CONTRACEPTION BIOGRAPHIES: WOMEN’S CONTRACEPTIVE METHOD
SWITCHING AND UNION STATUS
Larry St. J. P. Gibbs
A Dissertation
Submitted to the Graduate College of Bowling Green
State University in partial fulfillment of
the requirements for the degree of
DOCTOR OF PHILOSOPHY
August 2014
Committee:
Wendy D. Manning, Advisor
Roudabeh J. Jamasbi,
Graduate Faculty Representative
Karen B. Guzz.
Kara Joyner
Monica A. Longmore
© 2014
Larry Gibbs
All Rights Reserved
iii
ABSTRACT
Wendy D. Manning, Advisor
American women, on average, are having only two children requiring the
effective use of contraception for about 30 years. Relatively few women rely on the same type
of contraception throughout their entire reproductive life course, meaning a large proportion tend
to switch contraception. Prior research has not considered the contemporary context and has
been largely limited to married women. This is problematic given shifts in reproductive
behaviors of women along with changes in union formation in the United States. Using discretetime event history analysis and drawing on data from the NSFG 2006-10, (N=12,279) I analyze
three-year contraceptive switching behavior for single, cohabiting, and married women. Overall,
40% of women switch methods. While single, cohabiting and married women share similar
risks of contraceptive method switching behavior, analyses distinguishing stable users and stable
nonusers indicate that single women are more likely than married to switch contraception and
also to remain as stable users of contraception relative to stable non-use. Given parity is a strong
predictor of method switching, a set of analyses is limited to women at parity zero. However, the
findings indicate that across union status, women at parity zero share similar risks contraceptive
method switching. In terms of contraception used following switching, a greater proportion of
cohabiting women switch to least effective methods, more single women switch to the pill and
condom and married women switch to most effective methods. Results indicate that union status
differs to an extent depending on the originating contraceptive method. Among initial pill users,
cohabiting women compared to married women have lower odds of switching to most effective
iv
methods relative to least effective methods and cohabiting women who are initial least effective
methods users, compared to married women, have lower risk of switching to most effective
methods relative to condom. The results from this dissertation showcase the dynamics of
contraceptive behavior and provide evidence that it is important to distinguish unmarried single
and cohabiting women in the analysis of contraceptive switching behavior. Attention to
contraceptive switching is important in research addressing the correlates and implications of
intended and unintended pregnancies as well as studies of the meanings of cohabitation.
v
This dissertation is dedicated to Noel and Louise Gibbs for their steadfast support and love.
vi
ACKNOWLEDGMENTS
I am sincerely thankful to my astute advisor, teacher and mentor, Wendy D. Manning.
My understanding of sociology has broadened because of you. You always emphasized the
value of hard work and dedication to academic excellence. Regardless of the challenges I faced
during graduate school you were available to offer sound advice, encouragement and
constructive criticism, all of which enabled me to remain focused as a scholar-in-training. Your
support has not gone unnoticed and will always be valued. It is impossible to repay you for all
you have done for me. However, I will do my best to emulate you and mentor students by
conveying all the knowledge learned from you.
To Karen Guzzo, your comments and advice regarding future research emanating from
this project along with your diligence in the dissertation process is appreciated. To Kara Joyner,
thank you for exposing me to cutting edge techniques in demography. I am also grateful for the
opportunity you gave me to assist you on a research project. To Monica Longmore, thank you
for encouraging me and giving me the opportunity to collaborate with you on research projects.
Throughout my course of study you have always taken time to ask about my progress and I am
grateful to you for that. To Roudabeh Jamasbi, thank you for your words of encouragement and
thoughtful comments during my proposal and defense. Sincere thanks to all my committee
members for the time taken to contribute to this dissertation.
To my parents, Noel and Louise, you have been a tower of strength and words cannot
fully express my gratitude. I love you. To Andrew and Lance, I love you both and thank you for
always encouraging me and being the best brothers I could ever ask for. To the friends who
inspired me, too many to name, your prayers and support helped propel me to the finish line. To
Latoya, my best friend and wife, thank you for accompanying me on this journey. Even though
vii
we were miles apart you shared in my struggles and successes each day. Your prayers and
dedicated support enabled me to remain committed throughout my academic journey. I love you
so much – indeed you are a virtuous woman.
Finally, to the One who makes all things possible, Jesus Christ – to Him I give all glory,
honor and praise.
v
TABLE OF CONTENTS
Page
CHAPTER I. INTRODUCTION ..........................................................................................
