational characteristics include workers of all ages,
both male-and female. I include workers of all ages because I am attempting to capture
the opportunities available in the occupation overall rather than at to any one point in a
career. I include men because recent studies suggest successive cohorts of women have
exhibited labor market behavior that more closely approximates past cohorts of men than
women (Blau and Kahn 2007). Thus later cohortsjust entering a given occupation may
expect opportunities more similar to men who are already established in that occupation
than women. Furthermore, including men increases the precision of the estimated
occupational means.

Results

Table 1.1 contains the means and standard deviations of key covariates and

occupational characteristics considered by expected and actual childbearing. These

means are based on all childless married women across all cohorts at all ages. The

46

broken line separates tabulations of timing expectations group from tabulations of
motherhood expectations; as explained, timing expectations were only surveyed in a

subset of years (1976 through 1983) while motherhood expectations were surveyed

between 1971 and 1992.6 Thus comparisons of sample means across the broken line are
invalid.

Looking at the sample means to the left of the broken line, the childless women
who expect never to have children do not appear to conform to the stereotype of “career
women”. On average they are older than other childless women but they are also less
educated and less likely to be in the labor force. Their real mean weekly occupational
wages are higher than the wages in occupations women who expect to have children hold
but they are lower than those of women who are uncertain about having children. Yet.
none of these differences are particularly large when compared to the differences between
mothers and childless women.

Figure 1.2 plots expectations of motherhood among 20 to 23 year old childless,
married women against the actual percentage of ever married women who become
mothers before age 40. The lower series of expectations (the circles) include only those
women who say they do expect to have children and the top series (the squares) add the
percentage of women who say they are uncertain about ever having children. The series
of x’s show that the majority of ever married women do eventually have children. Also,

the circles show the majority of childless women expect to become mothers.

 

6 Motherhood expectations were surveyed in the 1998 June CPS but retrospective marital and childbearing
histories are not complete enough to support the analysis and although discrepancies in the questions were

not apparent from the codebooks tabulations suggest there may have been a change in the survey structure

or delivery that limits comparability.

47

When added to the percentage who are uncertain, the total percentage of women
who expect or are uncertain about childbearing roughly approximates the proportion who
become mothers. This sort of rough approximation has lead other cohort level studies to
conclude there is a reasonable correspondence between expected and actual fertility when
examining a single cohort or a small band of cohorts. Yet with the cross-cohort trends
visible in Figure 1.2, it seems the trend in expected motherhood is positive while the
cross-cohort trend is actual motherhood over the comparable region is flat. Although the
proportion of women who expect to have children increased by only 6.98 percentage
points from the first data point to the last data point, the slope of the OLS regression line
fit to the “Expect” series is 0.24 (or 'A of a percentage point) with a standard error of
0.07. Conversely, the 310pe of the OLS regression line fit to the actual motherhood data
across the cohorts where expectations were surveyed (1952 to 1972) is -0.08 with a
standard error of 0.09.

Figure 1.2 also displays the cross-cohort changes in the proportion of women who
are uncertain about childbearing (the difference between the squares and the circles) and
the cross-cohort changes in the proportion of women who expect never to have children
(the difference between 100 and the squares). Across cohorts, the percentage of women
who are uncertain appears to be roughly constant whereas the percentage of women who
expect never to have children is decreasing. Indeed the cross-cohort average percentage
of women age 20 to 23 who are uncertain about having children is 9.01 percent with a
standard deviation of 2.08 and no significant cross cohort trend (,8 = -0.03, s.e. = 0.08).

Therefore the positive trend in the percent who expect or are uncertain (the squares)

48

comes primarily from increases in the proportion of women expecting to have children
rather than increases in uncertainty.

Figure 1.3 compares expected and actual first birth timing by age at which the
expectations were surveyed. Specifically, the dark circles indicate the percentage of each
cohort that expected to postpone their first birth (i.e. expected to have children eventually
but not the in the next 5 years). The lighter colored diamonds plot the proportion of each
cohort that was married and childless at the age the timing expectations were surveyed,
did not have a birth in the ensuing five years, but did have children before age 40. The
actual timing data do not include women who never had children.

The large difference between the two series in each graph in Figure 1.3 shows that
women substantially underestimated the likelihood of postponement. For example, at
ages 18 to 22 only 3.58 percent of women in the 1950 birth cohort who were married
expected to wait five years until having their first child yet 36.8 percent of women in this
group actually did wait five years or more. Although, the trend in cohort expectations in
all age groups does indicate planned postponement was more prevalent in later cohorts.
The only group of women whose expectations of postponement closely approximated the
actual incidence of postponement was the group born after 1950 and surveyed at ages 27
to 35.

Figure 1.4 compares expected completed family size to actual family size by age.
The top left graph implies expected family size in most cohorts was quite close to actual
family size. Yet underlying that correspondence, women’s expectations of specific
family sizes and of remaining childless do not match well with actual incidence. The

other three graphs show cohorts consistently underestimated the likelihood of remaining

49

childless, having only one child or having more than two children. Thus, the close
correspondence in the upper left graph is primarily due to the large numbers of women
expecting and having two children. Across this particular range of the cohort fertility
distribution, the modal and average family size was two.

Underlying the seemingly close correspondence between expected and actual
family size in the top left graph of Figure 1.4, there are differences by educational

attainment. Figure 1.5 shows the expected family size of women in each cohort at ages

23 to 26 compared to their average actual completed family sizes.7 Women in each
cohort with less than a high school education had expectations that were systematically
lower than their actual fertility. High school educated women and those with some
college appear to have had expectations that approximate their cross cohort fertility
patterns quite well and women with a four year college degree slightly over estimate their
fertility, but not to the extent that the least educated women under estimate theirs.

In both Figures 1.4 and 1.5, there is a slight U-shaped pattern in the completed
family size series for all 20 to 23 year olds (Figure 1.4 top left graph) and for college
educated women (Figure 1.5 lower right graph). Interestingly, expectations at ages 20 to
23 (top left graph in Figure 1.4) and expeCtations of women with a four year college
degree at ages 23 to 26 (lower right graph in Figure 1.5) appear to match the U-shaped
pattern reasonable well. This suggests the cohorts born in the early to mid 19603 did
anticipate slightly larger families than the women born in the mid 19503 had. Indeed, this
is one of the few comparisons in which successive cohorts of women do appear to have

anticipated cross-cohort changes in childbearing behavior. At ages 20 to 23, average

 

7 . . . .

This comparison had to be made among women who were childless at ages 23 to 26 rather than earlier
because it relies on cross-sectional data and educational attainment cannot be reliably determined before
most women have finished college.

50

expected cohort family size changes from 1.91 children in the 1955 cohort to 2.34
children in the 1971 cohort. The inflection point in the average actual cohort family size
series also occurred at the 1955 cohort; actual family size reached a low of 2.00 children
and by the 1964 cohort (where the data series ends) had reached 2.09 children. The
average expected family size for the 1964 cohort was 2.10.

Table 1.2 contains the results of the occupational characteristics analysis. I
present these results as average cross-cohort differences rather than individual cohort
differences because there was vary little cross cohort variation in the relationships
between expected and actual fertility and occupational characteristics. Appendix Table
A1.2 contains the means and standard deviations of all characteristics for all groups
(women who expect, are uncertain, expect never, already have a child under age 6,
already have a child over age 6, expect to have a child within the next 5 years, expect to
postpone, did have a child within 5 years, did postpone) so that the reader may construct
any comparisons desired. All figures (differences reported in Table 1.2 and the means
and standard deviations reported in Table A1.2) are adjusted for race, ethnicity,
educational attainment, age at marriage, current age, and marital status (divorced,
separated, widowed or currently married).

Table 1.2 presents a clear pattern: the differences in occupational characteristics
between childless women and those who have already had children are generally larger
than the differences in occupational characteristics among childless women based on
their expectations of future childbearing. That is, the differences reported in each column
are larger than the difference between the figures in the lefi and right column of any row.

For example, in the second row of figures in the top panel of Table 1.2, both childless

51

women who expect to have children and those who expect never to have children work in
occupations that pay approximately $10 more per week (in real 1982—84 dollars) than
women who had their first child within the last 6 years, which implies virtually no
difference in real mean occupational wages among childless women by expectation.

Although, given the large number of observations, the differences in mean
occupational characteristics between women who expect to have children and those who
do not are nearly all significant (see Appendix Table A1.2), few remain significant after
controlling for race, ethnicity, educational attainment, marriage age, current age, marital
status and cohort labor force participation. For example, the difference in raw real mean
occupational wages between women who expect to have children and those who expect
never to have children is $7.12. However, after controlling for these other factors, the
difference due to childbearing expectations is only $0.33 and the estimates imply real
mean occupational wages are higher in the occupations held by women who expect to
have children. Women who expect to have children are more likely to work in
occupations with a higher proportion of female employees, but there is little evidence of
systematic selection into family friendly jobs or lower wage jobs prior to the birth.

This same conclusion holds when comparing childless women who expect to have
children in the next five years and those who expect to postpone. When other factors are
controlled for, the estimated differences in mean occupational characteristics by
expectation (the row differences in the lower panel of Table 1.2) do not reach
conventional levels of statistical significance. The raw means reported in Appendix
Table A1.2 suggest childless women who expect to have children earlier may be

somewhat less likely to be employed (7 percentage point lower labor force participation

52

rate) and may work in occupations with slightly shorter work weeks (0.97 fewer hours
per week on average) but these differences are small when compared to the labor force
participation rate and mean hours per week among women who have had children.
Discussion and Conclusion

Although, in keeping previous studies of expectations at the cohort level, the
foregoing analyses indicate the level of motherhood and completed family size
expectations may have been a decent approximation of actual fertility; with few
exceptions the cross cohort trends indicate women did not anticipate the patterns of actual
fertility which prevailed in the latter half of the twentieth century. From the 1952 to 1972
cohorts, the proportion of women who ever became mothers was constant. Yet, the
proportion of women ages 20 to 23 who expected to become mothers was increasing.
From the 1935 to the 1956 cohorts the proportion of women who postponed childbearing
was substantially higher than the proportion who expected to postpone across all age
groups, except among the 1951 to 1956 cohorts at ages 27 to 35.

The only measure by which cohort expectations appeared to lead changes in
cohort fertility is completed family size but this is likely due to the strong modality of
two child families over the time period considered. In nearly all cohorts, expectations of
large families (three or more children) were significantly lower than the actual occurrence
of large families. Expectations also under estimated the percentage of single child and
childless families across cohorts. Also, the least educated women in all cohorts
systematically underestimated their completed family size.

With respect to occupational characteristics, where differences exist they are

largest between childless women and mothers. There is little evidence of systematic

53

differences in the occupations held by childless women according to their expectations of
motherhood or first birth timing. These findings suggest most of the differences observed
between childless women and mothers results from job changes at or after the first birth.

The fact that no systematic difference was found in occupational characteristics
between women by expectations raises the question: who are the women who expect
never to have children or who expect a late first birth? In the simple descriptive statistics
presented in Table 1.1 it seemed many of our priors about the childless career women did
not apply to women who expect never to have children. Unfortunately for women who
expect to postpone their first birth there is little information available in the CPS data
beyond the descriptive statistics in Table 1.1. However, in the 1998 survey, the
expectations supplement included a follow up question about the influences behind the
expectations of having (or not having) children. Reponses to this question are tabulated.
by motherhood expectation in Table 1.3. Since this information was only surveyed in
1998 it cannot be compared across cohorts and ages alongside the rest of the analysis but
to the extent that some of these influences may be constant across cohorts and time, the
tabulations in Table 1.3 may be informative.

Not surprisingly, a high proportion of childless women who say they will never
have children indicate medical concerns are an influence (25.67 percent of those who
never expect to have children as compared to 10.86 percent of those who are uncertain
about having children). However, those who expect never to have children are also more
likely to reference childcare concerns, and are less likely across all age groups to
reference ability to afford children. Career, job, and educational decisions are no more

important among the “never” group than among the “expect” and “uncertain” groups.

54

With this additional information, it is less surprising that there were few differences in
occupational characteristics by expectation.

The foregoing analyses are subject to (at least) three important limitations. First,
although efforts were made to restrict the influence of changing marriage behavior, the
cross-cohort trends presented are arguably still confounded by changes in the timing of
marriage which were occurring over this time period as well. Since expectations data for
single women is available in only a subset of survey years the importance of this problem
is difficult to evaluate.

Second, the shortest forecasting horizon considered in the expectations data is five
years. Any changes in expected timing of births across cohorts over shorter intervals
cannot be captured. In the occupation characteristics analysis, the long forecasting
horizon may be part of the reason no significant differences were found among those who
expect a birth in the next five years and those who expect to postpone. Selection could be
occurring before the birth but only within a year or so (e. g. around the time the pregnancy
begins or when the couple is trying to conceive) and thus is not apparent in these
analyses. Selection in and out of the labor force and across jobs within twenty months of
the birth and up to 1.5 years after is directly examined using a different data set in
Chapter 2.

Third, since the level of analysis for the job characteristics considered is the
occupation, selection into certain types of jobs within a given occupation cannot be
captured. This concern is somewhat mitigated by the use of detailed occupation codes to
reduce within category variance in the nature of the work. Significant differences were

found between mothers and childless women for nearly all occupational characteristics.

55

Figures and Tables

Figure 1.1 Data Coverage Across Secular Trends in Family Size and Birth Timing

 

 

 

S,’ ' 'C
i . . ~ . 'completed Family Size . 1
l l
. I l l
0.00. ' “f. l .3,
i 0 0...... l “no j
on
00.0.. | .0. O... | ‘3‘
' We”...
' l
' I 7 I ‘l l 1 ' I x -.
" 1920 , - ' 1940 1960 4,, 1930..
. ,. Cohort . . 6
r . t ‘- 1 . .. , .-. % Had First Child by Age 25 ; ;:. w
80 | |
' l
70 . #0.. :
O . .
‘60 0.. “a. . . l O... '0’ l
. C
50 7.... O." I ‘ . “no ..
4b ~flfl.~" l I..."0'.fl.u ‘ A
i 1.960" .. ‘ .1929 I d 1940 H " '1960 :-‘ 19gb . ‘

Cohort

A

Source: June Current Population Survey 1979 through 2004, excluding survey years 1986 through 1988.

Notes: These graphs include all women regardless of past or current marital status. Prior to 1979 and in
1986 through 1988 the CPS surveyed family size among ever married women only so these years are
excluded from these figures. Completed family size is measured as the total number of children ever born
to women age 40 and above. The solid lines indicate the earliest and latest cohorts included in the analysis
when comparing expectations at 0 to 3 years of marriage among women in their twenties. The broken lines
indicate the earliest cohorts and latest cohorts included in the analysis when compared at ages 27 to 35.

56

Figure 1.2 Expected and Actual Motherhood Across Cohorts

Expectations at Ages 20 to 23

 

 

100~
I
I. .I.._._
90—..14" x a: -—‘--;r"-’ J] 'r' '
. ~-~-«.......,_.,._ 7.ix_._’<.fi.? r- :1..— .-
xxx" ‘_____,.______._I.."
80- L'Ir’t—’ . . Co
‘— o
70-
60-
50—
r ‘_F T ' I I
1935 ' 1945 1955 1965 1975
Cohort

 

O % Expect to Have Children
I % Expect + % Uncertain
x % Have Child by Age 40

 

 

 

Source: June Current Population Survey 1971 through 2004. (Expectations from 1971 through 1992 only).

Notes: These figures exclude never married women since expectations were not fielded among that group
in most survey years. The difference between the “% Expect + % Uncertain” series and 100% is equal to
the percentage of women who expect never to have children. The right panel includes expectations of

women in their twenties within three or fewer years of their first marriage.

57

Fi-ure 1. 3 Expected and Actual Timin- of First Birth Across Cohorts
. an»? ,“7 4.. we I”? in? Wt xvii "‘s ”retainer!

1“.” ‘6 9' (cow ,.

 

 

 

’QW““€Wm—,TWW: " "
’" “‘"““"Tr:":7fi“r*"""f:f:_ ‘

 

 

 

 

 

Source: June Current Population Survey 1971 through 1995. (Expectations from 1971 through 1983 only).

Notes: The figures plot the percentage of women who expect to have children but not in the next five years
against the percentage who actually delayed their first birth for five years. Ages indicate both the age at
which the expectation was given and the age at which the five year span begins (e.g. a representative
women in the lower graph may have been 28 when here expectation was surveyed and would have
responded about her expected fertility between age 28 and 33. Expectations data are restricted to currently
married childless women and actual fertility data are restricted to ever married women for comparability.
Maniage duration is not restricted in these figures; women in the 27 to 35 group could be just married or
could have married in their early twenties.

58

Fl

ure1.4 Ex-ected and Actual Famil Sze at A-e 20 to 23

    

 

 

 

 

 

 

 

 

 

 

 

 

'1‘

Source: Je Current Ptlopulaon Survey 1971 through 2004. (Expectations fl’om 1971 through 1992 only).

 

Notes: Expectations and actual fertility data are restricted to ever married women and include expectations
given at ages 20 to 23. Completed family size is measured afier age 40.

59

    

Figure 1. 5 Expected and Actual Completed Famil Size b Education
a! fit: s3"? fl ”#3 's‘

 

 

 

 

 

 

 

 

 

 

 

. - . ' ‘ .~. are" m 4.. ~~
Source: June current Population survey 1971 through 2004. (Expectations fiom 1971 through 1992 only)

Notes: Expectations are measured at 23 to 26 years of age. Completed family size is measured after age 40.

60

._.m0_m 3 9.305 92528 84 03qu ._

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

@0200 3099.380 m2: 3 4 @6880 .255 om 2an En: mwamfi

GEE—8m R 035.0% mm 002293 3 73908 a: _ @603 $31: mxuog 8 30:83 <<\

mx o2 C0833: mxvoo. Zn<2 GEE Aafim F282 0 <83 08625 035 A003
>mo N900 3.0m uNbu NT: . 3.2 5.3 3.00

8.50 8.08 8.08 81d _ 8.30 8.58 8.23
A Em: wago— 0.0A 0.00 0.3 0.00 _ 0.3 0.00 0.00

8&5 8.va 8.80 8.30 8me 8?: 8.0.3
Em: warez 0.3 0.3 0.8 0.3 _ 0.3 0.30 0x3

8me 8.33 8.58 8.03 _ 8.000 8.000 8&8
moan Oozmmo 0.3 0.8 0.8 0.: . 0N0 0.8 0.:

8.530 8.30 8.30 8.30 _ 8.30 8.30 8.30
Oozomnu. + 0.00 0.? ohm 0.8 _ 0.8 92 0b;

8.30 853 8.30 8.30 8.30 8.30 8.30
Zuiwmo >mo 3.30 8.3 For: 2.3 — 2.0m No.3 3.3

8.20 8.03 8.0.8 0.30 — 8.30 8.03 8.08
.503 038 080 0.3 0.3 0.3 0.00 0.00 0.0m
vwnwommmzo: 8.03 80$ 8.0.3 8.000 _ 8.3.: 8.30 8.08
0302303 3.03 0.5: 0.8» 3P0?“ _ 3.000 mow 3.05

 

 

208m“ 22 m: $530:

£30 08308 28: 9638203 om :2 353m 9:96: «<95 3.80 305 :6 2:15 om 033. $8338 :8 <m_:om 5 9m $559..

8E3: m8 :2 BE: 8 :6 Bow: <m_:om 8:25 Scan: :50 962: 8 :30 0 00:0 3 9o :8: :5 <33 Ba :58 «<00 868” 8 0869.0.

61

Table 1.2 Estimated Average Differences in Labor Force Participation and

Occupational Characteristics by Expected and Actual Motherhood and Timing of

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

First Birth
Expect — Moms <6yrs Never - Moms <6yrs
In Labor Force 0.30" 0.22M
(0.00) (0.01)
Ave. Wkly Wages 9.57" 9.24"
in Occ. (1.75) (1.96)
Average Hours in 0.88” 0.76M
Occ. (0.16) (0.18)
% Work Part-Time -2.06** -l .28“
in Occ. (0.39) (0.46)
% of Weeks per 0.41 0.34
Year P.T. in Occ. (0.28) (0.32)
Ave. Days per 0.14" 0.11"
Week in Occ. (0.01) (0.01)
% Have Flextime 0.01” 0.01 **
in Occ. (0.00) (0.00)
% Female in Occ. -1.78** -4.54**
(0.28) (0.32)
% Mothers in Occ. -l.92** -1.83**
_________wa______ _@Q___
Expect Early — Had Early Expect to Postpone — Had Early
[n Labor Force 0.32“ 0.35”
(0.02) (0.02)
Ave. Wkly Wages -4.35 -1.34
in Occ. (4.08) (6.50)
Average Hours in -0.97** -0.58
Occ. (0.37) (0.65)
% Work Part-Time 2.15““ 1.92
in Occ. (0.91) (1.62)
% of Weeks per 0.77 1.32
Year P.T. in Occ. (0.64) (1.07)
Ave. Days per -0.05* -0.06*
Week in Occ. (0.02) (0.03)
% Have Flextime 002* 004*
in Occ. (0.01) (0.02)
% Female in Occ. -2.70** -3.45**
(0.60) (1.07)
% Mothers in Occ. -3.62** -3.66**
(0.16) (0.28)

 

 

 

 

 

Notes: See Appendix Table A12 for number of observations, means and standard deviations of all
occupational characteristics. This information can be used to construct any test statistic desired. The
differences reported above are based on regressions which control for race, educational attainment,
marriage duration, age, marital status and cohort labor force participation. Means reported in Table A1 .2
are raw means.

62

 

63

Appendix A1

Description of Data Sources

Chapter 1 uses assorted years of data from the 1971 through 2004 Current
Population Survey (CPS) March Annual Social and Economic Study, May Work
Schedules Supplement and June Marriage and Fertility Supplement. The CPS is a
nationally representative survey of 60,000 households. Basic labor force data are
collected monthly and special topics information is collected in periodic supplements.

The June Marriage and Fertility Supplement is the source for all actual and
expected fertility data. Years used include 1971, 1973 — 1977, 1979 — 1983, 1985 —
1988, 1990, 1992, 1994 — 1995, 1998, 2000, 2002 and 2004. Since 1971 the June CPS
Supplement has collected fertility histories from ever married women. Beginning in
1976, never married women were also surveyed, however not all surveys during the
19803 included never married women. Women of childbearing age (generally 14 to 44)
are asked how many children they have ever had. Through 1995 they are also asked the
year of their first birth and the year of their first marriage. Using this information in
conjunction with age at time of the survey, I construct cohort age specific fertility
histories for women age 40 and older (the majority of whom have completed
childbearing) and can condition on timing and duration of marriage as well as
childlessness up to a given age.
Source and Use of Expectations Data

The June Supplement also surveys fertility expectations among women of
childbearing age. Expectations were surveyed in 1971, 1973 — 1977, 1979, 1981, 1983,

1985 -1988, 1990 and 1992. Expectations were also surveyed in 1998 but the questions

64

used changed slightly and tabulations suggested data from that year may not be
comparable to previous years. Although proxy respondents over the age of 15 are
permitted to answer other questions in the CPS basic and supplement surveys the
expectations questions had to be answered by the respondent herself. In most years
eligible respondents were between the ages of 18 and 39 but the lower bound was as low
as 14 in the early surveys and the upper bound was 44 in the early 19805. Since
expectations are tabulated by cohort and by age throughout most of the analysis, these
differences are not generally a problem. In the early surveys and in assorted years in the
19803, expectations questions were fielded to ever married (and sometimes currently
married) women only. Given these data limitations and the arguments for examining
ever married and never married women separately presented in the main body of this
paper, I exclude never married women from the analysis.
The expectations questions used in the analysis include:

1. Looking ahead, do you expect to have any (more) children?

2. How many (more) do you expect to have?

3. How many (more) do you expect to have in the next 5 years?
Possible responses to the first question are “yes”, “no” and “uncertain”. These responses
are tabulated among childless women. to generate the expectations of motherhood data.

In most years the second question is fielded as a follow-up to Question 1 and

asked only of women who respond “yes” to Question 1. In some survey years Question 1
was not fielded and Question 2 was the first question in the expectations battery.
Possible responses to Question 2 in most survey years are numbers between “1”

(meaning the respondent expects to have one child) to “6” (meaning the respondent

65

expects to have six or more children).8 When Question 1 is not fielded “0” (meaning the
respondent expects never to have children) is included among the possible responses.
Yet, in those survey years, “uncertain” is not a possible response to Question 2.
Furthermore, tabulations of the “0” responses in years when Question 1 was not fielded
do not appear to be comparable to the “Never” response to Question 1 in other years.
Therefore, I include only years in which both Question 1 and Question 2 were fielded.

Using responses to Questions 1 and 2, 1 construct the family size expectations
data. As mentioned in the main body of the text, it is not possible from these questions to
determine how many children women who respond “uncertain” to Question I expect to
have. To address this problem I assume women who respond “uncertain” to Question 1
have underlying expectations equal to the mean of expected family size among those who
respond “yes” to Question 1 and those who respond “No”. Those who respond “No” to
Question 1 implicitly indicate an expected family size of “0” and those who respond
“Yes” report their expected family size in response to Question 2.

Question 3, in conjunction with Question 1, provides the source for the first birth
timing questions. Question 3, like Question 2, is fielded only to women who respond
“Yes” to Question 1. Possible responses to Question 3 include “0”, “1”, “2” or “3”.
Women who respond “Yes” to Question 1 and “0” to Question 3 are coded as expecting
to postpone their first birth. That is, these women expect to have at least one child
eventually but they expect to wait at least five years to have their first child.

Unfortunately Question 3 was only fielded in survey years 1973 — 1974, 1975 — 1977,

 

8 As in Question 1, Question 2 asks the respondent to provide the expected number of additional children
(i.e. children they have already had are not counted in the expectation). Since responses to this question are
tabulated for childless women only throughout the analysis. the interpretation provided here is specific to
childless women.

66

1979, 1981 and 1983. Therefore, throughout the analysis the range of cohorts across
which I am able to comparing first birth timing expectations is smaller than the range
across which I can compare other expectations. Also, Question 3 was only fielded to
currently married women whereas Questions 1 and 2 were usually fielded to ever married
women (and in some years to all women regardless of marital status). Information on
separations, divorces and spousal deaths are not available consistently enough to match to
restrict the actual first birth timing data to only those women who were currently married
at the time expectations were surveyed. As in the rest of the analysis, I do restrict actual
fertility data to ever married women but the inability to restrict to women who were
currently married at the age expectations were surveyed may affect the comparisons
between expected and actual first birth timing.
Construction of Cohort Actual Fertility Data

Using the cohort birth and marriage histories of women ages 40 and older it is
straightforward to construct measures of actual fertility. Throughout the analysis I
restrict the actual fertility data to ever married women. When expectations are tabulated
by age group I restrict the actual fertility data to women who married for the first time
before the age at which expectations were surveyed. When expectations are tabulated by
marriage duration I restrict the actual fertility data to women for whom the specified
number of years (e.g. 0 to 3) had elapsed between their first marriage and the time at
which actual fertility is measured.

The questions used to construct actual motherhood and family size data are:

4. When was (respondent ’s) first child born?

5. When did (respondent) get marriedfor the first time?

67

6. How many live births, if any, has (respondent) ever had?
Unlike the expectations questions, these questions may be answered by a proxy
respondent. Construction of actual motherhood data and completed family size data is
straightforward. To construct the actual postponement measure I take all women who
were childless and had married for the first time before the age at which expectations
were surveyed, and code those who do not have their first birth in the ensuing five years
but do have at least one birth before age 40 as having “postponed” their first birth.
Again, I am unable to precisely observe marital status at the age when the expectations
were surveyed, so I am comparing the fertility history of ever married women who
married for the first time before the age at which expectations were surveyed to the
expectations of currently married women.