1
Research Goals………………………………………………………………………
4
Study Contributions…………………………………………………………………
5
Background: Historical Change in Contraceptive Use ..............................................
6
Theoretical Perspective ..............................................................................................
8
Why is Union Status Important? Application of Theoretical Perspective .................
9
Union Formation……………………………………………………………
9
Contraceptive Behavior and Union Status………………………………….
12
Reproductive Behavior and Union Status…………………………………..
13
Review of Literature on Contraceptive Use and Contraceptive Method Switching..
14
Contraceptive Use…………………………………………………………..
14
Contraceptive Method Switching…………………………………………… 15
Contraceptive Method Switching and Union Status:
Married Versus Unmarried Women…………………………………………………
20
Correlates of Contraceptive Method Switching…………………………………….
23
Sociodemographic Characteristics…………………………………………..
23
Age…………………………………………………………………..
23
Race/Ethnicity……………………………………………………….
24
Education……………………………………………………………. 25
Poverty Level………………………………………………………… 26
Background Characteristics………………………………………………….. 27
vi
Family Structure……………………………………………………..
27
Religious Affiliation………………………………………………… 28
Mother’s Education………………………………………………….
29
Fertility Characteristics……………………………………………………… 30
Age at First Sexual Intercourse……………………………………… 30
Parity………………………………………………………………… 30
CHAPTER II: CONTEMPORARY PORTRAIT OF STABLE USERS AND
CONTRACEPTIVE METHOD SWITCHERS………………. ............................................
32
Current Investigation………………………………………………………………..
32
Hypotheses…………………………………………………………………………..
33
Data and Methods…………………………………………………………………… 33
Analytic Sample……………………………………………………………... 34
Dependent Variable………………………………………………………….. 35
Independent Variable………………………………………………………… 36
Sociodemographic Characteristics…………………….……………………
36
Background Characteristics………………………………………………...
37
Fertility Characteristics……………………………………………………… 37
Measure of Time…………………………………………………………….. 38
Analytic Plan………………………………………………………………… 38
Results………………………………………………………………………………
38
Descriptive Results………………………………………………………….
38
Switching Contraceptive Methods: Bivariate and Multivariate Results…….
40
vii
Stability in Use and Nonuse of Contraceptive Methods:
Multivariate Discrete- Time Event History Results………………………… 42
Discussion…………………………………………………………………….........
46
CHAPTER III. CONTRACEPTIVE METHOD SWITCHING AND UNION STATUS AMONG
WOMEN WITHOUT CHILDREN .......................................................................................
49
Current Investigation .................................................................................................
49
Hypotheses….. ..................................................................................................…….
51
Data and Methods ......................................................................................................
51
Analytic Sample .............................................................................................
51
Dependent Variable .......................................................................................
52
Independent Variable .....................................................................................
52
Sociodemographic Characteristics………………………………………….
53
Background Characteristics………………………………………………...
53
Fertility Characteristic……………………………………………….........
54
Measure of Time………………………………………………....................
54
Analytic Plan………………………………………………..........................
54
Results………………………………………………................................................
55
Descriptive Results: Women at Parity Zero…………..................................
55
Switching Contraceptive Methods: Bivariate and Multivariate Results
(Women at Parity Zero) …………................................................................
56
Stability in Use and Nonuse of Contraceptive Methods:
Multivariate Discrete-Time Event History Results
(Women at Parity Zero) …………................................................................
58
viii
Discussion…………................................................................................................
61
CHAPTER IV: UNION STATUS AND CONTRACEPTIVE USE AMONG
CONTRACEPTIVE METHOD SWITCHERS ...................................................................
63
Current Investigation .................................................................................................
63
Hypotheses………………………………………. ............................................……
67
Data and Methods ......................................................................................................
67
Analytic Sample .............................................................................................
68
Dependent Variable .......................................................................................
68
Independent Variable .....................................................................................
68
Sociodemographic Characteristics .................................................................
69
Background Characteristics ...........................................................................
70
Fertility Characteristics ..................................................................................
70
Measure of Time ............................................................................................
70
Analytic Plan..................................................................................................
71
Results………………………………………………………………………………
71
Descriptive Results – All Women Who Switch Contraception .....................
71
Contraceptive Switching Patterns ..................................................................
73
Summary of Contraceptive Switching Patterns .............................................
75
Types of Contraceptive Methods Used among Contraceptive Switchers:
Multivariate Discrete-Time Event History Results........................................