As mentioned, to ensure I have observed all first births I use only women age 40
and older. Arguably restricting to an older age would be more precise since the average
woman is still biologically capable of bearing children at age 40. Yet since the vast
majority of first births occur before age 40 this assumption is unlikely to affect cohort
averages, and it allows for comparison across a broader range of cohorts.

Sources and Use of Occupational Characteristics Data

The March Annual Social and Economic Study and May Work Schedules
Supplement are used to obtain estimates of occupational characteristics at the time
expectations were surveyed. To do so I calculate the mean value of each characteristic by
detailed occupation, year and educational attainment among men and women of all ages
who reported working in that occupation in the week before the survey. These variables

are then matched to women in the expectations data by year, educational attainment, and

68

detailed occupation. Prior to 2003 the CPS used the same occupational codes used in the
Decennial Census. Table A1.1 provides a sample listing of occupations and codes along
with means of the occupational characteristics across years and educational levels. Mean

weekly wages are normalized to 1982-84 dollars using the CPI-U.

69

Table A1.1 Selected Occupational Categories and Mean Characteristics

A . . . . .

The number of observations listed includes women only. To calculate proportion of mothers in the
occupation, the percentage of female workers who are mothers of children under age 4 is first calculated
and then multiplied the estimated percentage of female workers in the occupation.

Notes: Not all occupational characteristics questions were fielded in all years. The observations reported
in the last column include all persons in the occupation; the observations reported in the last row include all
persons across all listed occupations with non-missing data for each characteristic. Because occupational
codes change across years the examples provided here include only women surveyed in 1983 through 1991
when consistent codes were used across years. The actual data include women surveyed in 1971 through
1992 and occupations are matched in the same survey year so no coding crosswalk across years was used.
Occupational characteristics are matched to expectations data by detailed occupation code, educational
attainment, and year. Since the May Supplement was not fielded in 1990 or 1992, the 1989 and 1991
surveys are used to generate occupational characteristics data to match with the June 1990 and 1992 data.

>033? .5 305.3 2:. .323

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

32m >. .. Maegan 08:02.05. 02m 2.8 mag .smm: Ormqmonm:m..om
303 Zea... 900.032: 3.03 7.3. $3.032: 0.33 .55 0325505
mcmmfiaaa 0.6.00. 03.0.
xx. $8.8 3 $9..» 2.3:
W8. Zoe: 0.... 02. Zn»: :5 pm: 036 02 .x. Im<m .x. .x. 2.09on
SQ E 5 .02 £8.» .23... «Snow 22255 355.0 a: 03.x. AA Zea. .Smm .Es.
.Smsmmoa. 03.9% mm. no.3 m.0o u. .N 0.0.3 No.0. No. ..0- .08 .30
2.3.8:: >a<oam§ Cooks 290d A. .58 Quad 8&3 $0.30 3908 Some
stmmoa. 2.8.0.30 mommom 9% from .o.om Mom .95 3.3 N... ~16 Pom. 03
8:. Egg 3.3.0.0 b8o0o bwmoo Oo.o0o God @9on $0.0loV A003
92:383. Marco. gumbo Sam no.3 .98 Pom Now 3.3 0.8 fimuo 930 903
0.8033 3 . 9000 0930 A. .98 $908 8.va CPod 690$ A. 0.me
moooaas mosoo. $0.3m 3.3 3.3 5.: P0. ~00 m~.mw won 93o obmw PM:
Humor»; 23.80 0“.me CNQNV 00.30 8.000 A303 $0.03 CNN:
woozxmlom m~N00o N..o0 098 No.3 P00 Goo 00.3 obu 8.00m .0.oom Shoo
Son... .0 6...me $.08 L303 80$ GALE A00: @908
”8833;; «SP... . wouo 00.3 3.0.. Poo ”.3 00.03 93 0.000 Po: Now“
22.30 Emu: A. .03 3.... .0 2.20 @008 2008 Good
Emaommom w.~o.m0 moow ~0.~m oo.uo P3 ...mo no.3 0.0. «05 0100.. $03
28.30 30.0. 20:30 3.3 2.08 3.30 00.03 35.23
000.8. 98.: moo: 5.3.8 30.. wabo Awuo 0.0m 0.0o meN woo 0.50 «Low ”room
02.9. 200.000 3903 c903 30.08 p..& $0.30 $0.00 A508
102. 095$... 29$ 0.: 8.0. «.00 Fou . SN 00.3 who ~08 Nouo .30
0955.: 3m... 3 60.80 A. .u .0 Swat: A. .30 Q . 000 3 . .30 20.30
Hoes...” maxim Emauo 3.: 30¢ 8.3 +00 M... 0...... Poo Abow Pug who.
Zuni—8 O 588 A. 36: 30.08 2:00 3..vo 80F Qwomo GHWM V8.03
mzoo 2.332.. wwwuow 3.8 00.8 .m.uo P0m o 0. .8. Mow A... ~0N .3
Omega; 80003 G . .000 G. .00 30$ 8.05 .o. buoy Cm.~00
ragga. $03. 0.3: m . mo. . m w . .oA 09.0 of... 93 0.0. a...» who ... .0 00. 3o
2.83:0 00m. 23.on $5.03 CPMMV $0.08 8.30 $0.30 $0.30 A730.
0.032233 00...... 0.5.: 3.08 3.08 marge mfg“. $.30 m..mm~>
mocqoam mg 38 .3 S29 7.8. 0:22: 00.3.2.0: 9.23. .03. .0 G I .03. mi. .00 I .00..

mop—38 3.. UmS :mao .: >=m.<m.mu :3. 9:42: veg—mac: mEéQ .00.. .000 I .03. .000. .02. .03. .08 I .03. BE .0m0 I .00..

 

.356 >..N 38:9 m.m:ama Um<.m=o:m. m2. MmSEm mnmm .2 08:02.03. O:mfimo.m1m..om.>:m_<m.m

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

7.23 A 2.03m N mxuoo. mxcoo. 8
mmmonfi 229. C3033: 803 3.; @010 08.00:: Im<o m2... 08630
.: 563 .528 9.3 o... o.m0 o... 9.0 o.00 o... o... odo
8.... 8.0. 8:00. 8.8 8.0... 8.0.. 8.... 8.3 8.00.
K. No.30 8.0800 8.0000 \.%....00 N02. 0000 00.0.00 5.0.0 8.0.0000 3.000
.020. $0.3. mmowbo $30..» $8.8. 3.0.0. 8...... 8.8.3 80.80 82... mum...
émmnm .: Ono. 2.8.... A..m.0.. A3888. 20...... 28.88. A. 3.8.. 28.0... 23.08. 8.90...
8.2.000 3.09.0 8.00.0 80.0800 30. 0000 5.20000 K0000 88.0000 8.8.00
><nqwmo 20:... .0. . 880 .0... 0.98. 80... w . .0. “8.0. 8.80 8.0.
.: 08. 29.6. A. .8». A. . . 8. 20.0. 2080. A. .88. A000. 20.03 A...ou.
890800 N0. 080 hug .0. 0800 00%....5 Fig.0 0...0 80.8000 U. 00%
.x. gear 00:- w..oo u..0o .06. .0... . . . . . 0o.o0 mm... ...o. 0...0.
flan .: 08. 8...... Awo. G. 8m.0o. 8.0.. 8.80. 8080. 890.. 8.0.. 80.0.0.
E... 0800 0. 0800 0.1.80 N. 0. 0800 000......0 5.0.30 \...%0 80.8000 N... 0000
.x. $08.6 02 5 ~0.m0 N...0 8o. 0.0.... 80.0. 80.... No.0. 8.80 8808
.u..... .: 08. A588. 2...... A.0.0.. 20.0. 20.2. A.m.wo. A.m.mo. 20.8.. A.0.N..
8.0.80.0 0.2.0.0 81.000 80.3.0 890000 8. 0000 \0000 80.00.00 8.880
>5. .0me 02 0.8 0.... 0... 0.0. 0.00 0.80 0.88 0.0 0...
£8: .: 08. 8.0... 8.00. 8.00. 8.0. 8.0... 8.0. 8.0. 8.00. 8....
3.00.0 50% 8.0000 8.18.. 5.0.030 8.0000 N030 8.0.0.80 8.00m
..\o :05 flax- :30 .m.00 .0... 3.0 .98.. 8.... .woo ...W0 .mo.
”.3... .: 08. 2.0... A3,... A300. 2.0.. 2.00. A803 :08... A...... 20....
x0688. @800? LEGF Rm. 0000! RM... 08.0 \0m.0 \0080 N0. 0.00 \0. 0800
.x. 02:03 .: am... .98 oo.o0 0.0. 0...0 00.... 0.... 08... .0...
08. 80.... 8.8.. 80.00. 88.0.. 8.00. 8.60.. 80.... 88.... 80:8.
8980 R000 #30? EQE 20.0.80 $0.300 U30 80.0.00 5.20.00
:0. 70.9.8; .: 0.0.. 0.00 0.0 .6. 0.0.. .0. ...o .00 ..0o
Ono. 8.0.. 8.00. 8.... A.. . N. 8.... 8.00. A0... A..0o. 860.
8.0.0 8.030 KMEQ \......0 83.0.... 50.800 .E ......0.0 0.0.0000

 

 

 

 

 

 

208” 0...0 $3.338 8.832. .: 00.0.0 .8 833:2. 0.2 88. £05.03: 8:80.03. $8.233. 33.8. .88.. mmo m0 335m? 2.40:. mmn min. 8:03

.32 mean 0.5.0.033: 38. damn 80039. 333 no :or

72

CHAPTER 2

Job Continuity Among Expecting and New Mothers

73

Introduction

Mother’s employment decisions around birth influence lifecycle earnings and
gender based economic equity. Job changes and labor force exits likely contribrite to
differences in wages between mothers and childless women (the family wage gap) and
between men and women (the gender wage gap). Mothers may have lower wages than
men or women who never have children because their life cycle labor supply is
punctuated with more absences (Ben-Porath 1967). Women who quit their jobs for any
reason lose firm-specific human capital. If they exit the labor force for an extended
period of time their skills may deteriorate (human capital deterioration). In order to
maintain employment women may seek out jobs with flexible working time policies and
other family friendly characteristics, and they may need to trade off wages, advancement
potential, or both to get them.

Mother’s employment behavior may also influence child development. A
substantial body of empirical evidence suggests a positive relationship between
household income and child outcomes. A mother’s ability to maintain paid employment
over the course of her pregnancy and while her child is young will affect the level of
household income. However, maternal employment reduces the amount of her time
available to care for her new baby, which may in turn have negative effects on child
development. For example, Waldfogel et al.(2002) and Baum (2003) find children whose
mothers worked in the first year after they were born have lower standardized test scores
through preschool and elementary school.

From the employer’s point of view quits can be costly. If flexible working time

policies and practices such as paid and unpaid leave improve job continuity and lead to a

74

net reduction in costs, then offering such policies would be efficient. Yet if the offer of
flexible working time policies leads women with a high demand for flexibility to seek
employment with the firm (adverse selection) then the costs of flexible policies and
practices may outweigh the benefits. Findings in Chapter 1 suggested women who
expect to have children are no more likely than those who do not expect to have children
to select occupations with family friendly characteristics early in their careers. However,
selecting out of less family friendly jobs and into more family friendly jobs may occur
closer to the time of birth or within occupations.

This paper addresses the following two questions: First, how prevalent are quits
prior to and just after having a child? Second, are women less likely to quit jobs which
provide paid or unpaid leave or a part-time work schedule? Of women employed one
year before pregnancy (or 21 months before birth), my estimates suggest 57.4 percent
remain in the same job 18 months after the birth. Of the remaining 42.6 percent,
approximately one-third leave their pre-pregnancy job for a new job and two-thirds leave
the labor force. Findings suggest women who are eligible for FMLA leave and those
whose pre-pregnancy jobs were part-time are less likely to change jobs between the
twenty-one months preceding and eighteen months following a birth. Those eligible for
FMLA leave are less likely to quit for any reason prior to the birth. Although women
who change jobs are more likely to obtain paid leave on the new job than to lose it, I do
not find any evidence of higher quit rates among women without paid leave.
Conceptual Framework

Job continuity, in this study, is defined as maintaining the same job with the same

employer across a specified interval of time. The decision to stay in one’s job (maintain

75

job continuity), change jobs, or leave the labor force can be conceptualized as follows.
An individual chooses between non-employment and a set of job opportunities; which are
characterized by combinations of wages, hours, and job attributes to maximize utility.
U=U(C,l,A;x) (1)
C is the quality adjusted level of consumption of both market and home produced goods
and services, which is determined by income and time spent in home production as well
as productivity in home production. I is the amount of leisure and x is a vector of taste
shifters. A is the level of amenities associated with the utility maximizing choice of job
or non-employment. Amenities for workers include job attributes J and other non-
monetary benefits of work such as socialization and self-efficacy. Amenities for non-
workers include utility from being at home, for example, value of greater flexibility and
autonomy over time and activities. Increased quantity or quality of home produced goods
when not working or when working a reduced schedule are captured in the level of C
associated with that choice. Utility is maximized subject to budget and time constraints:
Th+Y2w(T—h)+Ek (2)
T = h +1 + k (3)
Y is income, w is the wage and Ek is the monetary expenditure on child related goods and
services. T is the total time available for market work h, leisure l and home production of
goods and services including child care k.
I assume the value of a unit of leisure is constant across job and employment
choices and home production technology exhibits increasing returns to scale, but does not

differ across jobs and employment status. Thus, for a given individual, the differences in

the levels of k and 1 between jobs and when employed and not employed will be

76

determined by h alone. Time allocation across individuals will of course depend on the
value of leisure and relative productivity in home production.
Given these assumptions, the individual’s gains to changing jobs V(Chg) can be

expressed as a fimction of the hours, wages, and job attributes associated with the current
and alternative jobs, full income I, and the mobility cost Mchg associated with job change

(Altonji and Paxson 1988). The individual’s gains to leaving the labor force V(Lv) can be

expressed as a function of the hours, wage and job attributes of the current job, amenities
associated with non-employment A h and full income. Labor force exits are assumed to

carry no mobility costs.

V(Chg) = U(hl,wl,Jl,I)-U(h0,W0-,.J091)-Mchg

(2)
V(Lv) = U(A,,,1) — U(h0,w0,J0,I)

Mothers will choose change jobs if V (Chg) is greater than zero and greater than V (Lv).
Similarly, mothers will choose to leave the labor force if V(Lv) is greater than zero and
V(Lv) is greater than V(Chg).

Based on this framework, 1 choose to focus on quits (job changes and labor force
exits) as opposed to re-entry into the labor force or duration of leave as many previous
studies have done. Focusing on quits is advantageous because quits are invariant to the
interval studied so long as the quit itself is observed in the data. Conversely, a simple re-
entry measure would code women who begin a job with little or now employment
interruption the same as women who return to the labor force in a new job months later
the same way. Since pervious studies have shown women who maintain employment
continuity do not experience a motherhood wage penalty (Waldfogel 1997; Fuchs 1995),

such a measure is undesirable. Furthermore, most women who return to work at all in the

77

first year after having a child do so within the first three months (Klerman and Leibowitz
1990, 1994, 1999; Smith and Bachu 1999). Arguably the decision to maintain
employment or one’s current job with a brief leave will have a greater impact on long run
employment outcomes than the decision to take a one month versus three month leave.
Previous Studies

Although, there is a vast literature on the employment decisions of new mothers,
few studies have considered job continuity in particular. Still, since the decision to stay
in the same job, change jobs, or leave the labor force entirely is closely related to the
timing and duration of leaves or labor force separations, the broad return to work
literature is relevant to the present analysis. Return to work studies differ with respect to
the employment behavior studied and the sample design. Table 2.] summarizes key
characteristics of a subset of studies in the broader literature that were selected to
demonstrate these differences. As I will explain, given these differences it is difficult to
infer overall conclusions about women’s employment behavior, or job continuity in
particular.

In the existing literature, employment behaviors studied include employment at a
point in time, leave duration in general, instance and duration of unpaid leave in
particular, occupation changes, job changes, work schedule changes (for example, part-
time to full-time), time of re-entry into the labor force, labor force participation, whether
or not a woman returned to the same job, and whether or not she experienced any
employment interruption. The employment behavior considered, at least to a certain
extent, determines the sample design. For example, Bumpass and Sweet (1980) are

interested in employment. They measure employment at specific points in time, rather

78

than transitions in and out of employment, and can therefore include non-employed
women in their sample. Conversely, since both Desai and Waite (1991) and Han and
Waldfogel (2003) examine leave duration, they exclude women who are not employed
and must designate a time from which to measure leave.

These details are especially important for studying job continuity. Choice of time
from which to measure behavior, which I will refer to as the “initial state”, determines
which women, jobs, employment, and non-employment spells are in the sample. The last
job or employment state (employed, unemployed, out of the labor force) observed during
pregnancy or the year prior to birth are common initial state choices (e.g. Han and
Waldfogel 2003; Joshi and Hinde 1993; McRae 1993; Waldfogel et al.1999). By
defining the initial state close to the birth, these studies may miss early job changes or
transitions in and out of employment. Omitted transitions will clearly affect estimated
job or employment transition rates.

Given enough information, studies with different sample designs can be
comparable. For example Glass and Riley (1997) define the initial state as all women
working during pregnancy and estimate 84.2 percent of women who gave birth in 1991
and 1992 and worked during pregnancy were employed one year following the birth.
Waldfogel (1997) estimates employment one year after the birth among a similar sample
of women, but includes non-employed women. She finds 54 percent of all women who
gave birth were employed one year following the birth. 63 percent of women in
Waldfogel’s sample were working during pregnancy. Using this figure to convert Glass
and Riley’s estimate to the percentage of all women working one year after birth yields

53 percent (84.2 x 0.63 = 53), which is nearly identical to Waldfogel’s estimate. This

79

example is a best case scenario. The initial state definitions were clear, and comparable
estimates could be constructed from available information. Estimates from studies which
use more complex or ambiguous initial state definitions are harder to interpret and
compare.

Finally, sample design may also differ with respect to the start and end times of
the interval over which employment behavior is observed. Whereas the definition of the
initial state designates which women or jobs are included in the sample and which
previous transitions are excluded, the definition of the interval determines which
subsequent transitions are measured. The beginning of the interval need not be the same
as the time at which the initial state is measured, although it often is (e. g. Han and
Waldfogel 2003; Wenk and Garrett 1992). Even if two studies use the same definition
for the initial state, they may arrive at drastically different estimates of transitions in
employment or job changes if they measure transitions over different intervals.
Differences in interval studied generally produce sets of estimates that are more easily
comparable than estimates produced by studies with different initial state definitions,
because they represent different regions of the same distribution of transitions (assuming
the sample populations are otherwise comparable).

Excluding potentially relevant job changes before the birth is not only
problematic because the behavior is omitted but also because it may lead to endogenous
regressors. For example, in a comparative study of women in the U.S., Britain, and
Japan, Waldfogel et al. (1999) consider the effect of family leave coverage on the
likelihood of returning to the pre-birth job within six months of the birth, and they find

family leave coverage is positively associated with returns to work in all three countries.

80

Yet women may have changed jobs to obtain leave coverage before the initial state was
measured (six months prior to the birth). If so, family leave is endogenous in their
model.

The extent to which women are selecting into jobs with family leave coverage or
paid leave is unknown, in part because very few studies differentiate between returns to
the pre-birth employer and job changes. Klerman and Leibowitz (1999) estimate about
60 percent of women return to work in the year and a half following a birth and of those,
approximately one-third have changed jobs. However, their sample is drawn from
women in the 1979 National Longitudinal Survey of Youth, which covered births in the
19805. Using the 1969 NLS covering births during the 19705, Glass (1988) finds 53.5
percent of women who became pregnant during survey waves left the labor force and
12.3 percent changed jobs as compared to 16.4 percent exiting the labor force and 26.5
changing jobs when there was no family transition (pregnancy or marriage).

The findings in both Klerman and Leibowitz (1999) and Glass (1988), and in
most of the studies in Table 1.2 are all based on data from the 19705 and 19805 and thus
their findings may not generalize to more recent employment behavior. Between 1975
and 1996, the labor force participation rate among all women rose from 45.9 percent to
58.8 percent; and among women with children under age six the rate rose from 38.8
percent to 62.3 percent (Hayghe 1997). This striking change in labor force participation
among mothers with young children suggests employment behavior may have been
changing during the months preceding and following a birth as well.

The subset of return to work studies which have estimated relationships between

employment behavior and availability of part-time work and paid and unpaid leave have

81

found women with paid leave are likely to work longer into their pregnancy and to return
to work later (Joesch 1997), and the availability of liberal unpaid leave and part-time
work increases the likelihood of returning to the pre-birth employer (Hofferth 1996). For
FMLA, Han and Waldfogel (2003) and Waldfogel (1999) find limited impact on the
instance of unpaid leaves, and no evidence of a net change in women’s employment.
Empirical Strategy .

As stated previously, contrary to most studies in the return to work literature, the
dependent variables in my analysis are binary indicators of job changes and labor force
exits. These measures follow directly from the conceptual framework laid out in the
previous section. They also have practical advantages given common limitations across
relevant data sets. The timing of quits is generally easier to assess than the timing of re-
entry. Most surveys do not distinguish between time spent on leave and time spent at
work. Therefore, many respondents do not report a gap in employment around
childbearing or a measured state of non-work to re-enter from (Klerman and Leibowitz
1994). If there is no gap in behavior observed in the data, then the designation of a re-
entry time is arbitrary and the term “re-entry” itself is somewhat misleading given what
can feasibly be measured. Even if the measure used in the analysis is a discrete indicator
of transition behavior rather than duration of leave or time to quit measure, timing must
be clearly defined in order to appropriately designate the initial state and determine which
transitions fall within the analysis interval.

Although the conceptual framework I have laid out generally leads to multinomial probit
or logit estimation, it can also be operationalized in a competing hazard model. Since

quitting one’s initial job to leave the labor force precludes a quit to change jobs and vice

82

versa, quit behavior can be modeled in a competing hazard framework where the hazard
of each type of quit can estimated separately using Cox proportional hazard estimation
and defining jobs that end in labor force exits as censored in the job change estimation
and vice versa (Prentice et al 1978; Moeschberger 1978; Cox 1959). As will be
explained in the following section, the complex structure of the data make hazard
estimation more appropriate than multinomial probit.

To create the dependent variable, I define the initial state as employment in a job
that existed at a given previous point in time and in addition to I choose the specific point
in time based on patterns in the data which will be presented in the next section. In doing

so, I allow each woman who is employed at the time the initial state is defined to

contribute at most one observation.9 This sample design allows me to compare
differences in quit rates and relationships between each type of quit, paid and unpaid
leave and part-time work schedule in a well defined set of jobs at different points in time.
To make such comparisons, I estimate piecewise proportional hazard models allowing the
estimated effects of paid and unpaid leave, part-time work schedule and other covariates
pertaining to the job to differ before and after the birth.
Data

The analysis sample is drawn from the 1996 through 2005 Household
Components of the Medical Expenditure Panel Surveys (MEPS). The MEPS sample is a
sub-sample of households participating in the National Health Interview Survey (N HIS).
The NHIS is a nationally representative sample of the U.S. civilian non-institutionalized

population and includes over-samples of blacks and Hispanics. The MEPS also over

 

9 Multiple job holders contribute only the current main job (according to the Current Population Survey
definition). The incidence of multiple job holding was quite low (less than 1% of jobs in any initial state
definition considered).

83

samples Asians and low income households. The MEPS consists of five rounds of data
collection over a two year period. The Household Component contains socio—economic
data including job characteristics of the Cm'rent Main Job (analogous to the Current
Population Survey definition) and documents other “miscellaneous” jobs held. Start and
end dates for all jobs are recorded.

Initially, the sample includes all jobs for all women who had a birth during the
survey, have a child age 2 or younger, or have their first prenatal exam during the survey.
Not all of these observations are used in all parts of the analysis; again the intial state and
interval will determine which jobs are included and excluded.

For any jobs that end during the survey, a reason is recorded. I exclude jobs that
ended due to dissolution of the business, retirement or layoff. Only those quits for which
the respondent indicated she quit “to take another job” were coded as job changes. Quits
the respondent identified as due to “illness or injury”, “to have a baby”, “to take care of
home/family” or “because wanted time off” were all coded as leaving the labor force.
Illness or injury quits are included because they likely relate to pregnancy related
conditions. The ability to maintain employment through a complicated pregnancy is
likely to depend in part on flexible working time policies and practices, especially paid
_ leave. However, these observations constitute fewer than 5 percent of all quits (including
dissolution of businesses, retirements and layoffs) and quits to have a baby, to take care
of home or family, or to change jobs comprise 86 percent of all quits.

The MEPS is a relatively short panel. Respondents are surveyed five times over

the course of two years. Since births happen at all points during the survey, the quit data

is constructed of overlapping panels with time measured relative to the birth. Job tenure

84

information is used to backcast employment behavior. For example, women need not be
observed in the panel at 21 months prior to the birth in order to contribute an observation
to the analysis sample when the initial state is the job held 21 months prior to birth.
However, women who enter the sample at six months after the birth would only
contribute observations to the 21 month analysis sample if they were still in their 21
month jobs. Women who quit early and do not enter the sample until six months after the
birth would be excluded from the sample, that is, they are fully left censored. As is
generally the case, little can be done to recover information about those short spells but
given the overlapping panel structure of the data, I have no reason to believe the spells I
do observe have a different hazards than the ones I do not. Late entrants are treated as
left-truncated spells and their survival rates are adjusted for time at risk but not observed
in the data. '0
Results

Figure 2.1 displays the estimated monthly hazard rate of changing jobs (the
broken line) and combined hazard of changing jobs or leaving the labor force (the solid
line) among jobs held 21 months prior to the birth over the interval from 21 months
before to 1.5 years after the birth. The distance between the two lines represents the
hazard of leaving the labor force. The job change hazard is bi-modally distributed with
the first peak around one year prior to the birth. Job changes become increasingly rare
throughout the pregnancy and the hazard rate does not reach pre-birth levels again until
1.5 years following the birth. Conversely, the hazard of leaving the labor force increases

drastically just before the third trimester of pregnancy and peaks at birth. There is a

 

'0 Specifically, I use the stset command in Stata and specify the “origin” as the time defined as the initial
state and the “entry” as the time at which an individual enters the survey.

85

slight peak in the combined hazard around the start of the pregnancy which is due to a
higher rate of job changing one year prior to the pregnancy than during the pregnancy
itself and a higher rate of labor force exits at the beginning of the pregnancy than in the
second trimester.

Table 2.2 compares estimates of job continuity across different sets of jobs and
intervals to estimates provided in previous studies. Overall, I find 57.4 percent of jobs
held at 21 months prior to the birth continued on through 18 months after the birth. Of
the remaining 42.6 percent, approximately two-thirds left the labor force and one—third
changed jobs. The first row in Table 2.2 compares my estimates to Klerman and
Leibowitz’s (1999) estimates of job continuity for jobs held twelve months prior to the
birth over the year before and 1.5 years following the birth among women who had a
child between 1978 and 1990. Taken at face value, these estimates indicate women who
gave birth more recently are more likely to stay with their pre-birth employer but, among
those who quit their jobs, fewer change jobs. Comparing my estimates of job continuity
among jobs held at birth from birth to 1 year after to Desai and Waite’s (1991) estimates,
which were based on births between 1979 and 1985, also suggests the likelihood of
remaining with the pre-birth employer has increased over time.

The differences in job changing and labor force exit estimates may be due, in part,
different definitions of job changes. Klerman and Leibowitz (1999) code any women
who are working in a different job 1.5 years after the birth as having changed jobs, even
if they transitioned through non-employment first. My estimates code quits as job
changes only if the respondent indicated she was quitting to start a new job. She may or

may not be employed at the 1.5 year mark. Also, a respondent who said she was quitting

86

her job to leave the labor force and was coded as such in my data may have returned to a
new job by 1.5 years after the birth. In Klerman and Leibowitz’s data, that individual
would be coded as a job changer. Unfortunately, because my data are constructed of
short overlapping panels, it is not possible to construct a more comparable measure.
Thus, from this comparison one may infer job continuity has increased over time but
inferring any change in the distribution of quits between job changes and labor force exits
is less appropriate.

The differences in hazard rates in the months preceding and following the birth
shown in Figure 2.1 and the difference in my own and others estimates of job continuity
across sample designs in Table 2.3 highlight the importance of sample design. Table 2.4
provides a comparison to make this point even more clear. Consider a simple data set
containing all jobs held 21 months prior to the birth without the overlapping panel
structure and truncation in the actual data I use. If we choose to examine quits from 21
months prior to the birth through 18 months following, we would infer 57.4 percent of
women kept their jobs, 13.6 percent changed jobs and 29 percent left the labor force over
the interval. If instead we considered 21 month jobs from 12 months before the birth to
18 months after, we would omit 5.9 percent (4.6 + 1.3) of 21 month jobs because they
had already ended in a quit.

Note that in Table 2.3, the first and last row of figures seem to suggest we miss
little by looking at jobs from 12 months to 1.5 years versus 21 months to 1.5 years.
Indeed the estimated percentage of job changes is 12.8 percent in the shorter interval and
13.6 percent in the longer interval. However, this comparison is not based on a well

defined population of jobs. The 12 month to 1.5 year estimates are based on all jobs held

87

12 months prior to the birth whereas the 21 month to 1.5 year estimates are based on all
jobs held 21 months prior to the birth.

Returning to Table 2.4 and looking at 21 month jobs only, the estimated
percentage of women staying in their initial job over the 21 to 18 month interval is 57.4
percent as compared to 61.0 percent over the 12 to 18 month interval. Moreover,
approximately one-third of all job changes are missed when the interval is shortened from
21 months before to 12 months before. Looking only at the interval from birth to 18
months after, we would omit over 25 percent of 21 month jobs which constitute 52
percent of all job changes and 65 percent of all labor force exits among 21 month jobs
between 21 months before and 18 months after the birth..

Clearly these omissions change the estimates of job continuity based on
remaining jobs in each interval but as long as the intervals are reported and the number of
jobs remaining is reported, these differences are interpretable and the only source of
misclassification is due to truncated spells. However, if instead of analyzing the
remaining 21 month jobs only at each interval we were to examine the number of total
jobs which exist at the beginning of each interval, misclassification becomes a problem.
Of the original 1,922 women employed at 21 months prior to the birth 88 have new jobs
by the year prior to the birth and these jobs are now at risk for a quit. If the relevant quit
was the first quit these women will be misclassified. Similarly, at birth, approximately

7.1 percent (or about 136) of the original 1,922 women have new jobs and could end up
in any of the three categories by the 18‘h month. Those who previously exited the labor

force may also be re-employed by the beginning of the next interval, although since the

intervals considered here all begin at or before birth, re-entry is unlikely.

88

For these reasons, I define the initial state for the remainder of my analysis as jobs
held 21 months prior to the birth. Since one of the main purposes of this study is to
examine the relationships between paid leave and job changes, and since a large amount
of job changing occurs prior to the pregnancy, defining the initial state as jobs held 21
months prior to the birth is more appropriate than jobs held closer to the birth. In
particular, it captures the initial peak in job changing, but does not extend so far prior to
the pregnancy that it contains employment behavior that is more likely to be unrelated to
the birth.

Table 2.5 presents estimated availability of paid leave, means of other job
attributes, and socio-demographic characteristics of remaining 21 month jobs at various
intervals. The changes in means over time suggest women are more likely to stay in jobs
that offer paid leave or in which they are eligible for FMLA leave. However, they appear
to be less likely to stay in part-time jobs.

Table 2.6 contains the results of Cox proportional hazard estimation for each type
of quit (job change vs. labor force exit) reported as hazard ratios. The point estimates for
paid leave are less than 1 and thus imply that women with paid sick leave or paid
vacation leave are less likely to change jobs or leave the labor force than those without
but neither reaches statistical significance. F MLA eligibility, however, reduces the risk
of job changing by 37 percent. The risk of exiting the labor force does not appear to
differ among women who are eligible and those who are ineligible for F MLA leave.

Having a part-time work schedule differentially predicts job changes and labor
force exits. Women who work part-time have a 52 percent lower risk of changing jobs

but a 28.7 percent higher risk of leaving the labor force. Among women in general, not

89

just pregnant women and new mothers, labor force attachment tends to be lower among
part-time workers than among full-time workers (Blank 1989).

Interestingly, wages and income do not appear to have much impact on job
changing or labor force exits. Point estimates for wages imply women are less likely to
quit higher wage jobs for either reason. Point estimates for wages imply a 1% increase in
wages is related to a 20 percent reduction in the risk of changing jobs and a 10 percent
reduction in the risk of leaving the labor force but only the first estimate is statistically
significant and only at the 10 percent level. However, women who work in salaried jobs
have a 45 percent lower risk of leaving the labor force and that estimated effect is
significant at the 5 percent level. Point estimates for household income are very close to
1, meaning there is no difference in job changing or labor force exit likelihoods by
income after other factors are controlled for.

Table 2.7 presents the estimates of piece-wise proportional hazard estimation,
which allows the estimated relationships between each type of quit and paid and unpaid
leave and part-time work schedule to differ before and after the birth. Paid leave still
does not appear to have any affect on quits either before or after the birth. However, the
estimated effect of F MLA leave in Table 2.6 appears to be entirely primarily due to a
reduction in job changing behavior prior to the birth. Prior to the birth, women who are
eligible for FMLA have a 52 percent lower risk of changing jobs relative to those who are
ineligible whereas after the birth the estimated difference is only 21 percent and is not
statistically significant. Similarly, the risk of changing jobs among part-time workers is
over 80 percent lower than among full-time workers before the birth but there is no

significant difference in behavior after the birth. Interestingly, the entire positive

90

relationship between part-time work and labor force exits shown in Table 2.6 is due to
labor force exits prior to the birth. After the birth, part-time workers are no more likely
to exit the labor force than full-time workers. This finding is interesting because, as
stated previously, part-time workers in general have lower labor force attachment than
full-time workers. Blank (1989) finds, among all female workers, those initially working
part-time were approximately twice as likely to be out of the labor force when observed
again two years later. Thus, part-time may have no impact on labor force exits after the
birth because there are benefits to working part-time at that point in the lifecycle that lead
to a higher level of job continuity than is typical in that segment of the labor force.

Throughout the analysis I have assumed women change jobs around the time of a
birth to achieve more flexibility. Thus far, findings suggest FMLA eligible women and
those with part-time work schedules are less likely to change jobs prior to the birth. Table
2.7 further investigates the claim that women change jobs to obtain flexibility by
comparing the availability of paid and unpaid leave, part-time work and other job
attributes in old jobs and new jobs among all women who change jobs. Due to the short
panel structure of the data and missing data for some of the covariates among newjobs,
only 273 matched old and new job pairs could be identified.

Although having access to paid vacation or paid sick leave did not appear to
influence job changes (or labor force exits) in the competing hazard analysis, women who
changed jobs were more likely to obtain access to paid leave than to lose it. 19.19
percent of all job changers (or 37.14 percent of those who did not have it initially) gained
paid vacation leave on the new job. Figures for sick leave also imply a higher likelihood

of obtaining than losing sick leave but are not as pronounced as for vacation leave.

91

Although the hazard analysis suggested F MLA leave eligibility and part-time
work schedule deterred job changes, the tabulations in Table 2.7 seem to indicate those
who do change jobs are not seeking out FMLA covered employers or part-time work
schedules. Only 10.26 percent of all job changers entered a part-time job whereas 16.48
percent left part-time jobs. Additionally, there is virtually no difference in the proportion
of old and new jobs in which the employer is covered by F MLA. However, in the piece-
wise proportional hazard estimation, it became clear that the relationships between
FMLA eligibility, part-time work, and job changing was strongest prior to the birth. The
figures in Table 2.7 include all job changes between 21 months prior and 18 months after
the birth.

Although the sample sizes get quite small, when only oldjobs changes that
occurred prior to the birth are considered, 13.24 percent of job changer entered a part-
time job and 14.71 percent left a part-time job. Considering that only 37 percent of all
jobs held by women in the MEPS between 21 months prior to the birth and 18 months
and only 33 percent of those that began prior to the birth were part-time jobs, the roughly
equal transition rates in and out of part-time work among job changers are notable. Still,
this evidence is based on only 68 matched old and new job pairs.

Similar findings hold for transitions in and out of jobs with FMLA covered
employers. F MLA leave is likely to be most valuable at the time of the birth but
eligibility requires one year of tenure. Thus if selection into jobs where the employer is
covered by F MLA occurs it should occur between 20 and 12 months prior to the birth.
Comparing job changes in that six month window to those that occur after it, only 24

percent of jobs that were left during the six month window (old jobs) were covered by

FMLA as compared to 48 percent after the window. Furthermore, 28 percent of new jobs
that started during the six month window were covered by FMLA as compared to only 17
percent of those started after the window.

Discussion and Conclusion

Employment decisions associated with childbearing influence lifecycle earnings,
may affect child development, and may contribute to the gender and family wage gaps.
This paper analyzed the patterns in job changes and labor force exits among expecting
and new mothers and assessed the possible effects of paid and unpaid leave and part-time
work on these decisions. I contribute estimates of job continuity using a new and more
recent nationally representative sample of births, and examine not only transitions from
employment to non-employment but also job changes (job-to-job turnover).

When compared with previous estimates of job continuity, my estimates suggest
women who gave birth in more recent years have been more likely to stay in their pre-
birth jobs than those who had children in the 19705 and 19803. Comparisons of the
difference in the distribution of quits between job changes and labor force exits are more
tenuous, because my definition of job changes differs from previous studies and
construction of a more comparable measure was not feasible. However, taken at face
value, the estimates imply fewer women change jobs around childbearing in 1996. to 2005
than in the 19705 and 19803. Estimates for labor force exits, however, are very similar.

This paper also highlights the importance of sample design in the return to work
literature, and studies of women’s employment behavior around childbearing in general.
The interval over which employment behavior is evaluated and the designation of the

initial state from which employment transitions are measured vary greatly across studies.

93

My findings suggest there is relevant job changing behavior occurring as early as 21
months prior to the birth (1 year before the pregnancy). Studies that examine job
continuity among jobs that exist closer to the birth or at the time of the birth miss these
changes.

Furthermore, estimated relationships between job characteristics, and paid and
unpaid leave availability in particular, differ before and after the birth. For example,
early job changes are significantly less likely among FMLA eligible women when quits
before and after birth are analyzed together. But when analyzed separately, the overall
relationship is due to a large negative relationship between FMLA eligibility and job
changing before the birth; there is no significant relationship after the birth. The same is
true of part-time work.

Overall, I find no relationship between access to paid leave (either vacation or
sick leave) and job changing or labor force exits. However, I do find women who change
jobs are more likely to gain access to paid leave than lose it. Admittedly, they moved
into higher wage jobs that were more likely to offer health insurance and retirement plans
as well. In short, the possibility that women changed jobs for other reasons and happened

to obtain paid leave cannot be ruled out.

94

Figures and Tables

Figure 2.1 Monthly Hazard Rates for Job Changes and Labor Force Exits for
Jobs Held 1 Year Before Pre .nancy SI(21 Months Before Birth)
1'? ~ ’ '51“ EflifiiFiff If I. at!" 3: .4: £9 £1”!

    
 

    

 

 

Hazard of Quitting to Leave the Labor Force or Change Jobs
' 5 y ----- Hazard of Quitting to Change Jobs

is? .., . . ‘4‘; .1. Jr -. ‘i 3".“ V; W “$713; w l

 
 

 

95

406.0 N... 060.00.062.00 2 >:0.<0.0 m030.00 .: 0 m0.00..o: 2 0.05000 .0203 .0 <<o§ 02. 20.200 200.00

 

 

 

 

 

 

 

 

 

 

1002. 50.0 .000 <02... .: .w060<.0.. 52.3.20: on .60 .220. 920 m8... ....30 6.0.. 0.30 ..0..
066003020: 0.6.0.. . . 60.. >00.w.0.0 >00.v.0.0 0..
62.2203. 60.0%. .2260 30.0600 €030: ”0.0.0 .6... 30.00.00 0.. 0.. 506020.. 56010..
O 0 00 2. 2. 21. £01.03 3.20. .06 0..
020.. .0 .0026... 030.830... 000..
.0030000 22.30. m0..<0v. 0.. .000 I .00.. m30.8.302 <00 20 w2.0..0 N 0.00.0 010.. .0
02. m€02 003.... 03.2.6 303.0000 .00. 3.02.06
.000 Al0m0 .6. 6......
0000. 0:6 26m< .000 .000 I .006 00.2.0: 00 .1000 a: 20 .00. .06 0600200 . 0.000 N «0000 0:2
€020 .00. .20. 02. $03 .60 62.00 .60 6....6. 62.00
0.0....
:0: 02. m.00 .00. I .000 30.0600 02. 00.2.0: 0.. 20 .06 60.6 0 30260 .0 30260 0 30260 0.0.
20.26.00. .06 03.00.06 C600... 62.20 66.6 02. 62.00 6....6. 66.6.
308. .t00<0. 30.20.:06 .630m6 us
30260 0.400.
.00006 meo .000 I .06.. 00.2.0: 0.. 6000.00 20 >. 00000060: .02. 0 30260 0.00. 060020..
.00.. 300. 60<0 $0060.. .: 62.20 66.6. 0.02.... 60.
.60 ..06 0. 0.0338. 00.. 060022.30 0.0
2 .000. 6 30260.. 00:00.2.
60.0.00: 6
30260 0:6 0
N020 00.0. 66.6.
.006. 0:6 ZmIU .006 I .06. m30.8.30:. 02. .: 030.8.302 .100. ..06 60...:m w....6 .6 0.0000 0:00
1360 .000 02. .06. I 000002.30. O60:m0 30.0.0.0. v.00. .: 0003.38. .60 66.6.
.00.. 000002.30. 000002.30. 30.5.0
060:W0 30.0.0.0. 030..
:0.
.2230: zrm< .000. .0:0 .00.. I .000 08:02.30 603.00: <00 20 .~ 30260 .0 30260 06.0..
0:6 00m .060 02. 60.000: 62.0.0 6....6. 6....6.
62600.3 m30.8.302 02. 20:-
2000. 030.8.302 60. 00...-
..30 0:6 00..-..30
m.0.00
2.0.00 22.30.... .00.. I .00.. ”0-028.. .0023. 2 20 .00. .06 60... 600.:m w....6 N0 200.. 0:00
.00. 0000002270 30.. 2..-..30 0002038. 6....6.

 

0028.. 26.0.

 

 

 

 

 

 

 

 

96

._.06.0 N. 060.00.062.00 0. >:0_K0.0 m030_00 .: 0 m0_00..0: 0. .u..0<.000 .0203 .0 <<02 0:0 20.200 200.00 8023000.

 

 

 

 

0600.3 0:0 Zrm< .060 .060 I .000 .0600 0080 <00 20 .0 30260 .N 30260 00.0..
7.0.. .000 02.0.0000: 62.08 6...... 6....6.
20.00000. 2rm< .000. .000 I .000 20.0.: .0 0030 .06 20 P00. .06 00.00 .0 6....6. 9.3 .N 30260 2.2
0. 0.. 2000. 2000. 000 2.0.. 66.6 00.. c0.
.000 I .006 0:0 Ck. NA
8.0.. 30260 2.2
.000 COP05 6....6 00.. .0000.
200.. 0:0 Zrm< .000 .000 I .006 >8. 030.8.302 20 00:82.0: 0 30260 . 0.00.. 00.0..
00.8.. 3.03.0000: 0:0 3:0. 60008 66.6. 66.6.
.00~ 030.8302 0.200.

 

 

 

 

 

 

 

 

 

000” 00.82 0000.000: 0.020..
2000. 22.0:0. 06:0 0 08 003.0...

20.00” 22.0:0. 0....0 00830302 0.00.... . .20..

2605 22.0:0. 60:0..0030. 0028. 0.. <00... I .400. 00 30.00.00

206.0. 22.02.. 0023 00 003.... 032.6
20:0. 20:02.. 00200. 00 200:6 0:0 008.00302 .C..A..

0mm” 00:0. 003.0% 0: 0030303 2000:.

0.00. 0028. 0.. .:0030 0:0 0.0083 02.0.0000:

. 7.00. 020 00.0 00 :0. 0:0... .60 8000860.. .0 05082.20 60.200: £030: £60 08 030.8.00 0:0 0: 30.03.... .008 6.03 .6000 :60 08 030.8.00 0:0 0.
20.... 0.608008 .008 .0 00:000.... 02.300 00 0: 0:020 0600:00 #03 20.... 00 000.00 0.. :0:..030.8.302. 000 2.0.3.0: 0:0 60.60.20 3003 00.. 0 00.0..00

0.000003: 0.. .60 0:008:00 620.00: 0.0.... 0:0 030.8302 0.0.00 00:80.30 06:06.36.

97

32m NM 0033230: 8 @8305 mmaamfimm 0* gov man m3u_o<3m3 003352

 

3,285 mEEnm

Zw mmzamfim

Amem warm :23 I NoemV

 

 

 

 

 

 

 

203 .85 B 5833 63.38
41 l,[l It!

,I

 

 

S 783:; 629.0 8 E 39:3. ~59, 8:32am m8 85533 $2.: 81328 98:32. 5 +8.0 _ 3.5035: 25 55025 :83 no<m1=m E33

 

 

 

$50 .86 05.22: .86 OS 0% En $58 83 035mg ram 583
have”. 1080 86mm 108%
5 78.35 wnwoao 8 3 78:25 >32 Awmfi. 20.x. safe mnbfo _N.me\c 39X.
Cog :05 5 3833 8298
was 8 5 78:25 >mmn «when. 98.x. Sufi.
:8 $5 a was $9.. 2.x.
2 78:25 ammo?” 8 a 32:3 >mo~ Em 3.3.x. 58.x. No.9,}

 

 

:03 Sum 8 38. was 8 5 52:3 ~52. own—swam who 3me o: 08”: SE 238 Coo: no<oazm 233 was 3.3 8 .23.

N $032. :88 8an mm orwsmmsmbocm on; R 90% 598ng 33‘ run as: 8 035mm .83 2:. mm Emiam 9o :82 38a 0:? 1.28% :5?»an 93
2:581? Km :8: How Amma >253? >~ mom “38:8 anwnlvzo: 3. 9:3. .2035: BE romvoifi 05308le gov armammzm 8 335 m E2083 gov
E :w 3025 ~32 En En: 3m: 2 5 39:2; ~32 :8 E33. mo”: 523m: 25 r1833 :83 Ba Ummmm 85 (5:8 :00: 3328 :89. moan»
33.88 mm woman 9: 3. En Econ 38¢ a 30 28 83:0 318 cm 38238:. Zw maniac—a Saran s83»: Eco o<2 an :5: 55m: 86 8.. m :92
.86 2 i8 3:: 312.8 8 _nw<a :5 .32 38¢ 98" again 35 ER 5va mama—Sam $9: <81» m: m 83:55 23:33?" 852.0323 «ammarwmm
03:0: £25 3 :8 ma ow :5 38238: union. 308 833820 Eommcam <33 :2 «3&2» m2»: :5 m8: 251 38088 2.26 38.

98

Table 2.3 Potential Omission and Misclassification at Different Intervals for 21
Month Jobs

 

I 21 to 18 Months I 12 to 18 Months T Birth to 18 Months

 

Distribution among Total 2] Month Jobs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Stayed 57.4% 57.4% 57.4%
Changed Jobs 13.6% 9.0% 6.5%
Left Labor Force 29.0% 27.7% 10.2%
Previous Job Changes n/a 4.6% 7.1%
Previous Labor Force n/a 1.3% 18.8%
Exits

Total Jobs l,922 1,922 1,922
Distribution among Remaining Jobs 2] Month Jobs at Each Interval

Stayed 57.4% 61.0% 77.5%
Changed Jobs 13.6% 9.6% 8.8%
Left the Labor Force 29.0% 29.4% 13.8%
Remaining Jobs 1,922 1,809 1,424
Distribution among All Jobs that Exist at the Start of the Interval

Stayed 57.4% 54.9% 71.8%
Changed Jobs 13.6% 12.8% 9.4%
Lefi Labor Force 29.0% 32.3% 18.9%
Total Jobsl 1,922 2,233 2,303

 

Notes: The estimates in the top and bottom panels are derived from the actual data (with sample weights

applied). The middle panel of estimates is calculated from the top panel for illustrative purposes.
However, estimates in the actual data approximate these figures out to the third decimal place.

Changes in the total number of jobs are due in part to the overlapping panel structure of the data. They

do no reflect actual changes in the number of jobs held at each point among a constant number of

individuals.

99

 

Table 2.4 Descriptive Statistics by Interval for Jobs 21 Months Before (Standard

Errors in Parentheses)

 

A11 21 Month Jobs

Remaining at Birth

Remaining at 18
Months After

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Paid Sick 0.62 0.70 0.73
Leave (0.02) (0.01) (0.01)
Paid Vacation 0.74 0.81 0.80
(0.02) (0.01) (0.01)
FMLA 0.37 0.43 0.44
Eligible (0.01) (0.01) (0.01)
Part Time 0.26 0.23 0.21
(0.01) (0.01) (0.01)
Wage $13.34 $14.69 $15.59
(0.26) (0.25) (0.25)
Salaried 0.31 0.36 0.40
(0.01) (0.01) (0.01)
Retirement 0.51 0.60 0.61
Plan (0.02) (0.01) (0.01)
Union 0.12 0.14 0.14
Member (0.01) (0.01) (0.01)
Offered Health 0.81 0.83 0.83
Ins. (0.01) (0.01) (0.01)
Private 0.84 0.81 0.79
Employer (0.01) (0.01) (0.01)
Federal 0.02 0.03 0.03
Government (0.00) (0.00) (0.00)
State 0.07 0.08 0.09
Government (0.01) (0.01) (0.01)
Local 0.08 0.08 0.09
Government (0.01) (0.01) (0.01)
Management 0.12 0.13 0.13
(0.01) (0.01) (0.01)
Professional 0.24 0.27 0.30
(0.01) (0.01) (0.01)
Service 0.17 0.13 0.13
(0.01) (0.01) (0.01)
Sales 0.10 0.10 0.09
(0.01) (0.01) (0.01)
Clerical 0.19 0.19 0.21
(0.01) (0.01) (0.01)
Production 0.07 0.08 0.08
(0.01) (0.01) (0.01)
Other 0. 12 0.1 1 0.06
Occupation (0.01) (0.01) (0.01)
Total Income $63,627.38 $66,985.15 $69,945.20
(1,405.26) (1,197.26) (1,132.85)
MSA 0.84 0.82 0.82
(0.01) (0.01) (0.01)
Age 30.24 31.17 31.59
(0.19) (0.16) (0.14)
White 0.78 0.78 0.78
(0.01) (0.01) (0.01)
Black 0.16 0.16 0.15
(0.01) (0.01) (0.01)

 

 

 

 

100

 

Table 2.4 Descriptive Statistics by Interval for Jobs 21 Months Before

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Continued)
Other Race 0.05 0.06 0.07
(0.01) (0.01) (0.01)
Hispanic 0.24 0.20 0.21
(0.01) (0.01) (0.01)
Married 0.75 0.79 0.80
(0.01) (0.01) 40.01)
Less than High 0.12 0.09 0.09
School (0.01) (0.01) (0.01)
GED 0.04 0.04 0.04
(0.01) (0.01) (0.00)
High School 0.45 0.44 0.43
(0.02) (0.02) (0.01)
Bachelors 0.21 0.24 0.25
(0.01) (0.01) (0.01)
Graduate 0.09 0.10 0.1 1
Degree (0.01) (0.01) (0.01)
Other Degree 0.09 0.10 0.09
(0.01) (0.01) (0.01)
Grandmother 0.08 0.05 0.06
in House (0.01) (0.01) (0.01)
Non English 0.18 0.18 0.20
Speaking (0.01) (0.01) (0.01)
Household
Number of 1.48 1.56 1.60
Children (0.04) (0.04) (0.03)
Number of 0.91 1.02 1.05
Kids < 6yrs (0.03) (0.03) (0.02)
Tenure in 4.02 4.42 4.69
Years (0.09) (0. 10) (0. 10)
Observations 1,922 1,600 1,527

 

Notes: Figures are estimated population parameters and standard errors are reported in parentheses.

101

 

Table 2.5 Results 21 Month Jobs at 21 Months before to 18 Months After

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Changed Jobs Lefi Labor Force
Paid Sick Leave 0.763 0.921
(0.208) (0.144)
Paid Vacation 0.743 0.888
(0.191) (0.131)
FMLA Eligible 0.631M 0.867
(0.141) (0.120)
Part Time 0478* 1287+
(0.1 19) (0.184)
ln(Wage) 0800+ 0.917
(0.107) (0.090)
Salaried 0.592 0.550“
(0.215) (0.126)
Tenure in Years 0.925+ 0.951“
(0.037) (0.018)
Retirement Plan 0.895 0.863
(0.217) (0.123)
Union Member 0.793 0.955
(0.326) (0.191)
Offered Health Ins. 0.908 0.856
(0.207) (0.1 12)
Federal Government 0.643 0.554
(0.472) (0.255)
State Government 0.637 1.081
(0.290) (0.245)
Local Government 0.734 0.988
(0.355) (0.242)
Management 0.860 0.630+
(0.322) (0.176)
Professional 0.618 0628* *
(0.227) (0.139)
Sales 1.317 0.913
(0.401) (0.199)
Clerical 0.744 0.751
(0.236) (0.148)
Production 1 .042 0.938
(0.420) (0.232)
Other Occupation 1.176 2.81 1*
(0.382) (0.466)
Full Income (10 1.056 0.956+
Thousands) (0.064) (0.022)
Full Income2 (10 0997 1'00]
Thousands) (0.002) (0.000)
MSA 1.466 1.380“
(0.353) (0212)
Age 0921* 1.011
(0.021) (0.012)
Black 0.81 1 0.834
(0.212) (0.146)
Other Race 0.916 0.766
(0.346) (0.200)
Hispanic 0.503M 1.364M
(0.144) (0.196)

 

 

 

102

 

Table 2.5 Results 20 Month Jobs at 20 Months before to 18 Months After

 

 

 

 

 

 

 

 

 

 

 

(Continued)
Married 0.786 1.209
(0.184) (0.186)
Less than High 0.768 0.930
School (0.240) (0.161)
GED 1.416 0.845
(0.513) 10.232)
Bachelors 1.507 0.973
(0.495) (0.180)
Graduate Degree 4.145“ 1.108
(1.919) (0.318)
Other Degree 1.481 1.118
(0.533) 40.231)
Grandmother in 0.664 0.743
House (0.220) (0.156)
Non English 0.724 0625*
Speaking Household (0.209) (0.102)
Number of Children 1.088 1.034
(0.125) (0.070L
Number of Kids < 0.832 0.953
6yrs (0.136) (0.091)
Observations 1 922 1922

 

 

 

 

Notes: Figures in table are hazard ratios.