76
Summary of Method Switching Patterns Based on Any Contraception Used at
Start of Observation .......................................................................................
81
ix
Users of the Pill at the Start of the Observation Period Who Switch
Contraception: Multivariate Discrete-Time Event History Results..............
83
Users of Condom at the Start of the Observation Period Who Switch
Contraception: Multivariate Discrete-Time Event History Results………...
84
Users of Least Effective Methods at the Start of the Observation Period Who
Switch Contraception: Multivariate Discrete-Time Event History Results...
85
Summary of Method Switching Patterns Based on Specific Contraception Used at Start of
Observation………………………………………………………………………….
87
CHAPTER V. CONCLUSIONS ..........................................................................................
87
Key Findings ……...............................................................................................….
88
Contributions ............................................................................................................
90
Limitations and Future Research ...............................................................................
92
Summary…….. ..........................................................................................................
97
REFERENCES ...................................................................................................................... 100
APPENDIX A. Main articles on Method Switching in U.S. ............................................... 154
APPENDIX B. Data Construction for Event History Analysis ............................................ 158
x
LIST OF TABLES
Table
Page
2.1
Means (and standard errors) and Percentages of Women by Union Status ............... 118
2.2
Zero Order and Multivariate Discrete-Time Event History Models Predicting
Contraceptive Method Switching .............................................................................. 120
2.3
Zero Order Discrete-Time Event History Models Predicting
Contraceptive Outcomes ............................................................................................ 122
2.4
Multivariate Discrete-Time Event History Using Union Status and
Sociodemographic and Background Characteristics to Predict
Contraceptive Outcomes ............................................................................................ 124
2.5
Multivariate Discrete-Time Event History Using Union Status and Fertility
Characteristics to Predict Contraceptive Outcomes ................................................... 125
2.6
Multivariate Discrete-Time Event History Using Union Status and All
Characteristics to Predict Contraceptive Outcomes ................................................... 126
3.1
Means (and standard errors) and Percentages for Women at Parity Zero by
Union Status..... ......................................................................................................... 128
3.2
Zero Order and Multivariate Discrete-Time Event History Models Predicting
Contraceptive Method Switching among Women at Parity Zero .............................. 130
3.3
Zero Order Discrete-Time Event History Models Predicting Contraceptive
Outcomes among Women at Parity Zero ................................................................... 132
3.4
Multivariate Discrete-Time Event History Using Union Status and
Sociodemographic and Background Characteristics to Predict
Contraceptive Outcomes among Women at Parity Zero……………… ................... 134
xi
3.5
Multivariate Discrete-Time Event History Using Union Status and Fertility
Characteristics to Predict Contraceptive Outcomes among
Women at Parity Zero ................................................................................................ 135
3.6
Multivariate Discrete-Time Event History Using Union Status
and All Characteristics to Predict Contraceptive Outcomes among
Women at Parity Zero ................................................................................................ 136
4.1
Means (and standard errors) and Percentages of Women who Switch
Contraception by Union Status .................................................................................. 138
4.2
Percentage of Women Using Selected Contraceptive Methods Who Switch
Contraception …........................................................................................................ 140
4.3
Percentage of Married Women Using Selected Contraceptive Methods
Who Switch Contraception ........................................................................................ 140
4.4
Percentage of Cohabiting Women Using Selected Contraceptive Methods Who
Switch Contraception ................................................................................................. 141
4.5
Percentage of Single Women Using Selected Contraceptive Methods Who
Switch Contraception ................................................................................................. 141
4.6
Zero Order Discrete-Time Event History Models Predicting Contraceptive
Method Use among Women Who Switch Contraception .......................................... 142
4.7
Multivariate Discrete-Time Event History Using Union Status
and Sociodemographic and Background Characteristics to Predict Contraceptive
Method Use among Women Who Switch Contraception .......................................... 144
xii
4.8
Multivariate Discrete-Time Event History Using Union Status and Fertility
Characteristics to Predict Contraceptive Method Use among Women
Who Switch Contraception ........................................................................................ 145
4.9
Multivariate Discrete-Time Event History Using Union Status and All
Characteristics to Predict Contraceptive Method Use among Women
Who Switch Contraception ........................................................................................ 146
4.10
Multivariate Discrete-Time Event History Models Predicting Contraceptive
Method Use among Initial Pill Users Who Switch .................................................... 148
4.11
Multivariate Discrete-Time Event History Models Predicting Contraceptive
Method Use among Initial Condom Users Who Switch ............................................ 150
4.12
Multivariate Discrete-Time Event History Models Predicting Contraceptive
Method Use among Initial Least Effective Method Users Who Switch .................... 152
1
CHAPTER I: INTRODUCTION
As women make decisions regarding contraceptive practices, they confront the challenge
of what method(s) of contraception, if any, they and or their partners should use. On average,
women are at risk of pregnancy for nearly half of their life span (Hatcher et al. 2007). The
United States total fertility rate (TFR) is 1.9 children per woman (Population Reference Bureau
2012), which means that women, on average, have only two children. For this rate to be
maintained it requires women to effectively use contraception for about 30 years as they remain
at risk for unintended pregnancy (Frost, 2011). It is not likely that the same type of
contraception will be used throughout their life course and so a large proportion of women tend
to practice contraceptive method switching (Grady et al. 2002). Examining contraceptive
method switching is critical to understanding the dynamics of contraceptive behavior.