103

 

Table 2.6 Changes in Estimated Relationships Before and After Birth (Jobs Held

21 Months Before)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Changed Jobs Lefi Labor Force
Before Birth (BB) 0.643 0.042*
(0.589) (0.019)
Afier Birth (AB) 0,246+ 0.01 1*
(0.205) (0.006)
Paid Sick Leave BB 0.905 0.732
(0.400) (0.174)
Paid Sick Leave AB 0.740 0.985
(0.258) (0.208)
Paid Vacation BB 0.615 0.947
(0.249) (0.201)
Paid Vacation AB 0.829 1.041
(0.280) (0.227)
FMLA BB 0481+ 0.702
(0.184) (0.156)
FMLA AB 0.791 0.903
(0.225) (0.163)
Part-Time BB 0176* 1.472+
(0.072) (0.302)
Part-Time AB 0.851 1.041
(0.264) (0.207)
ln(wage) BB 0.821 0.912
(0.171) (0.103)
ln(wage) AB 0752+ 0.944
(0.123) (0.147)
Salaried BB 0367+ 0567+
(0.202) (0.165)
Salaried AB 0.707 0.655
(0.300) (0.186)
Tenure BB 0717* 0.970
(0.078) (0.029)
Tenure AB 0.981 0.949“
(0.042) (0.023)
Retirement Plan BB 0480+ 0680+
(0.209) (0.157)
Retirement Plan AB 1.151 1.030
(0.345) (0.192)
Union Member BB 1.023 0.926
(0.854) (0.321)
Union Member AB 0.693 0.984
(0.330) (0.243)
Offer Health Insurance BB 1.173 0.648“
(0.475) (0.118)
Offer Health Insurance AB 0.704 1.021
(0.198) (0.195)
Federal Govt BB 0.619 0.922
(0.662) (0.668)
Federal Govt AB 0.462 0344+
(0.477) (0.204)
State Govt BB 0.000 1.175
(0.001) (0.415)

 

104

 

Table 2.6 Changes in Estimated Relationships Before and After Birth (Jobs Held
21 Months Before) (Continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

State Govt AB 1.082 1.031
(0.516) (0.310)
Local Govt BB 1.165 1,961+
(1.025) (0.731)
Local Govt AB 0.755 0.742
(0.444) (0.247)
Management BB 0.525 0.889
(0.358) (0.306)
Management AB 0.988 0.373"
(0.456) (0.174)
Professional BB 0.803 0.705
(0.428) (0.215)
Professional AB 0449+ 0.638
(0.215) (0.210)
Sales BB 1.553 1.046
(0.713) (0.282)
Sales AB 1.236 0.708
(0.51 1) (0.268)
Clerical BB 0.836 0.772
(0.406) (0.202)
Clerical AB 0.705 0.774
(0.294) (0.237)
Production BB 1.189 0.923
(0.795) (0.327)
Production AB 0.979 0.964
(0.493) (0.341)
Other Occupation BB 1.51 1 1.010
(0.706) (0.261)
Other Occupation AB 0.967 5.761 *
(0.434) (1.437)
Observations 2,758 obs. for 1,922 individuals 2,758 obs. for 1,922 individuals

 

 

Notes: Figures in table are estimated hazard rations. Estimates also control for the full set of demographic
and other control variables as in Table 2.5.

105

Table 2.7 Characteristics of New and Old Jobs Among Job Changers

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Means Old Means New % Gained % Lost % of % of All
Jobs Jobs Without Had Lost
Gained

Sick Pay 0.39 0.46 17.71% 9.59% 28.74% 25.00%
(0.49) (0.50)

Paid Vacation 0.49 0.57 19.19% 10.33% 37.14% 21.37%
(0.50) (0.50)

FMLA Covered 0.46 0.47 17.34% 15.73% 31.39% 35.14%

Empmyerl (0.50) (0.50)

Part Time 0.30 0.24 10.26% 16.48% 14.74% 54.22%
(0.46) (0.43)

Wage $9.10 $10.48 64.58% 24.72% n/a
(5.16) (7.32)

Salaried 0.12 0.15 6.96% 3.30% 7.88% 28.13%
(0.33) (0.36)

Retirement Plan 0.21 0.30 18.08% 8.86% 23.00% 41.38%
(0.41) (0.46)

In Union 0.05 0.06 4.40% 2.93% 4.60% 66.67%
(0.21) (0.24)

Offered 0.76 0.80 15.75% 13.55% 70.49% 17.45%

Insurance (0.43) (0.40)

Federal 0.01 0.01 1 . 10% 0.37% 1.1 1% 50.00%

Government (0.09) (0.12)

State 0.02 0.01 0.3 7% 1 . 10% 0.3 7% 75.00%

Government (0.13) (0.09)

Local 0.04 0.06 3.30% 2.20% 3.45% 50.00%

Government (0.21) (0.23)

Management 0.07 0.07 2.93% 2.56% 3.14% 38.89%
(0.25) (0.23)

Professional 0.12 0.14 4.40% 2.93% 5.00% 24.24%
(0.33) (0.34)

Service 0.26 0.25 10.99% 1 1.36% 14.78% 44.29%
(0.44) (0.44)

Sales 0.21 0.16 10.99% 8.68% 55.56% 1 1.36%
(0.40) (0.36)

Clerical 0.16 0.20 6.67% 13.42% 45.24% 5.49%
(0.36) (0.40)

Production 0.10 0.1 1 5.49% 4.76% 6.15% 44.83%
(0.29) (0.32)

Other 0.09 0.07 0.73% 3.30% 0.81% 33.33%

Occupation (0.29) (0.26)

Observations 273 273 273 273 273 273

 

 

 

Notes: Standard deviations are reported in parentheses. Data include all quits that ended in job changes
among jobs held at any point within 21 months before the birth and 1.5 years following the birth and could
be matched with the subsequent job.

1
Measured as any employer with more than 50 employees

106

Chapter 3

Making Time for Well-Baby Care:
The Effect of Maternal Employment and Paid and Unpaid Leaves

107

Introduction

The American Academy of Pediatrics recommends children receive preventive
care visits or “well-baby visits” at regular intervals throughout the first twenty-four
months of life. Well-baby visits are intended to create and maintain a health history and
deliver anticipatory guidance for parents. They also provide an opportunity for health
screenings and vaccinations and include developmental and physical assessments.

Anticipatory guidance, health screening and vaccinations can play key roles in
reducing infant mortality. The mortality rate between birth and age 1 year is nearly 23
times higher than the rate for 1 to 4 year olds and only 25 percent lower than that of 55 to
64 year olds (Hoyert et a1. 2005). Although the risk of death is highest at the time of
birth and during the first few hours and days of life, 10 percent of all infant deaths in
2003 (or 34 percent of those which occurred after the perinatal period and were not
associated with congenital abnormalities) were due to Sudden Infant Death Syndrome
(SIDS), accidents or unintentional injury (Hoyert et al. 2003). Prevention through
anticipatory guidance of SIDS, accidents and injuries is a goal of well-baby care. Since
proper sleep position guidance was first publicized in 1992, the percentage of infants
placed on their backs to sleep increased from 13 percent to 72.8 percent and the rate of
Sudden Infant Death Syndrome (SIDS) fell by over 50 percent (National Institute of
Child Health and Human Development 2008). Although parents could have received
sleep position guidance from a variety of sources, well-baby care is specifically aimed at
delivering this sort of information.

Early diagnosis of existing conditions can improve prognosis. From birth to ages

3 to 5 years, visual acuity develops from below 20/200 to near 20/20, however, early

108

conditions such as strabismus (commonly known as “lazy eye”) can preclude
development of visual acuity (Daw 1998). The critical treatment period for strabismus
and other sources of limited visual acuity (amblyomia) is thought to be between birth and
age 2 years (Daw 1998). Children who receive early screening (before age 3) and
treatment for amblyomia have better long run visual acuity than children who are
screened and diagnosed at age 3 (Williams et a1. 2002). Similarly infants, who are found
to have congenital hearing loss in early infant hearing screening, have a higher likelihood
of developing speech (Kaye 2006). Screenings can occur in other medical settings but
well-baby care visits are designed to include them.

8.9 percent of 2003 deaths which occurred after the perinatal period and were not
associated with congenital abnormalities were the result of infectious diseases and
endocrine, nutritional or metabolic diseases (Hoyert et a1. 2003). Vaccines are available
for many infectious diseases and for some endocrine, nutritional and metabolic diseases,
treatment can be as simple as administering vitamin supplements and maintaining routine
follow-up care (Kaye 2006). The schedule of well-baby care visits includes screenings
for many of these conditions. For those that cannot be detected in medical tests, routine
physical examinations and developmental assessment included in all well-baby visits may
lead to earlier diagnosis.

Preventive care is recommended to continue throughout childhood, adolescence
and on into adulthood. However 11 of the 31 visits recommended between birth and age
twenty-one are to occur before age two. Given the higher mortality risk and the
concentration of health screenings and vaccinations during those early years, the marginal

benefit of preventative care is arguably highest for well-baby visits. Despite the

109

importance of these visits, most studies have found babies receive significantly less than
the recommended amount of care (Ronsaville and Hakim 2000; Byrd eta1.1999; Maisels
and Kring 1997). Yet others find compliance rates are quite high (around 80 percent)
(Selden 2006).

This paper attempts to reconcile differences in previous estimates of compliance
and examine the relationship between maternal employment, availability of employer
provided leaves and part-time work schedule, and receipt of well-baby care. As with any
form of healthcare, cost and access are important determinants of demand. Yet, federal
and state policies have made well-baby care virtually free for all publicly insured infants.
Furthermore, cost sharing under private insurance is also very low. If compliance rates
are indeed quite low despite the low cost of care then other factors, such as time
constraints, may be especially important.

One-third of mothers return to work during the first three months of their baby’s
life (Klerman and Leibowitz 1990, 1994, 1999; Smith and Bachu 1999). For these
mothers, well baby care must either be scheduled around work hours, or time off work
must be obtained. Therefore, paid and unpaid leave and part-time work may help to
enable employed mothers to take their babies to the doctor.

Background and Previous Studies

Currently, the American Academy of Pediatrics recommends a total of ten well-
baby visits before a child’s second birthday. These include a newborn evaluation, which
generally occurs in the hospital following delivery, and a visit three to five days after
birth to evaluate breast feeding behavior and to check for jaundice (Recommendations for

Preventive Pediatric Health Care 2008). Subsequent visits are timed to deliver on-time

110

vaccination, administer health screenings and deliver age apprOpriate anticipatory
guidance to parents

Previous studies examine compliance with AAP guidelines but produce
seemingly contradictory estimates. The discrepancies are likely due to the measure of
compliance used, differences in the sample population, or both. For example, Ronsaville
and Hakim (2000) find 58 percent of white, 35 percent of black and 37 percent of
Hispanic six month olds in the National Maternal and Infant Health Survey (NMIHS) are
“fully compliant” with recommendations. However, they count any doctor visit where an
immunization occurred as a well baby visit. They also code children who receive at least
three of the four visits recommended in the first six months as fully compliant. Using the
Medical Expenditure Panel Survey (MEPS), Seldon (2006) defines well-baby visits to
include general check-ups and immunizations or shots and measures compliance as the
percentage of age-specific recommended visits received. He concludes 83.2 percent of
infants are in compliance with care recommendations. Maisels and Kring (1997) use a
sample of 20 private pediatricians’ offices and examine compliance with the 2 to 4 day
follow-up visit recommended for infants who are discharged from the hospital less than
48 hours after birth. They find only 38 percent of early discharge infants received a
follow-up visit within four days. Bryd et a1. (1999) measure the percent of recommended
care intervals in which at least one visit occurred. Recommended care intervals are
defined as the range of ages during which a specific visit should occur. For example, the
2 month visit should occur in the interval between ages 2 and 3 months, inclusive. They
find privately only 49 percent of privately insured and 36 percent of publicly insured

infants received well-baby care in all recommended care intervals.

11]

The Maisels and Kring (1997) and Byrd et al. (1999) compliance measures are
affected by the timing of visits. Arguably the simplest measures would count the number
of visits a child received. However, timing of visits is important for (at least) three
reasons. First, by definition, anticipatory guidance must anticipate the need for
information. Second, as discussed previously, on-time screening for health conditions
can significantly improve prognosis among the diagnosed. Third, on-time immunizations
minimize the risk of viral infection.

The childhood immunization schedule corresponds to the levels of maternal
antibodies present in the infant’s body. At birth, maternal antibodies not only protect a
newborn from disease, but also prevent successful vaccination. These antibodies decline
over the first few months of life and thus the risk of infection increases until endogenous
antibodies reach a preventive threshold. Ideally, vaccination should occur as soon as the
maternal antibodies decline enough to permit successful vaccination. The immunization
schedule targets that window of time. Although subsequent doses of any given vaccine
are identical to previous doses (there is no additive effect), multiple doses increase the
likelihood of successful immunization. Late follow-up doses may lead to longer periods
of susceptibility to infection if early doses do not successfully establish immunity.

Studies that have used a strict definition of well-baby care and excluded other
types of doctor visits (e. g. Byrd et al 1999) may underestimate the among of care
received if infants are receiving some well—baby care during other doctor visits. For
example, a baby could obtain missed vaccinations during treatment for an ear infection.
Similarly, his mother could receive the anticipatory guidance that she would have

received in a well-baby visit. If parents often substitute incidental sick visits for regular

112

well-baby care, measures that exclude sick visits will underestimate the true amount of
care received.

Few previous studies attempt to determine why well-baby care is under-utilized.
Of those, only Vistnes and Hamilton (1995) specify the demand for medical care as a
function of prices and income. They find the number of visits a baby receives is
decreasing in hours worked per week and weeks worked per year and mothers with sick
leave to visit the doctor are no more likely to take their children to well-baby visits.
Estimated coefficients on out of pocket cost and time cost are negative but only weakly
significant and imply small own and cross price effects.

Berger et al. (2005) estimate the relationships between mothers’ early return to
work (within the first twelve weeks), initiation of well-baby care, and number of visits
over the first year, along with other infant health investments and outcome measures.
Their OLS results indicate infants whose mothers return to work early are 2.4 percentage
points less likely to receive any well-baby care in the first year of life. Intensive margin
results are weakly significant but imply infants whose mothers return to work within the
first twelve weeks receive 0.20 fewer visits. Propensity score estimates are very similar
to OLS results. However, they do not include any measure of out of pocket cost of care
in their model. Also, although, Berger et al. (2005) find negative relationships between
leave duration and well-baby care use, they estimate the relationship between leave
behavior and receipt of care. They do not examine the relationship between leave
availability and receipt of care and thus their estimates likely overstate the relationship

between leave and well-baby care use.

113

Other studies have found non-black, non-poor and privately insured infants,
infants with more educated parents (Yu et al. 2002; Mustin et al. 1994), infants from
families with fewer children, infants from households without transportation problems
(Moore and Hepworth 1994), infants who receive home health visits (Braveman et a1.
1996), and infants whose parents were sent visit reminders (Campbell et al. 1994)
received more care than other similar infants. Furthermore, McInemy et al. (2005) find
state level insurance reimbursement rates are significantly and positively related to
compliance with well-baby care recommendations. Kost et al. (1998) and Marsiglio and
Mott 1988 find no relationship between pregnancy “wantedness” and receipt of well baby
care. Kviz et al. (1985) find no relationship between well-baby care receipt and mother’s
health beliefs, defined as beliefs regarding susceptibility to illness, severity of risks,
benefits of care, and efficacy of care. Again, either price, income or both are missing
from the regressors in these studies.

In a related study, Colle and Grossman (1978) estimate a demand function for
pediatric care, including whether or not a child received any well-child visits, among
children ages 1 through 5. They exclude infants from their analysis because they argue
the benefits of care in infancy likely differ greatly from the benefits of care among older
children and thus the determinants of compliance also may differ by age. They base their
analytical framework on a production fiinction for child health where well-child care is
one of many inputs. From the production function they obtain a derived demand function
for well-child care that becomes their empirical model.

Colle and Grossman find income increases the likelihood of receiving a

preventive care visit, but the effect is smaller at higher income levels. This finding is not

114

surprising because the out of pocket cost of visits for most insured children is quite low.
Price effects are not separately estimated for preventive care visits, but they do find an
own price elasticity of -0.106 for total number of office based visits received.
Additionally, Medicaid recipients receive fewer visits but they find a positive relationship
between welfare receipt and receipt of preventive care. Time costs reduce the likelihood
of receiving preventive care and the estimated effect is largest for infants whose mothers
work. The potential endogeneity of maternal employment is not addressed.

In total, most existing studies estimate compliance rates or describe the
characteristics of infants who are likely to receive care. Of those that estimate demand
functions for well-baby care or related preventive care, both find evidence of competition
between maternal employment and receipt of care (Colle and Grossman 1978; Berger et
al. 2005). Yet, Colle and Grossman consider preventive care visits among older children
only and their data is from 1971. Berger (2005) use data from the NLSY 1979 covering
all births between 1988 and 1996 to women in the NLSY cohort who were ages 23
through 39 at that time. Since they focus on the relationship between receipt of care and
mother’s return to work, they restrict their analysis to women who were employed three
months prior to the birth.

I contribute nationally representative estimates of well-.baby care compliance
using a sample of babies born between 1994 and 2005. Furthermore, I estimate
relationships between availability of paid and unpaid leave and work schedules among
employed women and relationships between insurance coverage, price of care and
estimated time costs for all women (employed and non-employed).

Conceptual Framework

115

Following Colle and Grossman (1978) I assume household utility is a function of
the baby’s health B, mother’s and father’s leisure [m and If, and consumption of all other
goods X.

VI=W(B,L,X) (1)
Further, baby’s health is produced using well-baby visits V and full-income less
expenditure on well-baby care M ' and is affected by an exogenously determined

individual specific health endowment bi.

B=7<V.M‘,b.-) (2)
Derived utility can then be written as a function of well-baby visits, leisure, other
consumption and infant health endowment:
U = U(V,L,X,b,-)
The household chooses the number of well baby visits, amount of leisure and

consumption of other goods to maximize utility subject to the following budget and time

constraints and comer restrictions:

PX+WWL+W“V—W*S+my=WWT+Y=AI (»
T=V+L+H—S (Q

V20 L20
(5)

H20 .920

P is a vector of prices corresponding to the consumption goods vector X, W* is the vector
of maternal and paternal shadow prices of time, which for employed parents should be
equal to their wage, S is hours of paid time off available to each parent, Y is non-labor
income, M is full income, T is the total time allocated to all activities and H is hours spent

in market work.

116

To simplify the optimization problem and subsequent analysis I assume the
mother is solely responsible for taking the baby to the doctor and father’s employment
and income are exogenous. Mothers accompany children under age 16 to 91 percent of
doctor visits (Vistnes and Hamilton 1995). Since this proportion is likely to be even
higher for well-baby visits, this assumption is likely to be reasonably accurate. Even in
the case of a stay-at-home dad, the mother will likely need to be present for well-baby
care because evaluations of breast feeding and mother-infant interaction are key
objectives of care. With this simplification the optimization problem can be written as
follows:

aria} U(v,l,X,b,-) s.t.PxX+w*(1+v——S)+va=M
Va ,

T=v+l+h—s (6)
v20;120;h20;320

v is now the total number of visits received and w*, l, and s are now mother’s shadow
price of time, her time spent in leisure and the days of paid leave she has available.

Optimization yields the following conditions:

9::
ap,

<0

(7)

That is, the demand for visits should be decreasing in own price, increasing in paid time
off and increasing in full income. For working women, w* must equal the wage in
equilibrium (assuming no hours constraints or fixed costs of work) and thus the effect of
w* on the demand for visits will depend on the size of income and substitution effects

and the relative marginal values of time spent in leisure and well-baby care. Since the

117

out-of—pocket cost of well-baby care is $0 for the majority of privately and publicly
insured infants, the sign of the cross wage elasticity for working women is likely to be
negative (Acton 1973). For non-working women, w* is the value of time spent in leisure,
which must equal the reservation wage (Coffey 1983). Since there is no income effect
for non-working women, the relationship between w* and Vwill again depend on
whether or not well-baby visits and leisure are complements or substitutes.
Empirical Strategy

Based on the foregoing conceptual framework, the baseline empirical

specification of the demand function is as follows:

Vij- = ,30 + fllpvij + ,62'w* +,B3M,-j + Edy-A] + Kidsij-Az + ZijA3 + 5,-1- (8)
The dependent variable Vij is equal to 1 if infant i received at least one well-baby visit
during the jth care interval. pvij is the out of pocket cost , Mi]- is the household’s full
income, Ed is a vector of maternal education measures, Kidsij are variables summarizing

the number and age of children and Z0 is a vector of access to care, quality of care,

subjective and objective infant health endowment variables, and maternal and infant
demographic variables. w* is the shadow price of time and is equal to the wage for
workers and the reservation wage for non-workers, which is unobserved. In practice,
most authors either omit any measure of the value of time all together or use predicted
wage offers in lieu of w* for individuals who are not working. I discuss the interpretation
of parameters under each strategy below.

When any direct measure of the shadow price of time is omitted from the model,
maternal education and number of children are often regarded as proxies, however, their

predicted effects are ambiguous. Wage offers should be increasing in education and the

118

opportunity cost of time must be higher than the wage for women who are not working.
Thus education should be positively correlated with the shadow price of time and
negatively related to the receipt of well-baby care. Yet, more educated mothers may also
be more productive producers of infant health. Infants with more educated mothers may
receive a higher marginal benefit from well-baby care (productive efficiency) because
their mothers understand recommendations and communicate more effectively with
healthcare providers. Or more educated mothers may select a better mix of health inputs
including well-baby care (allocative efficiency) (Grossman 2003). If either the
productive or allocative efficiency hypotheses are correct then the overall relationship
between education and well-baby care could be positive.

Studies that use predicted wage as the value of time for non-workers implicitly
assume there are no inherent differences between workers and non-workers and their
wages are drawn from the same distribution as working women (Gronau 1973).
Furthermore, by definition, the predicted wage must be lower than the reservation wage.
The assumption of similarities between workers and non-workers is required even if
reservation wages are derived from estimates of female labor supply (e. g. Gronau 1973).
To address this problem, Coffey (1983) uses a survey that directly solicits estimates of
the reservation wage among respondents who are not currently working. She finds
estimates of time-price elasticities for female health care based on the direct survey
measure of reservation wages and on selection corrected predicted wages (Heckman full
information maximum likelihood FIML) are identical (Coffey 1983; 1980). Since there
is no direct measure of reservation wages equivalent to Coffey’s measure available in the

data, I test the sensitivity of my results to various time price measures including past

119

observed wages, predicted wages and selection corrected predicted wages. My preferred
measure uses a combination of past observed wages and predicted wages.

In addition to missing shadow price of time, out of pocket cost data is missing for
visits that do not occur. For infants who receive at least one well—baby visit, I use the
average out of pocket cost of previous visits in place of missing price data. For infants
who do not receive any well-baby visits I predict out of pocket cost using the following
equation:

Pv =50+61Age+1UAl +HiiA2+RijA3+e (10)

Since the distribution ofout of pocket costs is highly skewed to the right, and over 50
percent of visits have an out pocket cost of $0, I exclude observations with out of pocket
costs in the top 1 percent of the distribution when predicting missing price data. Also,
from 1997 forward I replace missing price data for Medicaid recipients with $0 because
cost sharing was eliminated under Medicaid/SCHIP in 1996.

To assess the relationship between maternal employment, paid leave and well-
baby visits I modify Equation 8 as follows and use all observations for working and non-

working women in the estimation:

VI] = p0 +plpv,-j +£22.14?” +p3l~1','j* Eli +p4EIi +p5Mij +EdiPl + KIdSijPZ

. ‘ Il
+PLUP3+JUP4+ZIIP5+H ( )

EU- is equal to 1 for employed mothers and {if}. is equal to the observed wage for

employed mothers and the estimated shadow price of time for non-working mothers. PL ,1-

is a vector of paid leave policies available to the mother. This specification allows for
comparisons among infants with non-employed mothers, employed mothers without paid

leave, and employed mothers with paid leave.

120

The decision to work is likely to be endogenous. To address this issue, I restrict
the sample to married women only and use father’s current wage, whether he is salaried
or hourly, and his hours of work (entered linearly and as a quadratic) as instruments for
mother’s employment status. If mothers are exclusively responsible for taking their
babies to well-baby visits then, after income is controlled for, father’s wages should not
have an effect on the demand for visits except through income. Admittedly, the validity
of this exclusion restriction rests entirely on the assumption that mothers are solely
responsible for taking the baby to the doctor. No more appealing identification strategy
exists in the previous literature although Berger et al. (2005) find OLS estimates of the
relationship between mother’s return to work are generally consistent with estimates
derived through propensity score matching.

To examine the effects of paid leave and other job characteristics among
infants with employed mothers I estimate the following equation only for recommended
care intervals when the mother is employed.

Vij = b0 +b1pvij +b2Wij +b3My' +EdijBI +KldSy'Bz +leB3

(12)
+PLijB4 ‘1' J1} BS “I'll

PLU is a vector of paid leave policies available to the mother and Jg- is a vector of
other job characteristics. If women who invest more in their infants’ health choose to
work in jobs that accommodate doctor’s visits then job characteristics will be endogenous
in this specification. Previous studies use past job characteristics as a proxy for current
job characteristics under the assumption that they are more plausibly exogenous (e. g.

J acknowitz 2004).

121

Unfortunately because the MEPS is only a two year panel, there are too few
recommended care intervals for infants whose mothers are observed prior to the birth to
use lagged job characteristics. I can, however, estimate the above model on a restricted
sample of infants whose mothers have worked in their current job for more than 1 year
before pregnancy. In Chapter 2, I found most job changes prior to birth occur between 1
year before the pregnancy through early pregnancy and there are very few job changes
from the second trimester through six months following the birth. This information, along
with the fact that most women who do acquire or lose paid leave do so by changing jobs,
suggests my restriction is similar to using lagged job characteristics.

Data

The analysis sample is drawn from the 1996 through 2005 Household
Component, Event Files, and Conditions Files of the Medical Expenditure Panel Surveys
(MEPS), which is the same data used in Chapter 2. The Household Component contains
socio-economic data including information about the Current Main Job (analogous to the
Current Population Survey definition), detailed current health status and background
information, monthly insurance coverage status and characteristics of respondents’ usual
care providers. The Event Files consist of records for each unique household-reported
medical event (e.g. office-based visit, emergency roomvisit, home health visit, outpatient
treatment). Although the majority of well-baby visits occur in an office or clinic setting,
the Emergency Room Visits File did contain some well-baby care records aside from
those at the time of birth. These visits are included in the analysis. Both the Office
Based Visit and Emergency Room Visit records contain the date of the visit, the broad

type of care received (e. g. well-baby care, diagnosis or treatment, emergency accident or

122

injury), total cost of the visit by source of payment and ICD9 condition and procedure
codes.

The expenditure data on the public use file are derived from both the Household
and Medical Provider Components. The Medical Provider Component surveys the health
care providers the household identifies in the Household Component. Data from medical
providers supplement missing household data.

Respondents identify visits as well-baby care when asked the main purpose of the
visit. In estimating compliance I consider a strict measure using only visits for which the
main purpose reported was well-baby care, a broader measure which includes visits
identified as “general check-up”, “immunization or shots”, and “maternal care
(prenatal/postnatal)”. I also consider the number of all office based visits (including sick
visits) received.

To construct the analysis sample, I use all infants age 24 months and younger and
determine the AAP recommended care intervals according to their dates of birth. Again,
recommended care intervals are the age ranges in which a given well-baby visit is
supposed to occur. Using the date of each well-baby visit on the event file, 1 match visits
to the recommended care intervals. Throughout the analysis, the dependent variable is an
indicator equal to 1 when at least one well-baby visit occurs during a recommended care _
interval.

Figure 3.1 is a histogram of all visits received by infants in the analysis sample by
their age in months at the visit. The peaks in the distribution of visits across age indicate

the‘timing of visits corresponds closely with the AAP recommended schedule (see Table

2 for the visit schedule). When visits are more than one month apart, the bins to the right

123

of each age at which the visit is to occur contain more visits the bins to the left. Thus, I
define the recommended care interval for each visit to include the month in which the
visit is recommended and subsequent months until the month of the next recommended
visit. For example, in the 9 month recommended care interval the dependent variable
would be equal to 1 for babies who received a well-baby visit at ages 9, 10 or 11 months.
The only exception to this coding rule is the 1 month visit; the dependent variable is
coded as l for infants who receive a well-baby visit before they reach 1 month of age
even if they do not receive another visit at age 1 month.

Covariates are matched to recommended care intervals based on the interview
round in which the recommended care interval falls. In cases where a recommended care
interval overlaps two survey rounds, information from the round in which the visit
occurred is used for infants who received the scheduled visit and information from the
round which contains the most days of the recommended care interval is used for those
who do not.

In total, the sample consists of 26,5 1 3 recommended care intervals for 7,251
infants. This sample is used to generate estimates of compliance with well-baby care
recommendations. The sample used to analyze the determinants of compliance is limited
to the subset of infants who have a usual care provider and received at least one office
based visit for any reason within the past year. I limit the sample in this way because out
of pocket costs must be estimated for all infants who do not receive any visits. Out of the
7,251 total infants in the sample 3,675 (50.7 percent) received no well-baby visits while
they were in the survey. Of those, 50.9 percent received at least one office-based visit for

which I have out of pocket cost information and had a usual care provider. While I have

124

attempted to predict out of pocket costs for the entire sample without using information
about the usual care provider or average out of pocket costs for other visits, the amount of
variance in the cost data I am able to explain without these variables is very low (R-
squared = 0.02). Among the subset of the sample who have non-missing usual care
provider and out of pocket cost for other visits data the R-squared for the price regression
increases to 0.25. The regression used to predict out of pocket costs is reported in
Appendix A3. After these sample restrictions and eliminating observations with missing
data I have 18,393 observations for 5,041 infants, 49.2 percent of whom received no
well-baby visits during the survey.

Results

Table 3.1 summarizes the timing of AAP recommended visits based on the
schedule in effect from 1993 through 2005 (the study period) and provides population
estimates of compliance for each recommended visit. Only 48 percent of infants received
a well-baby visit during the one-month recommended care interval, and estimated
compliance falls to under 25 percent by the eighteen month care interval.

Some infants may receive care during other visits that makes up for missed well-
baby care. The second and third rows of Table 3.1 present the percentage of infants who
received at least one other preventive care visit or sick care visit during each
recommended care interval. During the first few intervals, more infants receive well-
baby visits than other types of visits but by after the first year, more infants receive other
visits than well-baby care. However, a substantial portion of infants (between 24.6
percent and 39.8 percent depending on the interval.) receive no visits of any type during

the interval.

125

The next four rows in Table 3.1 reports the average total number of well-baby,
other preventive care and sick visits and the total number of all visits received to date.
These figures include all visits received; if infants receive an extra visit during a
recommended care interval that visit is counted in these figures.

Although the average number of well-baby visits received is well below the seven
recommended before the 18 months care interval, the average number of doctor visits
received is 10.90. Thus, if infants are receiving well-baby type care during other visits, it
would seem they are receiving an adequate amount of care. However, the sample
correlation between sick visits received and well-baby visits is only 0.42. Furthermore,
the likelihood of receiving a recommended well-baby visit is increasing in the number of
other doctors visits received during the recommended care interval. This is true whether
other visits are measured in the preceding or concurrent recommended care interval, as a
count or as an indicator. Adding infant fixed effects does not change the result; infants
who receive at least one non-well-baby visit during a recommended care interval have a
12 percentage point higher likelihood of receiving the well-baby visit during that interval
(t = 15.17). Both of these statistics indicate the same children who receive more well-
baby care visits receive more sick visits.

It in total, it is unlikely that infants are receiving all of the care recommended
during other types of visits. Nonetheless, although 20 percent of infants have not
received any well-baby care by the 18 month care interval, only 1.69 percent had not
received any type of preventive care. So looking at well-baby care only likely
understates compliance, but by the broader measure of preventive care infants receive

only 4.15 visits out of the 7 recommended prior to 18 months.

126

From Table 3.1 it is clear that the way in which compliance is measured can have
a large impact on conclusions drawn. Table 3.2 compares measures of compliance in
previous studies to estimates based on my sample. After adjusting for differences in
sample and measure used, my estimates are generally in keeping with previous estimates
of compliance with well-baby care. Although, compared to Ronsaville and Hakim (2000)
and Byrd et a1 (1999), my estimates are substantially higher for black and Hispanic
infants and for Medicaid recipients. The higher rates in my data could be due to the
elimination of cost sharing for well-baby care under Medicaid/SCHIP in 1996 whereas
both Ronsaville and Hakim (2000) and Byrd et a1 (1999) use data that predates the
elimination of cost sharing. The differences in my estimates of compliance for blacks,
Hispanics and Medicaid recipients and previous estimates suggest the gap in receipt of
medical care between minorities and whites and publicly and privately insured children
narrowed over the time period. My data also approximate Seldon’s (2006) estimate quite
well. However, if I use his measure and restrict the sample to children age two months
and younger, the estimate drops to 62 percent. This suggests, in keeping with my
estimates in Table 3.1, children are not receiving on-time care.

Table 3.3 presents sample means and standard deviations for the subset of the
sample used in theremaining analyses (infants with usual care providers who saw a
medical professional for any reason in the past year). The variables listed are a subset of
those used in the actual analysis. See Appendix 3A for a more complete listing. Average
out of pocket costs (co-pays) are less than $10 and lowest among infants whose mothers
are not working. Although, there is some variation by insurance status behind these

figures. Among privately insured infants, the average co-pay is $14.11, for Medicaid

127

recipients the average co-pay is $0.38 and among the uninsured the average out of pocket
cost is $15.97.

Household income is higher among infants with non-employed mothers. Wages
(or predicted wages), which proxy for time cost among the non-employed, are highest
among employed women. The $8.50 figure for non-employed women includes predicted
wages. If I restrict the sample to only non-employed women for whom I observe past
wages, the mean is $8.27 (standard deviation 5.09).

A substantial minority of mothers in the sample do not have access to leave. Only
63 percent have access to any paid vacation leave, 55 percent have sick leave. Yet 87.6
percent of those with paid sick leave or 48 percent overall can use their leave to see the
doctor. F MLA eligibility is rare. Only 26 percent of mothers both work for covered
employers and meet the hours and tenure requirements to be eligible for FMLA leave.

Table 3.4 presents the estimation results of the base model (Equation 8) among
infants of employed and non-employed mothers and for the entire sample. The reported
estimates are marginal effects after probit estimation. The bottom row of Table 3.4
contains estimated own price and income elasticities. Among all infants the estimates
suggest a 1 percent increase in the out of pocket price of care is related to a 10.6 percent
reduction in the probability of receiving a given visit, This figure is nearly identical to
Colle and Grossman’s (1978) estimated own-price elasticity of demand for all
physicians’ visits among children ages 1 to 5. The estimated income elasticity implies a
1 percent increase in income is related to a 19.6 percent increase in the probability of
receiving a visit, which is substantially smaller than Colle and Grossman’s (1978)

estimate (which was 0.379).

128

The estimated cross-price elasticity is positive in the full-sample, which suggests
employment does not compete with well-baby care, but it is not significantly different
from zero. Among non-employed women, the estimated cross-wage elasticity is negative
but again not significant at conventional levels. As stated previously, current wages are
unobserved for women who are not working. The measure used to proxy for wages may
affect the estimated cross-price elasticities.

Table 3.5 compares estimated cross-wage elasticities using OLS predicted wages,
past wages when observed and a combination of both for non-employed women.
Unfortunately I was unable to find a viable identification strategy to produce selection
corrected estimates of wages. Mean OLS predicted wages are about $0.40 higher among
non-employed women for whom I observe past wages and using OLS predicted wages
yields a much larger (in absolute value) point estimate of the cross-wage elasticity (-0.27
vs. -0.01). However, the OLS based estimate and the past wage based estimate are not
statistically significantly different from zero or each other. Since there is'limited
information from which to predict wages (see Appendix Table A3.] for the equation used
to predict wages), the past wage measure is arguably a better proxy of available wage
offers than the OLS prediction. Yet, I observe past wages for only 22.2 percent of non-
employed women. Therefore, my preferred measure is a combination of OLS predicted
wages for non—employed women who are never observed working in the panel and past
wages where available. The last column in Table 3.5 shows that this measure produces a
cross-wage elasticity estimate in the full-sample similar to the estimate obtained when

only past wages are used.

129

Table 3.6 presents the results for the effects of maternal employment. Without
attempting to address endogeneity, the Sign pattern on the employment variables suggests
infants whose mothers are employed are more likely to receive care than those whose
mothers are not. However, the employment variables and the paid leave variable are not
significantly different from zero and FMLA is only moderately significant. The sign
pattern of the instrumental variables estimates contradicts the probit estimates and
suggests infants whose mothers work full time are even less likely than those whose
mothers work part-time to receive care. Additionally, the point estimates imply having
access to paid leave or being eligible for FMLA leave mitigates the negative effect of
maternal employment. Yet, none of these estimates is individually or jointly significant.

The last column of Table 3.6 displays fixed effects results. Access to paid leave
and FMLA leave eligibility do not vary enough to identify parameters and are omitted
from this model. Among mothers who change employment status or work schedule, the
estimates indicate infants have a 4.8 percentage point (9.4 -— 14.2) lower likelihood of
receiving recommended visits during periods when their mothers are employed full-time
than when not working and 9.4 percentage point higher likelihood during periods when
their mothers are employed part time than when not working. These estimates are
significant.

Table 3.7 presents the estimated relationships between paid and unpaid leave and
work schedule among infants with employed mothers only. Paid leave and FMLA
eligibility do not appear to have any effect on receipt of care here either and moreover
few other employment related variables are significantly related to receipt of care. In the

base model (Column 1), no employment related variables are significantly different from

130

zero. Column 11 adds work schedule measures to the list of covariates; they are omitted
in Column 1 because hours of work are arguably less plausibly exogenous than other
variables. As in Table 3.6, women who work part-time appear to be more likely to take
their children to well-baby care than those who work full-time. Although only
moderately significant, the implied percentage point difference in the probability of
receiving care between mothers employed part-time and those employed full-time (the
benchmark category) is 3.6 percentage points, which this translates to a 10 percent
increase at the mean likelihood of receiving care.

The positive relationship between part-time work and receipt of well-baby care
persists when the lagged proportion of workdays missed in the last survey round and
lagged number of visits are added to the model in Column 111. The lagged work days
variable is intended to separately capture availability and use of leave from work. The
sign of the estimated coefficient does suggest women who have missed more days of
work to date are less likely to take their child to the current visit but the relationship is
insignificant. Furthennore, including this variable does change the sign and magnitude of
point estimates for paid and unpaid leave variables but they are still not significantly
different from zero. This also does not change when the sample is restricted to women
.who were employed in the same job two years prior to the birth in Column IV to attempt
to address the potential endogeneity of paid leave.

When the number of visits received in the last round is included in Columns III
and IV, there is little change in the point estimates for out of pocket cost, which implies
little change in the own-price elasticity. Thus, as was apparent in the estimates of

compliance, infants who have received care in the past are more likely to receive care in

131

the future. But this seems not to affect relationships between receipt of care and key
covariates.

From the results in Table 3.7, it seems access to paid leave and F MLA eligibility
have no relationship with well-baby care use. However, things change when
relationships are estimated separately by wage level and work schedule and by
occupation. Tables 3.8 and 3.9 present these results. Each column in Table 3.8 presents
estimates for women in different work schedule and wage level combinations. Estimates
in the first column are based on women working part-time in jobs that pay wages below
the occupational median wage.

While none of the estimates for paid leave are significantly different from zero in
any wage and work schedule group, there is an interesting pattern for FMLA eligibility.
Women who work 40 hours per week in jobs that pay below median wages have an 8.8
percentage point lower probability of taking their babies to well-baby visits than those
who are not eligible for F MLA, however, those working more than 40 hours per week in
low wage jobs have a 29.5 percentage point higher likelihood of taking their babies to
recommended visits. Admittedly, women working long hours in low wage jobs represent
a small portion of the sample. ’

Yet, the large positive relationship between F MLA eligibility among long hours
low wage women and the significant negative relationship among low wage women
working 40 hours a week seem contradictory. Error in the F MLA elgibility variable is
one plausible explanation for this sign pattern. Technically, FMLA is only available to
employees who have worked at least 1,250 hours in the past twelve months. The FMLA

variable I have constructed is based on the hours of work mothers say are “usual” at the

132

time they are surveyed. There is no measure of total hours per year available in the data.
Women who say they usually work long schedules are more likely to accumulate at least
1,250 per year than those who say they usually work 36 hours per week. Thus, the FMLA
indicator among women employed full-time (35 to 40) hours per week may reflect an
underlying negative relationship between continuous full-time work and well-baby care
rather than the relationship between FMLA eligibility and well-baby care. Alternatively,
the measure may be correct but longer hours employees may be more familiar with

F MLA because they face a more severe time crunch. Moreover, their employers may
perceive the of F MLA leave among long hours employees as more legitimate and so they
may be more willing to facilitate FMLA leave.

Table 3.9 further unpacks the relationships between paid leave and well-baby care
by analyzing occupations separately. From these estimates, one can see why the
estimates of paid leave in the full sample have been insignificant. For paid sick leave,
there are both positive and negative point estimates across occupations and work
schedules within occupations. Among mothers in managerial occupations, those working
part-time who have paid sick leave have a 53.4 percentage point lower probability than
other mothers working part-time (and 19.8 percentage point lower probability than those
working full-time) of taking their baby to a given well-baby visit. Those working more
than 40 hours per week who have paid sick leave are also more likely than other women
in management occupations to take their babies to visits but those who say they “can” use
their paid sick leave for doctors visits are less likely. Yet, paid sick leave that can be
used for doctors visits is related to higher rates of well-baby care use among mothers in

sales occupations who work long hours and mothers in clerical occupations who work

133

either part-time or long hours. In most occupations, paid vacation leave has no
significant relationship with well-baby care but among clerical and production workers
who work long hours, access to paid vacation leave more than offsets the negative
estimated effect of long hours employment.

Results for F MLA also vary substantially across occupations. First, F MLA
eligible mothers in sales occupations who work 35 to 40 hours per week have a 20.1
percentage point higher likelihood of taking their babies to visits than those working the
same number of hours who are ineligible for F MLA. But among production workers,
FMLA eligible mothers working 35 to 40 hours per week are less likely than ineligible
mothers to take their children to well-baby visits. These opposing occupational level
relationships led to the null finding for FMLA overall. Furthermore, the positive
relationship between FMLA leave eligibility and well-baby care among long hours
employees identified in Table 7 appears to emanate from women in professional, service
and production occupations. In all three occupational groups, the estimated effect of
F MLA is large enough to offset any negative effects of long hours work.

Hours of work have different relationships with well baby care across occupations
as well. Part-time workers in managerial and sales occupations and are significantly
more likely to take their babies to visits than full-time workers but in production
occupations they are significantly less likely to take their babies to visits. Somewhat
surprisingly, working over 40 hours per week has a significant negative relationship with
well-baby care only among production employees. However point estimates are negative
in all occupations except service occupations and only 8 percent of mothers in service

occupations worked over 40 hours per week.

134

Also, an interesting pattern in the relationship between wages and well-baby care
emerges in Table 3.9. Throughout the analysis, wages have been positively related to
receipt of well-baby care among employed mothers. Table 3.9 shows the relationship
between wages and visits is positive and significant for professional employees only.
Among managerial employees, there is a significant negative relationship that is almost
identical in magnitude to the relationship among professional employees. Estimates
across other occupational groups are small and statistically insignificant.

Why would the relationship between wages and well-baby care be positive for
professionals? It could be many of these women are nurses or other health care
professionals. Nursing is among the higher paying professional jobs available without an
advanced degree and is one of the more prevalent occupations among women.
Unfortunately the occupational categories used in Table 6.8 are the lowest level of
aggregation available in the MEPS so this hypothesis cannot be directly tested. But
according to Bureau of Labor Statistics estimates, the average hourly wage among
registered nurses is $30.04 and all but the lowest 10 percent of earners had wages over
$20.20. The average wage among women in production occupations in my sample is
$19.05. In 2000, 94 percent of all registered nurses were female and among females
employed in professional occupations (using the same classification as in the MEPS data)

17 percent were registered nurses and an additional 8 percent were in other health related
. rr
occupatrons.

Although Tables 3.8 and 3.9 provide insight to explain the null estimates for paid

and unpaid leave in the sample as a whole and they suggest some substantial positive

 

H Author’s tabulations using 2000 Current Population Survey.

135

relationships between paid leave availability and compliance with well-baby care in
certain jobs, these findings are based on small portions of the overall sample. As in any
small sample analysis, estimates will be especially sensitive to outliers and may not
generalize beyond the sample.

To provide an additional test of the relationships between paid and unpaid leave
and well-baby care, I take the proportion of the sample that received at least one well-
baby care visit and examine the relative likelihood of that visit occurring on a Monday
through Thursday rather than on a Friday or weekend. Employees with paid and unpaid
leave should be able to schedule weekday visits more easily. I treat Friday separately
under the assumption and incidental evidence that flextime and compressed work week
schedules commonly influence working time on Fridays more so than other days of the
week.

Table 3.10 presents the days of the week analysis. Estimates are reported as
marginal effects. The first column shows the percentage point change in the likelihood of
a doctor visit on a weekday other than Friday. Mothers who are employed full-time have
a 4 percentage point lower probability of taking their babies to well-baby care Monday
through Thursday as compared to mothers who work part-time and a 7 percentage point
lower probability than non-employed mothers. However, having paid vacation time
entirely offsets the negative effect of firll-time employment. Paid sick leave appears to
reduce the likelihood of Friday visits but the relationship with weekday visits other than
Friday is insignificant. There is also no evidence of a relationship between FMLA leave

eligibility and the distribution of visits across days of the week.

136

Mothers who say their paid sick leave can be used to visit the doctor have a 15
percentage point lower likelihood of taking their baby to the doctor Monday through
Thursday. This finding helps to explain some of the negative findings for paid leave to
see the doctor presented in Tables 3.8 and 3.9. It could be that respondents interpreted
the survey question as “must you use a paid sick day to visit the doctor” rather than “are
you permitted to use a paid sick day to visit the doctor”. If that is the case, this policy
actually limits schedule flexibility.

Discussion and Conclusion

This paper examined the determinants of well-baby care receipt in the first two
years of life among a nationally representative sample of babies born between 1994 and
2005. Most babies do not receive the AAP recommended amount of well-baby visits. I
estimate that at age 18 months the average child has received only 2.25 of 8
recommended visits. Although past studies have reported higher compliance rates, using
my data 1 demonstrate the differences in estimates result entirely from the measure of
compliance used. The fact that compliance rates are low despite the low out of pocket
cost of well-baby care and high insurance rate among children under two years old
suggests time constraints may be important. To consider this possibility, I examine
compliance with recommended well-baby care among employed and non-employed
mothers and by access to paid and unpaid leave and work schedule.

Despite the low cost of care, I find a 1 percent increase in co-pays (around $1 at
the mean) is related to a 10.6 percent reduction in the probability of receiving a given
visit. For income, I find a 1 percent increase is related to a 19.6 percent increase in the

probability of receiving a given visit. The sign of the cross-price elasticity varied across

137

specifications and estimates were generally insignificant. Yet when the relationship
between maternal employment and well-baby care is directly estimated, 1 find suggestive
evidence of a negative relationship between full-time maternal employment and well-
baby care.

For paid and unpaid leave, I find the implied effects vary substantially across
occupations and work schedules as well as earnings levels. FMLA leave is most
important for long hours (over 40 hours per week on average), low wage mothers.
Among these women, those with FMLA leave have a nearly 30 percentage point higher
probability of taking their baby to a given visit. Further investigation of this finding
reveals F MLA leave is most important for long hours employees in professional, service
and production occupations and for full-time (35 to 40 hours per week) employees in
sales occupations. The variation in the effects of F MLA leave suggests broad
government mandated workplace flexibility may have very different effects on different
types of jobs. Among the groups who are arguably most in need of flexibility (long
hours, low wage employees), my findings indicate F MLA leave may substantially
increase mother’s ability to get their babies to the doctor. The large estimated
relationships between F MLA leave eligibility and well-baby care are unique. Most
studies of FMLA leave have found significant effects of the law on leave coverage but
little association with changes in behavior.

The estimated relationships between paid leave and well-baby care also vary
dramatically by occupation and work schedule. Paid sick leave appears to matter most
for mothers employed full-time in sales occupations and part-time production workers.

Among mothers working long hours in production occupations, having paid sick leave

138

that can be used for doctors visits appears to more than offset the negative effect of there
long work week. Long hours production employees with paid vacation leave also have a
dramatically higher likelihood (estimated at 89.1 percentage points) of taking their babies
to well-baby visits than long hours production workers without paid vacation. Yet among
these positive estimates, there were many significant negative estimates as well. In
general, it seems the role of paid leave depends on the job. In some jobs it may enable
employees to take a child to the doctor, in others it may ration time off and inhibit doctor
visits. Nonetheless, I do find employees with paid vacation leave have a significantly
higher probability of taking their child to a well-baby visit on a weekday (as well as on a
weekday other than Friday) than a weekend.

The findings for maternal employment and for differences in the effects of paid
and unpaid leave across occupations, work schedules and wage groups must be
interpreted with caution. As discussed, the. identification strategy used to attempt to
obtain unbiased estimates for maternal employment is based on the assumption that
mothers bear the full responsibility for taking children to well-baby care. For some
groups, my analysis of paid and unpaid leave by occupation, work schedule and wage
group arguably cuts the data too thin. As in any small sample, estimates will be
especially sensitive to outliers and may not generalize to larger populations. Nonetheless,
the patterns I find indicate future studies should consider the possible differences in the
relationships between flexible working-time policies and practices across job types

whenever the sample size permits.

139

Figure and Tables

Figure 3.1 Correspondence between AAP Recommended Care Intervals and
Distribution of Actual Visits

w. ‘1‘ l
u) —l
‘—

10

Percent

 

 

10
Age in Months

Source: Author tabulation. See Table 1 for AAP Visit Schedule.

140

Table 3.1 Estimated Compliance with the AAP Well- Baby Visit Schedule

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Recommended 1 2 4 9 12 15 18
Visits Schedule Month Months Months Months Months Months Months Months
0 ‘ _
g’afiicjgfidinwe” 43.0% 40.3% 35.1% 34.3% 27.2% 30.7% 21.2% 23.1%
Interval (2.01) (1.6) (1.5) (1.2) (1.2) (1.7) (1.1) (1.2)
% Received 0 o o o o 0
Preventive Visit 28.1% 31.3/o 27.8/o 32.6% 23.9/o 30.2/o 23.2/o 27.8/o
in Interval (1.19) (1.01) (1.27) (1.00) (0.91) (1.29) (0.92) (1.27)
% Received Sick 22.0% 19.7% 23.7% 33.2% 38.7% 37.4% 34.7% 47.9%
Visit in Interval (1.03) (1.12) (1.01) (1.34) (1.36) (1.82) (1.00) (1.08)
% Received no
Visits of Any 24.6% 26.1% 30.2% 25.1% 33.0% 28.7% 39.8% 28.2%
Kind in Interval (1.38) (1.49) (1.00) (0.95) (1.38) (1.22) (1.03) (1.10)
Average # of
Well-Baby Visits n/a 0.74 1.21 1.53 1.80 2.03 2.25 2.25
1 (0.03) (0.04) (0.05) (0.04) (0.05) (0.05) (0.04)
to Date
Average # of
Other Preventive n/a 0.46 0.95 1.46 2.14 2.74 3.39 4.15
Care Visits to (0.03) (0.04) (0.05) (0.05) (0.05) (0.04) (0.04)
Date
Average # of Sick n/a 0.43 0.91 1.46 2.39 3.21 4.00 4.50
Visits to Date (0.03) (0.04) (0.05) (0.06) (0.06) (0.08) (0.05)
Average Total #
of Office-Based n/a 1.64 3.07 4.45 6.33 7.98 9.63 10.90
_ , 2 (0.03) (0.05) (0.06) (0.08) (0.09) (0.13) (0.13)
Vrsrts to Date
0 .
xi)! 3:331:63; ts n/a 52.0% 41.3% 35.5% 30.1% 26.7% 21.7% 20.0%
y (0.02) (1.1 I) (0.79) (0.55) (0.45) (0.54) (0.28)
to Date
% Received no
WB or Preventive n/a 29.3% 16.5% 10.8% 5.99% 4.17% 2.32% 1.69%
Care Visits to (1.31) (0.72) (0.27) (0.17) (0.08) (0.07) (0.03)
Date
0 .
$23333" n/a 23.6% 12.9% 7.37% 3.34% 1.74% 0.30% 0.50%
Kind to Date (1.38) (0.73) (0.24) (0.12) (0.07) (0.03) (0.01)
Sample Size 3,753 3,385 3,249 3,096 3,130 3,407 3,443 3,030

 

 

 

 

 

 

 

 

 

Notes: Sample weights are applied to estimate population parameters.

This figure includes additional unscheduled well-baby visits; a child who received two visits at age 2
months would contribute to visits to the average for that age.

F 1gures Include any office based vrsrts wrth any medical professronal 1nclud1ng srck vrsrts, well-baby care
visits, and other preventive care.

141

 

Table 3.2 Replication of Previous Compliance Estimates

 

 

 

Hispanic: 49%

Privately Insured: 53.9%

Ronsaville and Hakim Byrd et al. (1999) Selden (2006)
(2000)
Their Estimate White: 58%
Black: 35% Medicaid: 36% 83.2%
Hispanic: 37% Privately Insured: 49%
My Estimate White: 62%
Black: 49% Medicaid: 43.3% 79.0%

 

Measure Used

 

 

Receipt of three or more
well-baby visits or visits
where an immunization
occurred by age 6
months. 19

 

Percentage of
recommended care
intervals where at least
one visits was received
among children who
received at least 1 visit.

 

Mean percentage of
recommended well-baby
visits received among
infants. Visits include
general check-ups and
visits where
immunizations occurred.

 

142

 

Table 3.3 Sample Means and Standard Deviations for Selected Variables1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

All Mother Emgloyed Mother Non-Employed
$7.62 $9.92 $5.55
Out ofPocket Cost (17.01) (18.64) (15.11)
Diagnosed with Serious 0.21 0.22 0.21
Condition (0.36) (0.36) (0.67)
Uninsured 0.1 1 0.11 0.12
(0.31) (0.31) (0.32)
Income $36,937.3 $35,712.09 $38,038.51
(32975.20) (31 189.70) (34,465.60)
. . $10.41 $12.54 $8.50
Wage (or trme cost estimate) (6.74) (8.44) (3.78)
. 0.25
Salarred (0.43)
. 0.55
SICk Leave (0.50)
Paid Leave Used to See 0.48
Doctor (0.50)
. 0.63
Vacation Leave (0. 48)
0.26
FMLA (0.44)
Number of Em lo ees 1'51
p V (1.33)
- 0.09
Unron Member (0.28)
0.10
Management (0. 30)
. 0.25
Professronal (043)
Service 0'2]
(0.41)
0.14
Sales (03 5)
. 0.20
Clerical (0.40)
. 0.09
Productron (0. 2 8)
. 0.02
Other Occupatrons (0.13)
Federal Government 0.03
Employee (0. l 6)
State Government Em 10 ee 0'09
p V (0.29)
3 0.05
More than One Job (022)
Tenure 3'20
(3.78)
Observations 18,393 8,706 9,687

 

 

 

 

 

See Appendix A3 for complete listing of variables and definitions
Mean household income is reported here, full income is used in the analysis.