Much of the research on women’s contraceptive behavior has focused on risk of
unintended pregnancy, consistent use of contraception, method choice and satisfaction, gaps in
method use, and failure rates of contraceptive methods (Frost et al. 2007 a, b; Mosher and Jones
2010; Trussell 2011, Vaughan et al. 2008). While these studies expand the knowledge base of
women’s contraceptive behavior, fewer studies examine the switching of contraception during
women’s reproductive life course. Although 99% of sexually active women have used a form of
contraception (Daniels et al. 2013) maintaining consistent and effective contraceptive use over
time proves very difficult (Frost 2011). Additionally, women today have more options for birth
control (Rocca et al. 2013) yet the availability of these methods are not meeting the needs of
some women as is evident by the high rates of discontinuation (Lessard et al. 2012). All of these
factors lend itself to the examination of contraceptive method switching, which is driven in part
2
by the level and duration of contraceptive effectiveness as well as health risks associated with the
use of contraception (Grady et al. 2002).
Investigating contraceptive method switching among women is important from a public
health viewpoint. In the U.S. young adults (ages 18-24 years), on average, are not consistent
contraception users (Abma et al. 2004) resulting in high rates of unintended pregnancy and
unintended or poorly timed childbirth (Finer and Henshaw 2006). Additionally, women 30 years
and older account for approximately one-third of unintended pregnancies while those ages 35
and older have the greatest proportion of pregnancies ending in abortion compared to younger
women (Finer and Henshaw 2006). Further, studies find that older women (35 years and older)
who had unintended pregnancy resulting in live births are twice as likely to report an unwanted
than a mistimed pregnancy compared to women ages 25-34 years (D’Angelo et al. 2004).
One goal of the Healthy People 2020 initiative (United States Department of Health and
Human Services (HHS) (2012) is to improve pregnancy planning and spacing, and to prevent
unintended pregnancy. This is important as unintended pregnancies continue to be a serious
public health issue (Wildsmith et al. 2010) and financial concern in the U.S. (Monea and Thomas
2011; Trussell 2010; Trussell et al. 2013). Despite this family planning focus, almost half (49%)
of all pregnancies in the U.S. are unintended (Finer and Zolna 2011) and just under half of fertile
sexually active women of reproductive age who unintentionally become pregnant are
contraceptive users (Finer and Henshaw 2006).
Women’s contraceptive behavior allows for more in-depth analysis of their overall
reproductive behavior as contraceptive use is a proximate determinant of fertility (Bongaarts
1978). Research indicates that reproductive behavior of women (e.g., patterns of conception and
childbearing) varies by union status (married versus cohabiting), (Manning 2001; Musick 2002;
3
Raley 2001). Similarly, recent studies find variations in current contraceptive use by union
status among women in the United States (Jones et al. 2012; Sweeney 2010). These findings
indicate that more exploratory research is needed on other types of contraceptive behavior
among women. This dissertation seeks to fill this gap in the literature by providing an
exploratory and descriptive examination of differentials in patterns of contraceptive method
switching among married, cohabiting and single women, which to date has not been explored
thoroughly.
Contraceptive method switching emanates from a body of research on contraceptive
discontinuation. Contraceptive discontinuation, defined as ceasing to use a current method of
contraception, is a common phenomenon (Barden-O’Fallon and Speizer 2011). The motivations
underlying changes and/or discontinuation of contraception have been associated with a host of
factors including, side effects of specific methods, difficulties in using different procedures
(Ramstrom et al. 2002; Moreau et al. 2007), lack of intercourse (Ersek et al. 2011), difficulties
obtaining contraception (Stuart et al. 2013), ambivalence about avoiding pregnancy (Frost et al.