3 For multiple job holders, all other job related variables are based on the main job.

143

Table 3.4 Base Model (Equation 8) Results by Maternal Employment

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(1) (ll) (1")
Non- Employed All
Employed
-0.003+ -0.014* -0.005**
Out of Pocket Cost (0.002) (0.003) (0.002)
2 0.000+ 0000* 0.000+
** * *
Full Income (10 Thousands) (iooggfl 36000052) 36000062)
2 -0.000 -0.000 0000
Full Income (10 Thousands) (0.000) (0.000L (0.000)
. -0.028 0.01 1 0.005
ln(wage or time cost) (002]) (0.013) (0.0] 1)
Observations 9687 8706 1 8393
(Pseudo) R-squared 0.101 0.095 0.092
Own Price Elast -0.054+ -0.036** -0.106*
(0.033) (0.060) (0.046)
. -0.1 18 0.034 0.017
C’OSS'Pnce E'a“ (0.089) (0.040) (0.038)
Income Elast 0.218” 0.176" 0.196M
(0.074) (0.059) (0.050)

 

(1) Babies with non-employed mothers.
(11) Babies with employed mothers only.
(111) All babies.

 

Note: Estimates control for insurance coverage by source, characteristics of the usual care provider,
whether or not the child has previously been diagnosed with a serious health condition, mother’s subjective
rating of child’s health, mother’s educational attainment, number of children mother has ever had and
number of children in the household, recommended care interval, race and ethnicity, non-English speaking
household, mother’s marital status, region and urban residence.

144

Table 3.5 Sensitivity of Estimated Cross-Wage Elasticity to Measure of Wages

 

Combined Measure
in Among All non-

 

 

 

 

 

 

 

OLS Predicted Wage Average Past Wag Employed
Mean Predicted Value of Time $9.46 $9.09 $8.50
(Standard Deviation) (4.30) (3.78)
Estimated Marginal Effect for -0.27 -0.01 -0.03
ln(wage) in base model (0.17) (0.02) (0.02)
(Standard Error)
Observations 2,157 2,157 9,687

 

Notes: For comparability across methods, the estimates in the first two columns include only those women

for whom past wages are observed.

145

 

Table 3.6 Estimated Effects of Maternal Employment and Paid Time Off

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Probit IV Probit Fixed Effects
Out of Pocket -0.004* -0.006 -0.001
Cost (0.002) (0.009) (0.001 )
Out of Pocket 0.000* 0.000 0.000
C053 (0.000) (0.000) (0.000)
Full Income (10 0.006“ 0.006 -0.01 1
Thousands) (0.002) (0.012) (0.01 1)
Full Income2(10 -0000 -0000 0.000
Thousands) (0.000) (0.000) (0.000)
In (wage) -0.008 -0.223 -0.056
(0.019) (0.317) (0.043)
Employed .033 -0.960 0.094*
(0.025) (2.776) (0.045)
Employed Full- -0.009 -4.887 0142”
Time (0.026) (7.563) (0.050)
Any Paid Leave -0.004 3.890
Available (0.026) (4.817)
. . -0.041+ 2.041
FMLA Eligible (0.026) (3.848)
Observations 7,667 7,667

 

Notes: Reported estimates are marginal effects. Father’s wage, hours of work, hours of work squared and
whether or not he is salaried (as opposed to hours) are used to instrument for employment, full-time
employment, paid leave and FMLA. Estimates are based on infants whose father is present in the
household and for whom the father has non-missing wage and hours of work information. Paid leave and
FMLA variables are excluded from the fixed effects estimation because there is too little within person
variation to identify effects; insurance status, subjective health indicators and recommended care interval

are controlled for.

146

 

Table 3.7 Results for Paid and Unpaid Leave and Other Job Attributes

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(I) (II) (III) (IV)
0014* 0014* 0013* -0.016*
0‘" °f PM“ C0" (0.003) (0.003) (0.003 ) (0.004)
2 0000* 0000* 0000* 0000*
0‘" 0f ”“3““ C0“ (0.000) (0.000) (0.000) (0.000)
Full Income (10 0.005" 0.005” 0.003” 0.001
Thousands) (0.002) (0.002) (0.002) (0.003)
Full Income2 (,0 0000 0000 -0000“ 0000
Thousands) (0.000) (0.000) (0.000) (0.000)
In (wage) 0.012 0.012 0029+ 0.059**
(0.014) (0.014) (0.015) (0.027)
Salaried 0.000 0005 0021 0024
(0.020) (0021) (0.022) (0.034)
. . 0.041 0.050 0.001 -0.026
Pa‘d 5'“ Leave (0.032) (0.032) (0.035) (0.061)
Sick Leave used to 0023 -0.026 0.001 0076+
See Doctor (0.028) (0.028) (0.030) (0.044)
Paid Vacation Leave 0005 0.004 0004 0009
(0.021) (0.021) (0.023) (0.047)
. . 0010 0004 0.017 0.048
FMLA Ehg‘b'e (0.022) (0.023) (0.025) (0.041)
# Employees (in -0.016 -0.015 -0.026 -0.045
100s) (0.020) (0.020) (0022) (0.039)
# Employee} (in 0.003 0.003 0.004 0.006
1005) (0.004) (0.004) (0.004) (0.007)
Union Member 0031 0025 0041 -0.036
(0.029) (0.029) (0.032) (0.048)
Managerial -0005 0007 0009 -0.061
(0.030) (0.030) (0.034) (0.057)
Professional 0.009 0.004 0.015 0029
(0.026) (0.026) (0.029) (0055)
Sales 0.018 0.013 0.009 0009
(0.025) (0.025) (0.029) (0.060)
Clerical 0003 0003 0.007 0.053
(0.024) (0.024) (0.027) (0.053)
Production 0023 0020 0044 0139M
(0.030) (0.030) (0.034) (0.059)
Other Occupation 0030 0019 0037 0194"
(0.058) (0.062) (0.082) (0.088)
0.066 0.067 0.032 n/a
3"” Emphyed (0.083) (0.084) (0.088)
Federal Govt 0009 0012 0.006 0.051
Employee (0.045) (0.046) (0.044) (0.073)
0.004 0001 0.014 0.008
State Go“ Employee (0.029) (0029) (0.031) (0.046)
0042 0047 0035 0052
More “‘3" 1 10" (0.031) (0.031) (0.036) (0.072)
Tenure 0.008 0.003 0019* 0028+
(0.006) (0.006) (0.006) (0.016L
2 0000 0000 -0.001** 0001+
Tenure (0.000) (0.000) (0.000) (0.001)

 

 

 

 

 

147

 

Table 3.7 Results for Paid and Unpaid Leave and Other Job Attributes

 

 

 

 

 

(Continued)
Part Time (< 35 Hrs 0037+ 0036+ 0.052
per Week) (0.019) (0.022) (0.043)
Over Time (>40 Hrs 0.016 0.014 0.014
per Week) (0.023) (0.025) (0.038)
% Work Days Missed -0.065 0060
Last Round (0.042) (0.067)
Visits Received Last 0215* 0263*
Round (0.014) (0.023)
Observations 8 7 O 6 8 6 4 3 4736 1937

 

 

 

 

 

 

Note: All regressions also control for race/ethnicity, mother’s marital status, mother’s education, number

and ages of children, recommended care interval, region, characteristics of the usual care provider,

insurance coverage, and objective and subjective infant health measures.

148

 

Table 3.8 Results for Paid and Unpaid Leave and Other Job Attributes by Work
Schedule and Wage Level

Note: All regressions also control for race/ethnicity, mother’s marital status, mother’s education, number
and ages of children, recommended care interval, region, and insurance coverage (uninsured, privately
insured or Medicaid). Usual care provider variables, insurance coverage variables and characteristics of the
employer were collapsed due to the small number of observations in some wage/schedule groups. Below
and above median wage designations are with respect to the occupation. As before, characteristics of the
main job are used for multiple job holders.

1
Excludes self-employed.

I49

Table 3.8 Results for Paid and Unpaid Leave and Other Job Attributes by Work

Schedule and Wage Level

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Part Time Full Time Over Time (> 40 Hours)
Below Above Below Above Below Above
Median Median Median Median Median Median
Out of Pocket -0.018* -0.028* -0.014* -0.013* -0.044* 0031*
Cost (0.005) (0.006) (0.004) (0.004) (0.01 1) (0.007)
Out of Pocket 0000* 0.000* 0000* 0000* 0001* 0000*
Cost2 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Full Income (10 0028* -0.009 0.004 0.009 -0.031 0003
Thousands) (0.009) (0.010) (0.006) (0.006) (0.034) (0.009)
M, Income? (10 0001** 0.000 0000 0000 0.001 0000
Thousands) (0.000) (0.000) (0.000) (0.000) (0.001) (0.000)
Ln (wage) -0.0 l 3 -0.068 0.015 0.046 0124+ 0217+
(0.019) (0.089) (0.025) (0.060) (0.068) (0.1 17)
Salaried -0.092 0.148" -0.011 0.015 0.027 0.020
(0.058) (0.070) (0.038) (0.034) (0.109) (0.066)
Paid Sick Leave 0.113 0.034 0.068 0.040 0.166 -0.150
(0.099) (0.1 18) (0.064) (0.050) (0.152) (0.132)
Sick Leave to -0.019 -0.019 -0.030 -0.061 -0.1 16 0.054
See Doctor (0.091) (0.097) (0.059L (0.045) (0.1 15) (0.089)
Paid Vacation 0027 -0.075 0.029 0.065 0.147 0070
Leave (0.043) (0.077) (0.033) (0.045) (0.1 19) (0.089)
-0.088** -0.023 0.295** -0.037
FMLA (0.038) (0.037) (0.120) (0.101)
# Employees (in -0.009 -0.047 0.029 0.019 -0.322** 0.079
1005) (0.048) (0.059) (0.035) (0.038) (0.156) (0.102)
# Employeesz -0.003 0.008 -0.007 -0.002 0.062" -0.010
(in 1005) (0.010) (0.012) (0.007) (0.007) (0.029) (0.017)
Union Member 0.127 0.054 —0.001 -0.014 -0.006 -0.211**
(0.1 10) (0.079) (0.059) (0.045) (0.239) (0.090)
Managerial -0.020 0.135 0.028 0.029 -0.075 -0.092
(0.076) (0.1 12) (0.059) (0.070) (0.145) (0.127)
Professional 0038 0.069 0.043 0.029 -0.1 12 0.159
(0.049) (0.090) (0.060) (0.066) (0.180) (0.132)
Sales 0064 0.133** 0.056 -0.025 0.004 0.160
(0.040) (0.060) (0.062) (0.049) (0.284) (0.123)
Clerical 0.013 0.039 0.027 -0.031 -0.068 0.093
(0.048) (0.085) (0.047) (0.044) (0.177) (0.126)
Production 0134+ 0293+ 0.080 -0.107** -0.305 -0.031
(0.072) (0.155) (0.057) (0.043) (0.249) (0.137)
Other 0.147 -0.030 0.136 -0.215+ 0.024 -0.310*
Occyation (0.168) (0.247) (0.1 17) (0.1 17) (0.189) (0.099)
Private Sector 0006 -0.028 0.029 0.002 -0.244** 0.124
Employee] (0.059) (0.067) (0.043) (0.042) (0.098) (0.077)
More than 1 Job 0082 -0.050 0.015 0.025 0190 -0.090
(0.050) (0.074) (0.075) (0.072) (0.212) (0.105)
Tenure -0.018 -0.012 0.017 0.021“ 0.051 0.003
(0.015) (0.016) (0.010) (0.010) (0.051) (0.026)
2 0.001 0.002 -0.00 l + -0.001 -0.006 0.001
Tenure (0.001) (0.001) (0.001) (0.001) (0.004) (0.002)
Observations 1622 1 l 15 2152 2696 366 692

 

150

 

Table 3.9 Results for Paid and Unpaid Leave and Other Job Attributes by

Occupation and Work Schedule

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Managerial Professional Service Sales Clerical Production

Out of Pocket -0.060* -0.030* -0.013* -0.01 l** -0.007+ 0022*
Cost (0.009) (0.004) (0.004) (0.005) (0.003) (0.008)
Out of Pocket 0001* 0.000* 0.000* 0.000** 0000+ 0.000**
Costz (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Full Income (10 0.005 0.007 0.013+ 0023* -0.002 -0.006
Thousands) (0.004) (0.007) (0.007) (0.008) (0.007) (0.007)

Full Income2 0000" -0.000 -0.000 -0.000** -0.000 0.000
('0 Thousands) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

In (wage) -0.071* * 0,070+ -0.002 -0.039 0.006 0.102
(0.032) (0.037) (0.023) (0.032) (0.031) (0.080)

Salaried 0108+ -0.005 -0.026 -0.042 0.050 -0.209*
(0.062) (0.038) (0.043) (0.068) (0.045) (0.028)

PT 0336* 0.017 0.026 0123+ 0.025 -0. 178*
(0.1 1 1) (0.091) (0.039) (0.063) (0.067) (0.052)

OT 0102 -0.123 0.007 -0.046 -0.1 17 -0.454*
(0.219) (0.208) (0.077) (0.143) (0.172) (0.041)

Paid Sick Leave -0.040 0.136 0087 0169+ 0.064 -0.015
(0.180) (0.1 15) (0.072) (0.102) (0.082) (0.061)

Paid Sick -0.534* -0.072 0.106 0.140 -0.263** 0789*
Leave*PT (0.094) (0.181) (0.164) (0.166) (0.115) (0.113)

Paid Sick 0380+ 0.016 -0.267* 0167 -0.359* -0.123
Leave*OT (0.215) (0.231) (0.014) (0.149) (0.035) (0.148)
Sick Leave to 0.144 0 140+ 0.077 -0.104 -0.043 0077
See Doctor (0.131) (0.076) (0.082) (0.080) (0.071) (0.065)

Doc Leave*PT 0.025 0.145 -0.076 -0.003 0.507“ -0.121
(0.218) (0.138) (0.1 17) (0.148) (0.205) (0.150)

Doc Leave*OT -0.421* 0.1 16 0.777* 0.258 0.406** 0.714*
(0.161) (0.100) (0.012) (0.233) (0.166) (0.144)

Paid Vacation 0.036 0.068 0.022 0.043 -0.093 0.077
Leave (0.121) (0.071) (0.049) (0.084) (0.074) (0.054)
Paid Vacation 0.215 -0. 142 -0.038 -0.144 -0.017 -0.146
Leave*PT (0.170) (0.109) (0.070) (0.092) (0.1 17) (0.170)
Paid Vacation 0.192 -0.071 —0.090 0.109 0540* 0.891 *
Leave*OT (0.238) (0.1 14) (0.1 18) (0.181) (0.191) (0.016)
FMLA -0.1 10 -0.058 0.015 0.201" —0.027 -0.141**
p (0.078) (0.050) (0.064) (0,097) (0.050) (0.057)

* -0.238** 0.174** 0.351" -0.108 -0.070 0774*
FMLA OT (0.108) (0.084) (0.175) (0.119) (0.099) (0.077)
# Employees (in 0123+ 0075+ -0.022 0.013 -0.029 0,085+
1005) (0.069) (0.045) (0.047) (0.050) (0.042) (0.050)
# Employees? 0019 0.011 0.006 0002 0.004 0014
(in 1005) (0.012) (0.008) (0.009) (0.010) (0.008) (0.009)

Union Member 0323* 0094+ 0.063 -0.050 0131” 0.088
(0.1 13) (0.052) (0.064) (0.101) (0.065) (0.075)

Private Sector 0,165+ 0.010 -0.024 0.000 -0.020 0.074
Employee] (0.095) (0.039) (0.052) (0.143) (0.062) (0.066)

Tenure 0.217** -0.1 15+ 0072 0.018 0.128 -0.005
(0.104) (0.067 ) (0.047) (0.091 ) (0.087) (0.147)

 

151

 

Table 3.9 Results for Paid and Unpaid Leave and Other Job Attributes by
Occupation and Work Schedule (Continued)

 

 

 

2 0.005 0.003 0000 0041** 0063* 0007
Tenure (0.019) (0.013) (0.01 1) (0.017) (0.013) (0.016)
Obs in Group
(Total 8,514) 870 2152 1814 1203 1732 755

 

 

 

 

 

 

 

 

Note: All regressions also control for race/ethnicity, mother’s marital status, mother’s education, number
and ages of children, recommended care interval, region, and insurance coverage (uninsured, privately
insured or Medicaid). Usual care provider variables, insurance coverage variables and characteristics of the
employer were collapsed due to the small number of observations in some wage/schedule groups. Below
and above median wage designations are with respect to the occupation. As before, characteristics of the
main job are used for multiple job holders. Women working in “other occupations” (farming and
construction) are omitted due to small sample size.

I
Excludes self—employed.

152

Table 3.10 Probability of Weekend, Weekday (except Friday) and Friday Visits by

Employment Status, Paid Leave and Occupation (Multinomial Probit)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Weekday Friday Friday Weekend

Employed -0.03 0.08 0.02

(0.03) (0.02) (0.01)
Employed 35 hours or -0.04** 0.03* 0.01
more per week (0.02) (0.01) (0.01)
Employed with Paid 007“ -0.05* -0.03*
Vacation Leave (0.022) (0.02) (0.01)
Employed with Paid 0.056 -0.06* 0.01
Sick Leave (0.03) (0.03) (0.02)
Employed with Paid -0.15** 0.13** 0.02
Leave to See Doc. (0.03) (0.03) (0.02)
Employed with FMLA 0.01 -0.01 0.00
Leave (0.02) (0.02) (0.01)
Employed in 0.04 -0.01 -0.03**
Management Occ. (0.02) (0.03) (0.01)
Employed in 0.03 0.00 -0.04**
Professional Occ. (0.03) (0.03) (0.01)
Employed in Service 0.04 -0.02 -0.02
Occ. (0.03) (0.02) (0.01)
Employed in Sales Occ. -0.06+ 0.05 0.01

(0.03) (0.03) (0.02)
Employed in Clerical 0.05+ -0.03 -0.02+
Occ. (0.03) (0.02) (0.01)
Employed in -0.03 0.03 -0.00
Production Occ. (0.04) (0.04) (0.02)
Employed in Other -0.05 0.05 -0.01
Occupation] (0.07) (0.07) (0.03)
Observations 6,440 6,440 6,440

 

 

 

 

 

Notes: Significance is for 2 tailed Z test

I . . . .
Other occupation includes farm and construction occupations.

153

Appendix A3

Prediction of Time Cost for Non-Employed Women

To predict the time costs for non-employed women used in the analysis in
Chapter 3 I use an OLS regression of wages among employed women on maternal
education, age, region of residence, number of children, race, ethnicity and marital status.
The results of this estimation are presented in Table A3.]. Using these covariates, I
obtain an R-squared of 0.29. Although these estimates are based on a selected sample,
there are no readily apparent valid exclusion restrictions to enable selection corrected
predictions.
Prediction of Out of Pocket Cost for Infants who do not Receive any Visits

As explained in the text, 50 percent of infants do not receive any well-baby visits
during the survey. Among infants who do receive at least one visit I can use price data
from past visits as a proxy for price of visits not received but I must predict out of pocket
costs for infants who did not receive any visits. To do so, I restrict the sample to infants
who have a usual care provider and received one office based medical visit of any kind
within the past year. Without doing so, the amount of variation in the cost of visits I can
explain is too low to convincingly predict missing out of pocket cost data (R-squared =
0.02). With these variables I can explain approximately 25 percent of the variation in out
of pocket costs. The out of pocket cost regression is reported in Table A3.2. Because of
the highly right skewed distribution of out of pocket costs I omit the top 1 percent. The
high proportion of 0’s in the actual price data preclude a log transformation. Also, since
cost sharing was eliminated under Medicaid in 1996, I assign $0 out of pocket cost to all

Medicaid infants with missing price data from 1997 forward.

154

Table A3.1 Equation for Predicted Wages for Non-Employed Mothers: OLS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ln(wage)
Less than High School 0148”
0.015
GED -0.086**
(0.023)
4 Year College Degree 0.484"
(0.016)
Graduate Degree 0732“
(0.022)
Mother’s Age 0.059"
(0.008)
Mother’s AgeT '0'001"
(0.003
Urban 0.] 15**
(0.013)
NE 0043*
(0.018)
MW 0064"
(0.016)
S 0070"
(0.014)
Number of Children -0.008
(0.006)
Number of Children < 6 -0.017*
(0021)
Black 0091 **
(0.015)
Asian, American Indian, Pacific Islander 0.017
(0.021)
Hispanic -0.040**
(0.015)
Non-English Speaking -0.176**
(0.020)
Married 0090"
(0.013)
Observations l 4,3 70
R-squared 0.290

 

155

 

Table A3.2 Equation for Predicted Out of Pocket Cost

 

 

 

 

 

 

 

Out of Pocket Cost
Recommended Care Interval (e.g. 1 month) 0,26]+
(0.145)
Private Insurance through Employer -0.862
(1.000)
Private Non-Group Insurance 3.184
(2.141)
Private Insurance through Self Employment 2.270
(3.457)
Other Private Group Insurance 5.715
(7.684)
Other Private Insurance 13.235
(9.659)

 

Usual C are Provider is: (Omitted C ategory is Pediatrician)

 

 

 

 

 

General Practitioner -2.016
(1.266)
OB/GYN 3.779+
(2.138)
Other Medical Doctor -4.713
(8.879)
Nurse Practitioner 2.914
(6.750)
Physician’s Assistant 18.693“
(7.133)

 

Type of Practice: (Omitted Category is Individual Doctor in Group Practice)

 

 

 

 

 

 

 

 

 

 

 

 

Individual Doctor’s Office -1 .084
(1.152)
Group Practice -2.500*
(1.003)
Average Cost per Office Based Visit (Visits of Any 0.444**
Type Received During the Year) (0.077)
Urban 1.266
(0.952)
NE -1 .928*
(0.835)
MW 2386*
(1.097)
S 1.394
(1.019)
Constant 6.022
(1.619)
Observations 3,935
R-Squared 0.253

 

 

Note: The distribution of out of pocket costs is very skewed to the right. To predict missing price data I
exclude the top 1% of observed out of pocket costs. Also, out of pocket cost is not predicted for those
covered by Medicaid; due to the elimination of cost sharing under Medicaid/SCHIP a $0 out of pocket cost
is assumed.

156

Table A3.3 Means and Standard Deviations of Independent Variables by

Maternal Employment

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Employed Non-Employed
Received Visit in Care 0.33 0.26
Interval (0.47) (0.44)
$9.92 $5.55
Out of Pocket Cost (18.64) (15“)
Household Income $35,712.09 $38,038.51
(31,189.70) (34,465.60)
Wa e $12.54 $8.50
g (8.44) (3.78)
. 0.25
Salaried (0.43)
Paid Sick Leave (3'33)
Sick Leave used to See 0.48
Doctor (0.50)
. . 0.63
Paid Vacation Leave (0. 48)
FMLA Eligible (3'32)
# Employees (in 100s) (: 2;)
2 . 5.64
# Employees (in 100s) (9.39)
. 0.09
Union Member (0.28)
. 0.10
Managerial (0.30)
. 0.25
Professmnal (0.43)
0.14
Sales (03 5)
. 0.20
Clerical (040)
Production (3'32)
Other Occupation (3(1)?)
Self Employed (3'33?)
Federal Govt Employee (3");
State Govt Employee (8'33)
More than 1 Job (3'23)
3.20
Tenure (3 78)
2 24.55
Tenure (5297)
Part Time (< 35 Hrs per 0.32
Week) (0.47)

 

 

 

157

 

Table A3.4 Means and Standard Deviations of Independent Variables by
Maternal Employment (Continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Over Time (>40 Hrs per 0.12
Week) (0.33)
% Work Days Missed Last 0.08
Round (0.22)
Visits Received Last Round 0.43 0.34
(0.70) (0.63)
Private Employer Provided 0.56 0.25
Insurance (0.50) (0.43)
Private Non-Group 0.01 0.01
Insurance (0.10) (0.1 1)
Medicaid 0.34 0.62
(0.47) (0.49)
UCP is General Practitioner 0.1 1 0.12
(0.31) (0.33)
UCP is OB/GYN 0.00 0.00
(0.05) (0.06)
UCP is an Individual Doctor 0.25 0.28
in Own Office (0.43) (0.45)
UCP is Group Practice (No 0.62 0.64
Individual Doctor) (0.49) (0.48)
Diagnosed with Serious 0.22 0.21
Condition (0.36) (0.37)
Subjective Health Rating = 0.52 0.49
Excellent (0.50) (0.50)
Subjective Health Rating = 0.29 0.27
Very Good (0.46) (0.44)
Subjective Health Rating = 0.03 0.04
Fair (0.17) (0.19)
Subjective Health Rating = 0.00 0.01
Poor (0.07) (0.08)
Mother Has Less than High 0.16 0.42
School Education (0.36) (0.49)
Mother has GED 0.05 0.05
(0.22) (0.21)
Mother has Bachelor‘s 0.19 0.09
(0.39) (0.28)
Mother has Graduate Degree 0.08 0.02
(0.27) (0.14)
Number of Children Mother 1.71 2.21
has Ever Had (1.25) (1.45)
Number of Children Ever 1.28 1.59
Had Currently Under Age 6 (0.98) (1.15)
Mother’s Age 28.46 26.82
(5.83) (6.34)
Mother’s Marital Status 0.68 0.64
(0.47) (0.48)
White 0.75 0.81
(0.43) (0.39)
Black 0.18 0.13
(0.39) (0.34)
Hispanic 0.25 0.46
(0.43) (0.50)

 

158

 

Table A3.4 Means and Standard Deviations of Independent Variables by
Maternal Employment (Continued)

 

 

 

 

 

 

 

 

 

Non-English 0.09 0.29
(0.29) (0.45)
Urban 0.80 0.80
(0.40) (0.40)
NE 0.15 14.07
(0.36) (0.35)
MW 0.25 15.05
(0.43) (0.36)
S 0.37 0.37
(0.48) (0.48)
Observations 8,706 9,687

 

 

159

Conclusion

Summary of Findings

This dissertation examined the actual and expected childbearing, employment
behavior and compliance with well-baby care visit schedule the relationship between
access to paid and unpaid leave and other job attributes and these behaviors. Chapter 1
provided an examination of the cross-cohort trends in expected and actual childbearing
and compared the characteristics of women’s chosen occupations prior to the birth of
their first child by their childbearing expectations. Previous studies that examine subsets
of the cohorts studied in Chapter I conclude expectations are a fairly good approximation
of actual fertility at the cohort level. However with few exceptions, the cross-cohort
trends among women born in the 19305 and 19405 through the 19705 do not approximate
the cross-cohort trend in actual fertility. While the proportion of women who expected to
become mothers was increasing from the 1947 through 1972 cohorts, the trend in actual
motherhood was flat or even negative. In nearly all cohorts expectations underestimate
the likelihoods of postponing first births and having more than three children. The only
group among whom the trend in expected fertility appears to lead the trend in actual
fertility is college educated women born after 1950.

Chapter 1 also compared occupational characteristics by expected and actual
motherhood and first birth timing as a first step towards understanding relationships
between childbearing expectations and early employment behavior. For all occupational
characteristics considered, the largest differences existed between childless women and
mothers. I found no evidence of systematic differences between childless women’s

occupations based on their birth expectations. From this finding I infer the differences in

160

mothers’ and childless women’s occupations are due to sorting at and around the time of
the first birth rather than early in women’s careers.

Chapter 2 directly examines job changing and labor force exits among women up
to 1 year before pregnancy and eighteen months after. Although Chapter 2 focuses on
job quitting rather than returns to work it is closely related to the return to work literature.
Most return to work studies have used the NLSY 1979 and have not examined job
changing behavior; women in the NLSY 1997 are not yet far enough into their
childbearing years to support this kind of analysis. Chapter 2 uses the Medical
Expenditure Panel Data Surveys from 1996 through 2005, which constitute a more recent
and nationally representative sample of expecting and new mothers. Based on this
sample, I estimate approximately 55 percent of mothers remain in the jobs they held 1

year before pregnancy through eighteen months after the birth, 16 percent change jobs

and 35 percent leave the labor force.‘2 Compared to Klerman and Leibowitz’s (1999)
estimates of job continuity among women who had children during the 1980s, fewer
women who had children between 1993 and 2005 quit their jobs. Most labor force exits
are clustered within the three months preceding and three months following the birth but
job changes occur prior to the pregnancy, in the first trimester and after the baby is
twelve weeks old. Very few women change jobs with the months immediately preceding
and following the birth. This finding indicates future studies need to consider a wider
time frame than is customary in existing literature in order to capture all relevant

employment behavior to correctly classify women as having stayed in their job, changed

 

12 . . . . . . . .
Estimates are based on surv1val probabilities and use sample weights to generate estimates of population
parameters. They do not sum to 100% due to estimation error.

161

jobs or left the Iaborforce and more plausibly assume the jobs observed are not a selected
sample.

As expected, wages in the prebirth job decrease the likelihood of quitting to
change jobs or leave the labor force. In general, differences in employer provided
benefits, including paid leave, explain more of the variation in quit behavior prior to the
birth than after whereas differences in occupation, sector of employment (public or
private) and wages appear to matter most for quit decisions after the birth.

In Chapter 2 the overall evidence for sorting to obtain paid leave is weak. While I
do find women without paid vacation leave are more likely to change jobs prior to the
birth, I find no significant differences among women with and without sick leave.
Furthermore, having paid vacation does not appear to deter quits after the birth. In my
descriptive analysis of old and new jobs among job changers, I do find women were more
likely to acquire paid leave of either type (sick or vacation) than to lose it. However, they
also moved into higher wage jobs that were more likely to offer health insurance and
retirement plans. In short, the possibility that women changed jobs for other reasons and
happened to obtain paid leave cannot be ruled out and the results as a whole suggest
wages are a more important determinant of quit behavior than paid leave.

In Chapter 3 I estimated compliance with the American Academy of Pediatrics
recommended schedule of well-baby care visits and attempted to explain why compliance
rates are so low. Well-baby care is preventive care for children from birth through two
years of age (when it becomes well-child care) and includes vaccinations, health
screenings, physical examination and anticipatory guidance for parents. An estimated 52

percent of infants do not receive their first well-baby care visit at age one month and

162

compliance rates fall from there. Out of 8 recommended visits over the first 18 months,
the average baby receives just over 2. This is somewhat surprising considering the low
level of cost sharing under both private and public health insurance plans and the high
level of insurance coverage among very young children. Thus, I examine the
relationships between well-baby care, maternal employment and paid and unpaid leave
available to mothers to see if time constraints and competition between work and taking
one’s child to the doctor can explain low compliance rates.

Even after controlling for maternal employment and access to paid and unpaid
leave, copays do reduce the likelihood of receiving a given recommended visit. A 1%
increase in copays, which would be about $1 on average, reduces the likelihood of
receiving a given visit by 19.6 percent. Among children whose mothers work outside the
home, mother’s access to paid vacation, sick or F MLA leaves does increase the
likelihood of receiving a visit but only in certain types of jobs. For example, effects of
FMLA are large among mothers in long hours, low wage jobs but insignificant across all
jobs. Paid sick and vacation leave matter most for women in professional, clerical and
production occupations. In other occupations mothers’ access to paid leave is negatively
related to receipt of well-baby care. These findings suggest paid leave may act as a
rationing policy rather than a flexibility enabling policy in some jobs. Additionally, I
find mothers who enter or leave the labor force are less likely to take their children to
visits during recommended care intervals when they are employed. However, cross-
sectional estimates of relationships between maternal employment and well-baby care
were inconclusive.

Implications for Future Research

163

Job changing behavior has largely been ignored in the return to work literature.
The findings in Chapters 1 and 2 indicate job continuity among women of childbearing
age and new mothers may be more complex than a simple model of compensating wage
differentials and lifecycle labor supply would imply. In Chapter 1 I find no evidence of
sorting into occupations with more family friendly benefits and characteristics prior to the
birth and my analysis of expectations suggests women may not have enough information
to accurately anticipate their fertility and engage in efficient sorting. But there could still
be important sorting behavior occurring within occupations. In Chapter 2, using
availability of paid and unpaid leave and other attributes of women’s pre-pregnancy jobs,
.I find mixed evidence regarding sorting behavior in the months preceding and following
a birth. Availability of paid and unpaid leave do not appear to influence job changing
behavior in the regression analysis but simple tabulations of the characteristics of old and
new jobs are suggestive of sorting into jobs with paid leave. Further analysis of job
changes before and after births as the NLSY 1997 becomes available and with other data
sets which contain information about recent employment and childbearing behavior is
needed to better identify the sorting that is occurring and the influence of family friendly
job characteristics.

Additionally, Chapter 2 demonstrates that the time interval over which
employment behavior is observed may lead to misclassification of employment behavior
and in turn affect the estimated relationships between employment behavior and variables
of interest. Also, relationships between key covariates and employment behavior may
change over the months preceding and following a birth. For example, I find F MLA

leave eligible women are less likely to quit their jobs before giving birth but their quit

164

behavior after the birth does not differ from the quit behavior of women ineligible for
F MLA leave. The distribution of job changes and labor force exits suggests future
research should aim to analyze behavior from at least twelve months before and
preferably twenty months before the birth and be cognizant of the potential for
misclassification when shorter intervals must be used. Furthermore, separately
estimating relationships of interest before and after the birth or over even more targeted
intervals may provide a richer understanding of the interaction between key covariates
and employment behavior.

Findings in Chapter 3 suggest there may be competition between maternal
employment and well-baby care. However, the identification strategy used to estimate
the effect of maternal employment on well-baby care is not ideal. Further examination of
the relationship between compliance with well-baby recommendations and maternal
employment is needed to confirm the results found in Chapter 3. If there does in fact
seem to be competition between maternal employment and well-baby care, one might
wonder about the relationships between employment and other health behaviors including
adults’ compliance with their own preventive care schedule. Timing and adequate
preventive care may help to ward off more serious and costly future medical conditions
and thus lower healthcare costs. Rising healthcare costs are a focal issue in current policy
debates and a challenge to employers attempting to manage benefits costs.

Finally, one of the important barriers to empirically analyzing the relationships between
family friendly workplace policies and practices and any behavior is the lack of detailed
information about workplace policies in most large scale surveys. Indeed, the only

policies I was able to consider in this dissertation were paid vacation and sick leave and

165

FMLA leave. Even then I had information about the availability of paid and sick leave
but not the amount available or rules for accumulation and use. Short of having better
questions in the large scale national surveys, researchers may be able to make use of the
quality of questions available in smaller surveys such as the National Study of the
Changing Workforce, which is still nationally representative and has richer information
regarding workplace policies and practices.
Policy Implications

As stated in the introduction to this dissertation, in the U.S. there are few public
policies aimed at reconciliation of work and family time and no universal entitlement to
paid leave. This contrasts sharply with nearly all other developed countries. In the U.S.
employer policies and practices determine the organization of work and non-work time
with very little intervention from the government or organized labor. This institutional
structure has lead to disparities across jobs and occupations in the availability of paid
leaves and family friendly policies. Yet even among jobs which offer paid leave or in
which workers are eligible for F MLA leave, Chapter 3 suggests the relationships between
paid and unpaid leave and care-giving behavior may vary greatly across jobs. In short, in
the U.S. there is substantial inequity in access to paid and unpaid leaves and may also be
important differences in the terms and conditions of paid and unpaid leave policies
among those who are covered.

Given the disparities in access to and characteristics of paid and unpaid leave
policies, we might expect to see workers with a high demand for these benefits sort into
jobs with the best paid leave policies. The birth of a child and associated acquisition of

time intensive care giving responsibilities should increase the demand for these benefits

166

and, if mobility were costless, should lead to sorting. Yet job changing can be costly in
terms of destruction of specific human capital, mobility costs and forfeiture of any
accrued non-transferable benefits. A rational agent would only change jobs when the
benefits of doing so outweigh these costs. Furthermore, a rational agent with perfect
foresight would choose to pursue a sequence of jobs which maximized utility over her
lifecycle. Thus a woman who knew she was going to have a child in the next five years
should be less likely to take on a job with any sort of deferred compensation scheme or
specific human capital investments if the job has a schedule that will not be workable
after she has her child or if she intends to quit for other reasons. Or a woman who knew
she was going to have multiple children and require repeated labor force absences should
select into an occupation with less skill atrophy. Early sorting behavior need not
contribute to differences in lifecycle earnings between mothers and childless women; it
could help to reduce them.

The fact that cohort level childbearing expectations, and expectations of first birth
timing in particular, were found to have a poor correspondence with actual cohort fertility
suggests a poor correspondence at the individual level as well. If women cannot predict
their future fertility with reasonable accuracy then the employment decisions they make
may be optimal in the short run but suboptimal over the life cycle. Indeed, the fact that
there are no observable differences in the characteristics of occupations chosen by
childless women who expect to become mothers and those who do not only suggests
women are not sorting in to lower wage jobs before having children on based on the
expectations of becoming mothers. It suggests that they are not sorting prior to thebirth

at all. Certainly the results of Chapter 1 are far from definitive and there could be a large

167

amount of sorting that occurs within occupations. But these results do warrant further
investigation of common assumptions about women’s job changing behavior.

If in fact women do not seem to have enough information to accurately anticipate
their fertility and sort efficiently, less diversity in working time policies and practices
across jobs and a broader set of basic entitlements and policies regarding the combination
of work and non-work time might help to alleviate some of the uncertainty women
experience when attempting to plan childbearing and make career decisions. If there is
excess mobility, more workable and stable employment relationships and childcare
strategies should improve children’s well-being, promote gender based pay equity and
increase household incomes.

Findings in Chapter 3 have strong implications for child health and development,
and more broadly management of public health. Public policies already provide
widespread subsidies to reduce the cost of well-baby care and promote compliance with
recommended care schedules. Yet compliance remains appallingly low. Among women
with the longest hours, my findings indicate women with FMLA leave eligibility are
substantially more likely to take their children to recommended visits. Furthermore, I
find women with paid sick or vacation leave and those eligible for F MLA are more likely
to schedule weekday rather than weekend doctors visits than those without paid and
unpaid leave. However, women who indicate their paid sick leave “can be used to see the
doctor” are less likely to schedule weekday appointments and, throughout the analysis,
this response was often negatively related to receipt of care. One explanation for this
finding is women answered the question “must you use your paid sick time to visit the

doctor.”

I68

Based on tabulations in the NSCW (see Introduction Table 1.1) Only 68 percent of
mothers with children under age 5 have any paid time off to care for their own illness and
only 45 have paid time off to care for a sick child. Of those who have paid time off for
their own illness, 23 percent state they do not receive enough. Similarly, of those who
have paid time off to care for a sick child, 17 percent state they do not receive enough. If
paid sick leave is scarce, parents are unlikely to be willing to use it to take a healthy child
to the doctor. Although “intermittent FMLA leave” (use of F MLA to obtain part of a day
off or arrange a reduced schedule for a qualifying purpose) is permitted for FMLA
eligible employees, paid sick and vacation days are often doled out by the day or half
day. Furthermore, employers may require employees to use all banked sick and vacation
time before using FMLA leave. Arguably this creates a disincentive to use because
parents may wish to reserve their sick time for illnesses and their vacation time for
vacations or unforeseen needs for time off. A separate entitlement to FMLA leave would
resolve this disincentive. Finally, under the strictest interpretation of the law, F MLA
does not extend to preventive care or minor illnesses, except as related to pregnancy or in
proximity to the birth.

Although my findings for well-baby care use imply substantial non-compliance
. with recommended well-baby care, full compliance may not be an appropriate policy
goal. Preventive care is often proposed as part of the solution to rising medical care costs
(Hensrud 2000; Fries et al 1993). However, not all preventive care is cost-effective
(improve health enough to justify their cost) and even those which are may result in a net
increase in medical care costs (Russel 1993). For example, Tucker et a1 (1 998) evaluate

the costs and benefits of introducing universal vaccination for rotavirus (the most

169

conunon cause of severe diarrhea in children). Rotavirus vaccine costs $20 per dose and
they estimate the cost of universal immunization would be $289 million and would
prevent 1.08 million cases of diarrhea, 34,000 hospitalization, 95,000 emergency
department visits and 227,000 doctors visits among children age 5 and younger.
Assuming vaccination rates similar to those for DPT (diphtheria, pertussis and tetanus),
The medical costs associated with rotavirus include increased doctor visits, emergency
department visits, hospitalizations and medical costs associated with death. Societal
costs of rotavirus include caregivers’ loss of earnings and lifetime productivity loss due
to deaths. Costs of the vaccine program include both administrative costs and the $20
cost per dose. In total, they estimate the program would result in a net reduction of $296
million in societal costs and a net gain of $107 million in medical costs.

There are few studies that estimate the cost-effectiveness of well-baby care or
other preventive care visits, probably because anticipatory guidance and other less
objective aims of care are not as easy to quantify as immunization and the incidence rate
of unhealthy behaviors in the absence of anticipatory guidance is difficult to measure.
Using the 1992 Pennsylvania Port Authority Transit strike as a source of identification,
Evans and Lien (2005) attempt to provide estimates of the causal effect of prenatal care
on pregnancy and infantoutcomes (birth weight, gestation, maternal weight gain and
smoking) among black inner-city women. The find missed visits early in pregnancy
negatively affect some outcomes but late visits appear to be less influential. Hoekelman
(1975) compares the gain in maternal knowledge, level of maternal satisfaction, degree of
compliance and attainment of planned health supervision between infants who receive 3

and 6 well-baby visits in the first year of life and finds no significant differences in these

170

measures between the two groups. Gilbert et al (1984) conduct a similar study in Canada
where they compare infants who were assigned to receive 10 well-baby visits over the
first two years of life (the current recommended number in Ontario) and 5 well-baby
visits. They find no differences in the incidence of illness or prevalence of undetected
abnormalities between the two groups. However, babies in the 5 visit group received an
average of 4.77 on-time visits. My estimates suggest American children receive 2.03
visits on average by age 12 months and 2.25 visits by age 18 months, including
unscheduled visits. No study has compared outcomes for reduced care at that level.
Current and Future Public and Workplace Policy Trends

Issues involving working families and children continue to be at the forefront of
policy discussions in the U.S. and other OECD countries. Yet other OECD countries
have more policies currently in place and work-family issues receive more attention in
current policy debates. Japan and Korea are especially concerned with the reconciliation
of work and family life because of their extremely low fertility rates. In countries with
higher fertility, concerns over female labor force participation tend to receive more
attention than levels of fertility but both are important issues. The EU, for example, has
set a female labor force participation rate target of 60% in each member state by 2010
(OECD 2007).

European countries are currently debating individualization of parental leave
benefits with the specific aim of promoting gender equity at home and in the labor force
(OECD 2007). Since paid parental leaves are generally state financed in European
countries, many countries grant a certain portion of maternity leave to mothers and

additional amounts of leave are transferrable between parents but fathers have no separate

I71

right to parental leave. Most Scandinavian countries have already adopted individualized
benefits. Iceland, for example, has individualized rights to parental leave; mothers and
fathers are each separately entitled to three months of parental leave and jointly entitled
to an additional three months which may be divided in any way the couple chooses
(Rostgaard 2002). If individualization reduces the differences in leave taking behavior
between mothers and fathers it may reduce gender inequality in the labor force.

Scandinavian countries have the most generous family policy in the world and in
order to finance those policies they also have some of the highest tax rates (OECD 2007).
Among EU. countries, the policy regime and fiscal perspective in the U.K. is probably
the most similar to the U.S. Yet even in the U.K., all working mothers are entitled to job
protected leave of up to 26 weeks and 60% are entitled to some form of maternity
payment based on their work history (Hudson et al., 2004) and pay is replaced at 90% for
the first six weeks (OECD 2007).

In 2003, the U.K. introduced “right to request” legislation which granted
employees with young or disabled children the right to request “flexible working”
(Department for Business Enterprise and Regulatory Reform 2008). Flexible working
requests may include but are not limited to changes in hours of work, changes in times
when required to work or working from home (Advisory, Conciliation and Arbitration
Service 2007). When presented with a request, employers are required to consider it and
either agree to the proposed work schedule changes or provide a business rationale for
refusal. Employees have protection from reprisal or dismissal for filing a request, the

right to appeal refusals, and in some cases the right to bring a refusal before a tribunal.

172

 

Although the U.S. has historically provided significantly less public support for
the reconciliation of work and family time than most other developed countries, recent
policy changes and current initiatives may begin to close this gap. For example, in 2002
California passed the first paid family leave law in the nation. The law went into effect in
2004 and provides up to 6 weeks of leave with 55 percent pay up to a maximum of $728
per week to both male and female employees who have a new child either by birth or
adoption or need to care for a seriously ill family member (Milkman and Appelbaum
2004). These benefits are entirely employee financed through Califomia’s existing State
Disability Insurance program (Milkman and Appelbaum 2004).

The Alfred P. Sloan Foundation has sponsored a National Initiative on Workplace
Flexibility. The goals of this initiative is to make workplace flexibility “the standard of
the American workplace” and the Sloan Foundation has sought to accomplish this goal
by providing funding for projects at the national, state and local level which advance
flexible work arrangements (Christensen 2004). The Workplace Flexibility 2010 policy
initiative, which one of the projects funded by the Sloan Foundation, has set forth the
ambitious goal of creating consensus based national policy solutions in the areas of
flexible work arrangements, time off and career exit/reentry by the year 2010 (Workplace
Flexibility 2010, 2004).

In the 2008 election, Democratic presidential candidate Barack Obama has laid
out a platform which would introduce new policies and expand the FMLA to address
work-family issues. He states he would work to enact an employer mandate that would
require the provision of seven paid sick days per year. A similar proposal was previously

introduced as “The Healthy Families Act” by Senator Edward Kennedy (D-Mass.) and

173

Representative Rosa DeLauro (D-Conn.) in the 108th, 109th and 110th Congresses but

never made it to the floor (GovTrack.us 2007). Furthermore, Obama says he would
support the expansion of FMLA, which currently covers businesses with 50 or more
employees, to cover businesses with 25 or more and to extend coverage for more
purposes including time parents choose to spend participating in their children’s
academic activities. (Obama’08 2008). No mention is made of providing pay during
F MLA leave but Obama does propose federal funds would be allocated to assist states
with the establishment of paid-leave systems, presumably similar to the California
system.

Although unionization rates in the U.S. are generally quite low, the labor
movement has taken up the issue of workplace flexibility and achieved some important
benefits for their members. For example, in 1999 UAW negotiations with the Big Three
automakers established the Alliance for Children and Working Families which included
funding for training of child care providers, summer camp, after-school programs, and
back-up child care (Lazarovici 2000). District 31 of the American Federation of State,
County and Municipal Employees (AFSCME) negotiated 1 day of work at home per
week for new parents with children less than one year old and hotel and members of
Hotel and Restaurant Employees Local 2 in San Francisco negotiated a child and elder
care fund and flexible paid-time-off policy in their 1994 contract (Lazarovici 2000).

Many private sector employers are choosing to provide their employees with
policies that help to reconcile work and family roles as part of an attraction and retention
strategy. However, those with the most generous policies disproportionately employ

highly paid professionals. This tendency leads to disproportionate access to flexibility

I74

throughout the labor market by pay grade. For example, companies who made Working
Mother Magazine’s 100 Best list for fifteen years or more were primarily large financial
companies, including Bank of America and Citi, pharmaceuticals like Merck & Co. and
Procter and Gamble and high tech corporations including IBM and Hewlett-Packard. Yet
in 2005 IKEA, as large Swedish owned retail store, made a notable appearance on the list
as one of the few companies in the retail industry to offer medical and dental coverage to
all employees, including part time workers (Business Wire 2005).

Table C 1 compares the family benefits provided in the 2007 Working Mother
Magazine’s 100 Best companies to the national availability of benefits as measured in the
2007 benefits survey of the Society for Human Resource Management members
(SHRM). The membership of SHRM is disproportionately made up of human resource
managers and executives from larger companies and thus the differences in Table Cl ,
although striking, likely understate the true differences between benefits available in jobs
at the100 Best and the average U.S. job.

Future advances in public and employer flexible working-time policies will
depend on many factors, perhaps the most important of which is the strength of the
economy. Flexibility can be costly to employers, employees and taxpayers. Even in
Europe, where willingness to pay appears to be quite high, tension over the costs family
friendly benefits impose upon employers is an important public concern and has lead to
“flexicurity” initiative. F lexicurity is a policy strategy aimed at enhancing the flexibility
of labor markets, work organizations and labor relations while ensuring employment and
income security for workers (European Comission 2008). More directly, flexicurity

would allow employers to more freely hire labor than under existing life-long

I75

 

employment regimes in order to compete in changing global markets. The “security”
component of flexicurity is the govemment’s commitment to providing worker retraining
and income support to ensure employment and income security in the face of lessoned
job security. Flexicurity is a highly contested issue in the EU. and in particular between
employee and employer organizations. But, since many family benefits are delivered
through public entitlements and funded with public dollars rather than through
employment contracts, reduced job security may not have much effect on a new mother
or father’s ability to take leave and arrange for on-going care. The U.S., on the other
hand, faces many of the same economic pressures but has a much less developed social
safety net. Furthermore, unlike in Europe, family benefits are almost entirely determined
by one’s current employment contract and eligibility for F MLA leave is contingent upon
having worked at least one year full-time with one’s current employer. In this regime,
increased job insecurity may lead employees to experience substantial changes in
working-time arrangements and access to flexible policies as they transition through jobs
and more workers may find themselves in the periphery of the labor market without
access to FMLA and other benefits contingent upon continuous, full-time job tenure.
This possibility, along with the marked inequalities under the current employer provided
benefit regime, point to the need for public entitlement to flexible working-time
arrangements.

Furthermore, while the focus of this dissertation has been the need for work and
family time reconciliation among new parents and mothers in particular, parents are not
the only persons in need of flexibility. As the baby boom generation ages a large

proportion of the working population will take on elder-care duties that will need to be

176

 

reconciled with their work schedules. Also, the baby boomers themselves may seek a
more flexible transition into retirement. As people live longer, the pursuit of portfolio
careers (careers which include various jobs and work arrangements to suit each stage of
the lifecycle) may increase (Platman 2004). Indeed there is some suggestive evidence to
indicate paid work may actually increase longevity. In a study of cohorts affected Social
Security benefit reform Snyder and Evans (2006) find cohorts who received lower benefit
payments had significantly lower mortality rates. One explanation for this finding would
be these cohorts had to wait longer to retire and work may have actually improved
longevity.

Future research can help to inform work and family reconciliation policies
through better understanding of the choices and challenges women face when making
employment and childbearing decisions over the lifecycle. A key goal of this research
should be to further investigate the role of workplace policies and practices in shaping
those decisions. Furthermore, although much of the existing literature and this
dissertation focused exclusively on women and childbearing, the influence and
importance of flexible working-time policies and practices extends to fathers and other
care-givers as well. Since the dual earner family has become the standard and since
demographic trends will lead many American workers to take on elder-care
responsibilities in the future, new studies of the effects of workplace and public policies
on fathers and other care-givers are needed.

In conclusion, there is vast need for better reconciliation of work and family time
in the U.S. and to date public policy intervention is minimal. The conflict between work

and family time may contribute to gender inequities in the labor market and negatively

177

affect the health and development of young children. The results of this dissertation
suggest access to paid leave may help women to maintain job matches during
childbearing years and improve health outcomes for young children by encouraging
mothers to take their children to well-baby care. The findings invite further investigation
of the extent to which women are able to optimally plan their childbearing and careers
over the lifecycle and further analyses of flexible working—time policies and practices and

their effects on fathers and other care-givers as well as mothers.

I78

Table C1 Comparison of Flexible Working-Time Policies at the Working Mother
Magazine “100 Best Companies” and the Average Company1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100 Best Average
F lextime 100% 58%
Telecommuting 100% 3 3%
Child-Care Resource and 98% 74%
Referral
Job-Sharing 98% 20%
Lactation Program/ 98% 26%
Designated Area
Compressed Work Week 97% 38%
Elder-Care Resource and 97% 22%
Referral
Prenatal Program 97% 70%
Adoption Assistance 91% 20%
Stress-Reduction Program 88% 15%
Paid Adoption Leave 75% 20%
Parental Leave Beyond 73% 27%
FMLA
Paid Paternity Leave 69% 17%
On-Site Child-Care 53% 6%
Health-Care Insurance for 99% 41%
Part-Time Workers

 

 

 

Source: Adapted from Working Mother Magazine “National Snapshot: The Best vs. the Rest.” 1 July 2008.
<http://www.workingm other.com/web?service=vpage/ 1046>

l . . .
Average company statistics are based on surveys conducted by the Soc1ety for Human Resource
Management.

179

 

References

Acton, Jan Paul. Demand for Health Care When Time Prices Vary More than Money
Prices. New York. Rand Institute, 1973.

Altonji, Joseph G. and Rebecca M. Blank. “Race and Gender in the Labor Market.”
Handbook of Labor Economics. Eds. Orley Ashenfelter and David Card.
Elsevier, 1999. 3,143 — 3,259.

Averett, Susan L., H. Elizabeth Peters and Donald M. Waldman. “Tax Credits, Labor
Supply, and Child Care.” The Review of Economics and Statistics. 79(1997): 125
— 135.

Averett, Susan L. and Leslie A. Whittington. “Does Maternity Leave Induce Births?”
Southern Economic Journal 68 (2001): 403 — 417.

Baber, Kristine M. and Patricia Monaghan. “College Women’s Career and Motherhood
Expectations: New Options, Old Dilemmas.” Sex Roles 19(1988): 189 — 203.

Bailey, Martha. “More Power to the Pill: The Impact of Contraceptive Freedom on
Women’s Life Cycle Labor Supply.” Quarterly Journal of Economics 121
(2006): 289 -— 320.

 

Baum, Charles L. “Does Early Maternal Employment Harm Child Development? An
Analysis of the Potential Benefits of Leave Taking.” Journal of Labor Economics
21(2003): 409 — 448.

Ben-Porath, Yoram. “The Production of Human Capital and the Life Cycle of Earnings.”
The Journal of Political Economy 75(1967): 352 — 365.

Berg, Peter, Eileen Appelbaum, Tom Bailey and Arne L. Kalleberg. “Contesting Time:
International Comparisons of Employee Control of Working Time.” Industrial
and Labor Relations Review 57(2004): 331 — 349.

. Berger, Lawrence M. and Jane Waldfogel. “Maternity Leave and the Employment of
New Mothers in the United States.” Journal of Population Economics 17(2004):
1432 — 1475.

Berger, Lawrence M. Jennifer Hill and Jane Waldfogel. “Maternity Leave, Early
Maternal Employment and Child Health and Development in the US.” The
Economic Journal 115(2005): 29 —- 47.

Bianchi, Suzanne M. and Daphne Spain. “Women, Work and Family in America.”
Population Bulletin 51 (1996): 2 — 46.

I80

Blakemore, Arthur E. and Stuart A. Low. “Sex Differences in Occupational Selection:
The Case of College Majors.” The Review of Economics and Statistics 66(1984):
157 — 163.

Blank, Rebecca M. “The Role of Part-Time Work in Women’s Labor Market Choices
Over Time.” The American Economic Review 79(1989): 295 - 299.

Blau, Francine D. and Lawrence M. Kahn. “Swimming Upstream: Trends in the Gender
Wage Differential in the 1980s.” Journal of Labor Economics 15 (1997): 1 — 42.

 

. “Changes in the Labor Supply Behavior of Married Women: 1980 — 2000.”
Journal of Labor Economics 25 (2007): 393 — 438.

Blau, Francine D. “Trends in the Well-Being of American Women, 1970 — 1995.”
Journal of Economic Literature 36(1998): 112 — 165.

Bowler, Mary. “Women’s Earnings: An Overview.” Monthly Labor Review
December(1999): 13 — 21.

 

Brandth, Berit and Elin Kvande. “Flexible Work and Flexible Fathers.” Work,
Employment and Society 15(2001): 251 — 267.

. “Reflexive Fathers: Negotiating Parental Leave and Working Life.” Gender
Work and Organization 9(2002): 186 — 203.

 

Braveman, P. C. Miller, S. Egerter, T. Bennett, P. English, P. Katz and J. Showstack.
“Health Service Use Among Low-Risk Newborns after Early Discharge with and

without Nurse Home Visiting.” Journal of the American Board of Family
Practitioners Jul-Aug(1996): 245 — 260.

 

Brewster, Karin L. and Ronald R. Rindfuss. “Fertility and Women’s Employment in
Industrialized Nations.” Annual Review of Sociology 26(2000): 271 - 296.

Buchmueller, Thomas C. and Robert G. Valletta. “The Effect of Health Insurance on
Married F emale. Labor Supply.” The Journal of Human Resources 34(1999): 42 —
70.

Budd, John W. and Angela M. Brey. “Unions and Family Leave: Early Experience Under
the Family and Medical Leave Act.” Labor Studies Journal 28(2003): 85 — 105.

Bumpass, Larry L. and James A. Sweet. “Patterns of Employment Before and After
Childbirth.” DHEW Publication No. (PHS) 80-1980. U.S. Department of Health,
Education and Welfare, 1980.

. “National Estimates of Cohabitation.” Demography 26(1989): 615 — 625.

 

181

Bumpass, Larry L., James A. Sweet and Andrew Cherlin. “The Role of Cohabitation in
Declining Rates of Marriage.” Journal of Marriage and the Family 53 (1991): 913
— 927.

“IKEA U.S. Named a 2005 Working Mother 100 Best Company by Working Mother
Magazine.” Business Wire 22 Sep. 2005. 15 June 2008. <http://www.allbusiness.
com/labor-employment/working-hours-pattems-flextime/S073 1 54-1 .html>

Byrd, Robert S., Robert A. Hoekelman, and Peggy Auinger. “Adherence to AAP
Guidelines for Well-Child Care Under Managed Care.” Pediatrics 104(1999): 536
— 540.

Cable, Daniel M. and Timothy A. Judge. “Pay Preferences and Job Search Decisions: A
Person-Organization Fit Perspective.” Personnel Psychology 47(1994): 317 — 348.

Campbell, James R., Peter G. Szilagui, Lance E. Rodewald, Cynthia Doane and Klaus J.
Roghmann. “Patient-Specific Reminder Letters and Pediatric Well-Child-Care
Show Rates.” Clinical Pediatrics 33(1994): 268 — 272.

Cantor, David, Jane Waldfogel, Jeffrey Kerwin, Mareena McKinley Wright, Kerry Levin,
John Rauch, Tracey Hagerty, Martha Stapleton Kudela. “Balancing the Needs of
Families and Employers: Family and Medical Leave Surveys.” Washington DC,
U.S. Department of Labor. Office of the Assistant Secretary for Policy, 2001.

Cherlin, Andrew J. Marriage, Divorce, Remarriage. Harvard University Press, 1992.

Coffey, Rosanna M. “The Effect of Time on the Demand for Female Medical Care
Services.” Diss. Southern Methodist University, 1980.

. “The Effect of Time Price on the Demand for Medical-Care Services.” 1112
Journal of Human Resources 18(1983): 407 — 424.

Colle, Ann D. and Michael Grossman. “Determinants of Pediatric Care Utilization.” _T_I_1§
Journal of Human Resources 13(1978): 115 — 158.

Comfort, Derrick, Karen. Johnson and David. Wallace. “Part-time Work and Family-
Friendly Practices in Canadian Workplaces.” The Evolving Workplace Series #6
Statistics Canada and Human Resources Development Canada, Ottawa.

Commission on Family and Medical Leave. “A Workable Balance: Report to the
Congress on Family and Medical Leave Policies.” Washington, DC: Women’s
Bureau, U.S. Department of Labor, 1996.

Daw, Nigel W. “Critical Periods and Amblyopia.” Archives of Ophthalmology.
116(1998):502 - 505.

I82

Desai, Sonalde and Linda J. Waite. “Women’s Employment During Pregnancy and After
the First Birth: Occupational Characteristics and Work Commitment.” American
Sociological Review 56(1991): 551 — 566.

Dwight Geduldig, etc. v. Carolyn Aiello et al. 1974. 41 L.Ed.2d 256, 417 U.S. 484, 94
S.Ct. 2485.

Errnisch, John F. and Robert E. Wright. “Wage Offers and F ull-Time and Part-Time
Employment by British Women.” The Journal of Human Resources 28(1993):
Ill — 133.

European Commission. “Towards Common Principles of F lexicurity: More and Better
Jobs through Flexibility and Secuirty.” Directorate-General for Employment,
Social Affairs and Equal Opportunities, 2007.

Evans, William N. and Diana S. Lien. “The Benefits of Prenatal Care: Evidence form the
PAT Bus Strike.” Journal of Econometrics 125(2005): 207 —- 239.

Facts from EBRI: Flexible Spending Accounts. Washington, DC: Employee Benefit
Research Institute, 2007.

Felmlee, Diane H. “Causes and Consequences of Women’s Employment Discontinuity,
1967 — 1973.” Work and Occupations 22(1995): I67 — 187.

Filer, Randall K. “Male-Female Wage Differences: The Importance of Compensating
Differentials.” Industrial and Labor Relations Review 38(1985): 426 — 438.

Fischhoff, Baruch, Aandrew M. Parker, Wandi Bruine de Bruin, Julie Downs, Claire
Palmgren, Robyn Dawes and Charles F. Manski. “Teen Expectations for
Significant Life Events.” The Public Opinion Quarterly 64 (2000): 189 — 205.

F orgionne, Guisseppi A. and Vivian E. Peeters. “Differences in Job Motivation and
Satisfaction Among Female and Male Managers.” Human Relations 35(1982):
101 — 118.

Freedman, Ronald, Deborah S. Freedman, and Arland D. Thorton. “Changes in Fertility
Expectations and Preferences between 1962 and 1977: Their Relation to Final
Parity.” Demography 17 (1980): 365 — 378.

Freedman, Ronald, David Goldberg and Larry Bumpass. “Current Fertility Expectations
of Married Couples in the United States: 1963.” Population Index 31 (1965): 3 —
20.

Fries, James F ., C. Everett Koop, Carson E. Beadle, Paul P. Cooper, Mary Jane England,
Roger F. Greaves, Jacque J. Sokolov and Daniel Wright. “Reducing Health Care

183

Costs by Reducing the Need and Demand for Medical Services.” The New
England Journal of Medicine 329(1993): 321 — 325.

 

Fuchs, Victor R. “Women’s Quest for Economic Equality.” The Journal of Economic
Perspectives 3(1989): 25 — 41.

 

General Electric v. Martha V. Gilbert et al. 1976. 50 L.Ed.2d 343, 429 U.S. 125, 97 S.Ct.
401.

Gilbert, J. Raymond, William Feldman, Linda S. Siege], Dorothy-Anne Mills, Charles
Dunnett and Greg Stoddart. “How Many Well-Baby Visits are Necessary in the
First 2 Years of Life?” Canadian Medical Association Journal 130(1984): 857 —
861.

Glass, Jennifer. “Job Quits and Job Changes: The Effects of Young Women’s Work
Conditions and Family Factors.” Gender and Society 2(1998): 228 — 240.

Glass, Jennifer L. and Lisa Riley. “Family Responsive Policies and Employee Retention
Following Childbirth.” Social Forces 76(1997): 1401 — 1435.

 

Gomick, Janet C., Marcia K. Meyers and Katherin E. Ross. “Supporting the Employment
of Mothers: Policy Variation Across Fourteen Welfare States.” Journal of
European Social Policy 7(1997): 45 —- 70.

Greenwood, Jeremy, Ananth Seshadri and Guillaume Vandenbroucke. “The Baby Boom
and Baby Bust.” American Economic Review 95 (2005): 183 — 207.

 

Golden, Lonnie. “Flexible Work Schedules: Which Workers Get Them?” American
Behavioral Scientist 44(2001): 1 157 — l 178.

 

Goldin, Claudia. “The Meaning of College in the Lives of American Women: The Past
One-Hundred Years.” NBER Working Papers 4099, National Bureau of
Economic Research, 1992.

 

. “Career and Family: College Women Look to the Past.” NBER Workingjapers
5188, National Bureau of Economic Research, 1995.

. “Exploring the ‘Present Through the Past’: Career and Family Across the Last
Century.” The American Economic Review 87(1997): 396 — 399.

Goldin, Claudia and Lawrence F. Katz. “Career and Marriage in the Age of the Pill.” T_he_
American Economic Review 90 (2000): 461 — 465.

 

. “The Power of the Pill: Oral Contraceptives and Women’s Career and Marriage
Decisions.” Journal of Political Economy 110 (2002): 730 — 770.

 

184

 

Grimshaw, Damian, Kevin G. Ward, Jill Rubery and Huw Beynon. “Organizations and
the Transformation of the Internal Labour Market.” Work, Employment and
Sociegg 15(2001): 25 — 54.

Gronau, Reuben. “The Intrafamily Allocation of Time: The Value of Housewives’
Time.” The American Economic Review 63(1973): 634 — 651.

Grossman, Michael. “Household Production and Health.” Review of Economics of the
Household 1(2003): 331 — 342.

 

Gruber, Jonathan. “The Incidence of Mandated Maternity Benefits.” The American
Economic Review. 84(1994): 622 - 641.

 

Guthrie, Doug and Louise Marie Roth. “The State, Courts, and Maternity Policies in U.S.

Organizations: Specifying Institutional Mechanisms.” American Sociological
Review 64(1999): 41 — 63.

Hagewen, Kellie J. and S. Philip Morgan. “Intended and Ideal Family Size in the United
States, 1970 — 2002.” Population and Development Review 31 (2005): 507 — 527.

Hakim, Catherine. “Lifestyle Preferences as Determinants of Women’s Differentiated
Labor Market Careers.” Work and Occupations 4(2002): 428 — 459.

Han, Wen-Jui and Jane Waldfogel. “Parental Leave: The Impact of Recent Legislation on
Parents’ Leave Taking.” Demography 40(1): 191 — 200.

 

Hayghe, Howard V. “Developments in Women’s Labor Force Participation” Monthly
Labor Review Sept(l 997): 41 — 46.

Hensrud, Donald D. “Clinical Preventive Medicine in Primary Care: Background and
Practice: 1. Rationale and Current Preventive Practices.” Mayo Clinic
Proceedings 75(2000): 165 — I72.

 

Hernandez, Donald J. Trends in the Well-Being of America’s Children and Youth.
Washington DC: U.S. Department of Health and Human Services, 1996.

Hill, E. Jeffrey, Vjollca K. Martinson, Maria Ferris and Robin Zenger Baker. “Beyond
the Mommy Track: The Influence of New—Concept Part-Time Work for

Professional Women on Work and Family.” Journal of Family and Economic
Issues 25(2004): 121 -— 136.

Hoekelman, Robert A, H. George DeCancq, Marsden Fox, Elizabeth McAnarney,
Barbara O’Brien, Charles Olin, Ellen Perrin, Cornelia Porter, Carol Samuelson
and Helen Stutzman. “What Constitutes Adequate Well-Baby Care?” Pediatrics
55(1975): 313 — 326.

185

 

Hoem, Britta. “The Way to the Gender-Segregated Swedish Labour Market.” Gender and
Family Change in Industrialized Countries. Eds. Karen 0. Mason and An-Magritt
Jensen. Oxford: Clarendon Press, 1995. 279 — 296.

Hofferth, Sandra L. “Effects of Public and Private Policies on Working after Childbirth.”
Work and Occupations 23(1996): 378 — 404.

 

Hotz, Joseph V. and M. Rebecca Kilbum. “Regulating Child Care: The Effects of State
Regulations on Child Care Demand and its Cost.” RAND unrestricted draft,
1995.

Hoyer, Donna L., Hsiang-Ching Kung and Bettery L. Smith. “Deaths: Preliminary Data
for 2003.” National Vital Statistics Reports 53.15(2005):

Hudson, Maria, Stephen Lissenburgh and Melahat Sahin-Dikmen. Matemityand
Paternity Rights in Britain 2002: A Survey of Parents IAD Social Research

Division in-house report 131. Department for Work and Pensions, London, UK.
2004.

 

J acknowitz, Allison. “An Investigation of the Factors Influencing Breastfeeding
Patterns.” Diss. RAND Graduate School, 2004.

Jacobs, Jerry A. “Long Term Trends in Occupational Segregation by Sex.” The American
Journal of Sociology 95(1989): 160 — 173.

 

Jacobs, Shelia C. “Changing Patterns of Sex Segregated Occupations throughout the
Life-Course.” European Sociological Review 11(1995): 157 — 171.

Joesch, Jutta M. “Paid Leave and the Timing of Women’s Employment Before and After
Birth.” Journal of Marriage and the Family 59(1997): 1008 — 1021.

Joshi, Heather and PR. Andrew Hinde. “Employment after Childbearing in Post—War
Britain: Cohort-Study Evidence on Contrasts within and across Generations.”

European Sociological Review 9(1993): 203 — 227.

Judge, Timothy A. and Robert D. Bretz. “The Effects of Work Values on Job Choice
Decisions.” Journal of Applied Psychology 77(1992): 261 — 271.

Juhn, Chinhui and Kevin M. Murphy. “Wage Inequality and Family Labor Supply.”
Journal of Labor Economics 15(1997): 72 — 97.

Kalleberg, Arne L. “Nonstandard Employment Relations: Part-time, Temporary and
Contract Work.” Annual Review of Sociology 26(2000): 341 — 365.

. “Flexible Firms and Labor Market Segmentation.” Work and Occupations
30(2003): 154 — I75.

186

 

Kaye, Celia I. “Newborn Screening Fact Sheets.” Pediatrics 118(2006): 934 — 963.

Klerrnan, Jacob Alex and Arleen Leibowitz. “Child Care and Women’s Return to Work
After Childbirth.” The American Economic Review 80(1990): 284 - 288.

. “The Work-Employment Distinction among New Mothers.” The Journal of
Human Resources 29(1994): 277 — 303.

 

 

. “Job Continuity among New Mothers.” Demography 36(1999): 145 — 155.

Korenman, Sanders and David Neumark. “Marriage, Motherhood, and Wages.” IE
Journal of Human Resources 1992(2): 233 — 255.

Kost, Kathryn, David J. Landry and Jacqueline E. Darroch. “The Effects of Pregnancy
Planning Status on Birth Outcomes and Infant Care.” Family Planning
Perspectives 30(1998): 223 — 230.

 

Kviz, F J , CE Dawkins and NE Ervin. “Mothers’ Health Beliefs and Use of Well-Baby
Services among a High-Risk Population.” Research in Nursing and Health
8(1985): 381 — 387.

Lazarovici, Laureen. “Takiong Care of My Family. . .and Do My Job?” American@Work
February, 2000.

Maag, Elaine. “The Disappearing Child Care Credit.” Tax Notes October(2007) 177.

Maisels, M. Jeffrey and Elizabeth Kring. “Early Discharge from the Newborn Nursery —
Effect on Scheduling of Follow-up Visits by Pediatricians.” Pediatrics 100(1997):
72 — 74.

Mare, Robert D. “Five Decades of Educational Assortative Mating.” American
Sociological Review 56 (1991): 15 — 32.

Marsiglio, William and Frank L. Mott. “Does Wanting to Become Pregnant with a First
Child Affect Subsequent Maternal Behaviors?” Journal of Marriage and the
Family 50(1988): 1023 — 1036.

 

Martin, Steven P. “Diverging Fertility among U.S. Women Who Delay Childbearing Past
Age 30.” Demography 37(2000): 523 — 533.

Martin, Teresa Castro and Larry L. Bumpass. “Recent Trends in Marital Disruption.”
Demograplm 26 (1989): 37 — 51.

187

McInemy, Thomas K., William L. Cull and Beth K. Yudkowsky. “Physician
Reimbursement Levels and Adherence to American Academy of Pediatrics Well-

Visit and Immunization Recommendations.” Pediatrics 115(2005): 833 —— 838.

McRae, Susan. “Returning to Work after Childbirth: Opportunities and Inequalities.”
European Sociological Review 9(1993): 125 - 138.

Michalopoulos, Charles, Philip K. Robins and Irwin Garfinkel. “A Structural Model of
Labor Supply and Child Care Demand.” The Journal of Human Resources
27(1992): 166 — 203.

Milkman, Ruth and Eileen Appelbaum. “Paid Family Leave in California: New Research
Findings.” California Center for Population Research. On—Line Working Paper
Series. Paper ccpr—03 1-04 <http://repositories.cdlib.org/ccpr/olwp/ccpr-031-04>,
2004.

Miller- Wohl Company, Inc. v. Commissioner of Labor & Industry 515 F. Supp. 1264(D.
Mont. 1981), vacated, 685 F.2d 1088 (9‘h Cir. 1982).

Mincer, Jacob. “Labor Force Participation of Married Women: A Study of Labor
Supply.” The Economics of Women and Work. Ed. Alice H. Amsden. New York:
St. Martin’s Press, 1980. 41 - 51.

Mincer, Jacob and Soloman Polachek. “Family Investments in Human Capital: Earnings
of Women.” Journal of Political Economy 82(1974): S76 - 8108.

Moore, Maurice J. “Findings from Census-Bureau Surveys.” Predicting Fertilim
Demographic Studies of Birth Expectations. Eds. Gerry E. Hendershot and Paul
J. Placek. Lexington: D.C. Health, 1981. 153 — 168.

Moore, P. and J .T. Hepworth. “Use of Perinatal and Infant Health Services by Mexican-
American Medicaid Enrollees.” The Journal of the American Medical Association
27(1994): 297 — 304.

Morgan, Philip S. “Parity-Specific Fertility Intentions and Uncertainty: The United
States, 1970 to 1976.” Population Association of America 19(1982): 315 — 334.

Mott, Frank L. and Susan H. Mott. “Prospective Life Style Congruence among American
Adolescents: Variations in the Association between Fertility Expectations and
Ideas Regarding Women’s Roles.” Social Forces 63 (1984): 184 — 208.

 

Mustin, H.D. V.L. Holt and FA. Connell. “Adequacy of Well-Child Care and
Immunizations in U.S. Infants Born in 1988.” The Journal of the American
Medical Association 272 (1994): 1111 - 1115.

188

 

National Association of Child Care Resource and Referral Agencies. 2007 Price of Child
Care. Mar. 2008. 6 Jun. 2008. <http://www.naccrra.org/randd/docs/2007_Price_
of.Child__Care.pdf.>

 

National Center for Health Statistics. Healtthnited States, 2007: With Chartbook on
Trends in the Health of Americans Hyattsville, MD: 2007.

Obama ’08: Families 15 Jun. 2008. <http://www.barackobama.com/issues/farnily
/#support-families>

O’Connell, Martin and Carolyn C. Rogers. “Assessing Cohort Birth Expectations Data
from the Current Population Survey, 1971 - I981.” Demography 20 (1983): 369
— 384.

O’Neill, June and Soloman Polachek. “Why the Gender Gap in Wages Narrowed in the
19808.” Journal of Labor Economics 11 (I993): 205 — 228.

Oppenheimer, Valerie Kincade. “Women’s Rising Employment and the Future of the
Family in Industrial Societies.” Population and Development Review 20(1994):
293 - 342.

Organization for Economic Co-Operation and Development. Babies and Bosses:
Reconciling Work and Family Life A Sythesis of Findings for OECD Countries.
OECD, 2007.

Perrons, Diane. “Flexible Working Patterns and Equal Opportunities in the European
Union: Conflict or Compatibility?” European Journal of Women’s Studies
6(1999):391—418.

Peterson, Linda S. “Birth Expectations of Women in the United States: 1973 — 1988.”
National Center for Health Statistics. Vital Health Stat 23. 17 (1995): l — 36.

Platman, Kerry. “Portfolio Careers and the Search for Flexibility in Later Life.” Work,
Emrioyment and Society 18(2004): 573 — 599.

Powers, Mary G. and Joseph J. Salvo. “Fertility and Child Care Arrangements as
Mechanisms of Status Articulation.” Journal of Marriage and the Family 44
(1982): 21 - 34.

Presser, Harriet B. and Wendy Baldwin. “Child Care as a Constraint on Employment:
Prevalence, Correlates, and Bearing on the Work and Fertility Nexus.” IE
American Journal of Sociology 85 (1980): 1,202 — 1,213.

Quesnel-Vallée and S. Philip Morgan. “Missing the Target? Correspondence of Fertility
Intentions and Behavior in the U.S.” Population Research and Policy Review 22
(2003): 497 — 525.

I89

 

Recommendations for Preventive Pediatric Health Care. American Academy of
Pediatrics, 2008.

Rindfuss, Ronald R., S. Philip Morgan and Kate Offutt. “Education and the Changing
Age Pattern of American Fertility: 1963 — 1989.” Demography 33(1996): 277 —
290.

Ronsaville, Donna S. and Rosemarie B. Hakim. “Well Child Care in the United States:
Racial Differences in Compliance With Guidelines.” American Journal of Public
Health 90(2000): 1436 — 1443.

 

Rostgaard, Tine. “Caring for Children and Older People in Europe — A Comparison of
European Policies and Practice.” Policy Studies 23(2002): 51 — 68.

Ruhm, Christopher. “Policy Watch: The Family and Medical Leave Act.” The Journal of
Economic Perspectives 1 1(1997): 175 — 186.

 

. “The Economic Consequences of Parental Leave Mandates: Lessons from
Europe.” marterly Journal of Economics 113(1998): 285 — 318.

. “How Well do Parents with Young Children Combine Work and Family Life.”
(January 2004). NBER Working Paper No. W10247. Available at SSRN:
http://ssm.com/abstract =492360

Russell, Louse B. “The Role of Prevention in Health Reform.” New England Journal of
Medicine 329(1993): 352 - 354.

 

Selden, Thomas M. “Compliance with Well-Child Visit Recommendations: Evidence
From the Medican Expenditure Panel Survey, 2000 — 2002.” Pediatrics
118(2006): 1766 — 1778.

Smith, Kristen E. and Amara Bachu. “Women’s Labor Force Attachment Patterns and
Maternity Leave: A Review of the Literature.” U.S. Bureau of the Census,
Population Division Working Paper No. 32, 1999.

Snyder, Stephen E. and William N. Evans. “The Effect of Income on Mortality: Evidence
from the Social Security Notch.” The Review of Economics and Statistics
88(2006): 482 — 485.

Stier, Haya, Noah Lewin-Epstein, and Michael Braun. “Welfare Regimes, Family-
Supportive Policies, and Women’s Employment along the Life-Course.”

American Journal of Sociology 106(2001): 1731 — 1760.

Stolzenberg, Ross M. and Linda J. Waite. “Age, Fertility Expectations and Plans for
Employment.” American Sociological Review 42 (1977): 769 — 783.

190

 

The World Fact Book. Central Intelligence Agency. 15 Jun. 2008. <https://www.cia.gov/
library/publications/the-world-factbook>

Trent, Katherine and Kyle Crowder. “Adolescent Birth Intentions, Social Disadvantage,
and Behavioral Outcomes.” Journal of Marriage and the Family 59 (1997): 523 —
535.

Trzcinski, Eileen and William T. Alpert. “Pregnancy and Parental Leave Benefits in the
United States and Canada: Judicial Decisions and Legislation.” The Journal of
Human Resources 29(1994): 535 — 554.

 

Tucker, Andrew W., Anne C. Haddix, Joseph S. Bresee, Robert C. Holman, Umesh D.
Parashar, Roger 1. Glass. “Cost-Effectiveness Analysis of a Rotavirus

Immunization Program for the United States.” The Journal of the American
Medical Association 1998(279): 1371 — 1376.

United Kingdom. Department for Business and Enterprise and Regulatory Reform.
Flexible Working: The Right to Request and the Duty to Consider. <http://www.

berr.gov.uk/employment/employment-legislation/employment-guidance/page
35663.html>

United States. U.S. Department of Labor. The Family and Medical Leave Act of 1993.
<http://www.dol.gov/esa/whd/regs/statutes/fmla.htm>

Van Horn, Susan Householder. Women, Work and Fertility, 1900 — 1986. New York:
New York University Press, 1988.

Vistnes, Jessica Primoff and Vivian Hamilton. “The Time and Monetary Costs of
Outpatient Care for Children.” The American Economic Review 85(1995): l 17 —
121.

Waite, Linda J. and Ross M. Stolzenberg. “Intended Childbearing and Labor Force
Participation of Young Women: Insights from Nonrecursive Models.” American

Sociological Review 41 (I976): 235 — 252.

. “Age, Fertility Expectations and Plans for Employment.” American
Sociological Review 42(1977): 769 — 783.

Waldfogel, Jane. “The Effect of Children on Women’s Wages.” American Sociological
Review 62(1997): 209 — 217.

. “Understanding the ‘Family Gap’ in Pay for Women with Children.” fie
Journal of Economic Perspectives 1998(1): I37 — 156.

I91

 

. “The Impact of the Family and Medical Leave Act.” Journal of Policy Analysis
and Management 18(1999): 281 — 302.

 

. “International Policies toward Parental Leave and Child Care.” The Future of
Children 11(2001): 98 — 111.

 

Waldfogel, Jane, Wen-Jui Han, and Jeanne Brooks-Gunn. “The Effects of Early Maternal
Employment on Child Cognitive Development.” Demography 39(2002): 369 —
392.

Waldfogel, Jane, Yoshio Higuchi and Masahiro Abe. “Family Leave Policies and
Women’s Retention after Childbirth: Evidence from the United States, Britain,
and Japan.” Journal of Population Economics 12(1999): 523 — 545.

Wenk, Deeann and Patricia Garrett. “Having a Baby: Some Predictions of Maternal
Employment around Childbirth.” Gender and Socieyy 6(1992): 49 — 65.

Westoff, Charles F. “The Validity of Birth Intentions: Evidence from the U.S.
Longitudinal Studies.” Predicting Fertility: Demographic Studies of Birth

Expectations. Eds. Gerry E. Hendershot and Paul J. Placek. Lexington: D.C.
Health, 1981. 51 — 59.

Westoff, Charles F. and Norman B. Ryder. “The Predictive Validity of Reproductive
Intentions.” Demography 14(1977): 431 — 453.

Williams, C., K. Northstone, R. A. Harrad, J .M. Sparrow and 1. Harvey. “Amblyopia
Treatment Outcomes after Screening before or at Age 3 Years: Follow Up from
Randomized Trial.” British Medical Journal 324(2002): 1549 — 1551.

Williams, Wendy W. “Equality’s Riddle: Pregnancy and the Equal Treatment/Special
Treatment Debate.” NYU Review of Law and Social Change 13(1984): 325 -
380.

Yu, Stella M., Hilary A. Bellamy, Michael D. Kogan, Jennifer L. Dunbar, Renee H.
Schwalberg and Mark A. Schuster. “Factors that Influence Receipt of

Recommended Preventive Pediatric Health and Dental Care.” Pediatrics
110.6(2002): 1 — 8.

192

 

 

 

 

v A. ‘_
Wit-ills; 11h

iii

0

H l 11 Ill

2956 9435