2007a), and changes in life situations (Frost et al. 2007a; Vaughan et al. 2008). There are at least
three types of contraceptive discontinuation associated with inconsistent contraceptive use and
women’s overall reproductive health. They include contraceptive failure, abandonment of
contraceptive use (while in need of contraception) and method switching. Contraceptive failure
is based on the probability of pregnancy during the first year of typical use of any contraception
and highlights the effectiveness of the method among persons who may fail to use methods
correctly or consistently, if any at all (Trussell 2009). Abandonment of contraception is the
stopping of any contraceptive use because of contraceptive failure or for any other reason than
nonexposure or trying to get pregnant (Vaughan et al. 2008). The use of different contraceptive
4
methods in consecutive months is defined as method switching (Grady et al. 2002). While
contraceptive failure and abandonment are two important outcomes to study because they lead to
immediate risk of unintended pregnancy, a key limitation when analyzing these outcomes is that
women’s choice of new method are not taken into account. In contrast, method switching is
important within a woman’s reproductive life course because the rate of switching and type of
switching negatively or positively affects reproductive health outcomes (Steele and Curtis 2003).
On one hand, contraceptive method switching may compromise women’s protection and, as
such, make them more vulnerable to unintended pregnancy (Grady et al. 2002; Vaughan et al.
2008). Conversely, it may improve women’s overall reproductive health if they switch from less
to more effective contraception.
This study uses recently collected data from the National Survey of Family Growth
(NSFG) (2006-10) to examine specifically contraceptive method switching behavior among
women in the United States. The overarching goal of this dissertation is to provide new
descriptive, in-depth and current analysis of the association between contraceptive method
switching and union status of reproductive age women.
Research Goals
This dissertation addresses three sets of research questions focusing on contraceptive
switching. First, in Chapter II, I examine the prevalence and correlates of contraceptive methods
switching according to union status (i.e., married, cohabiting and single). This analysis predicts
the likelihood of women switching versus not switching, net of other sociodemographic,
background, and fertility characteristics. The second research question, examined in Chapter III,
investigates contraceptive method switching among women without children (i.e., women at
parity zero) and examines whether this behavior differs across union status. The final research
5
question in Chapter IV tests if contraceptive method switchers differ in the method of
contraception used and, if so, whether there are differences across union status.
Study Contributions
There are four main major contributions of the current study. First, prior research on
contraceptive method switching typically examines method switching using small, clinical and
disadvantaged samples to establish switching patterns. The value of these results is therefore
limited because of the study population, which is unrepresentative of all women in the United
States. This dissertation examines a contemporary nationally representative sample of
reproductive age women to determine the prevalence and predictors of contraceptive method
switching.
Second, earlier research on contraceptive method switching examined behavior among
married women as the context of childbearing was mostly highlighted through marital unions
(Grady et al. 2002). However, with changes in union formation attributed mostly to increases in
the age at first marriage for men and women (Manning et al. 2014b) and the normative nature of
cohabitation prior to first marriage (Manning 2013), understanding the dynamic contraceptive
process of women becomes even more germane. I investigate contraceptive method switching
by examining union status (married, cohabiting and single) and not marital status (married versus
unmarried) at the start of the observation period.
Third, several studies on different contraceptive behaviors of women have included parity as
a key factor influencing both contraceptive outcomes as well as independent variables (Frost et
al. 2007 a, b; Grady et al. 2002; Jacob and Stanfors 2013; Kavanaugh et al. 2011). To my
knowledge, no research has examined contraceptive method switching behavior specifically
among women who had not yet had children (i.e., where parity is zero at observation). This
6
particular area of enquiry is important as there has been a historical shift towards delayed
childbirths, which has increased the amount of time in which women are at risk for unintended
pregnancy (Kavanaugh et al. 2011).
Fourth, measurement is a significant issue when investigating contraceptive method
switching. Earlier studies have measured contraceptive method switching using different
duration periods of observation (Barber et al. 2011; Frost et. al 2007a, b; Grady et al. 2002;
Manlove et al. 2013). I contribute to the literature on contraceptive method switching by using
the detailed contraceptive method history calendar to estimate contraceptive method switching
for women. I use the contraceptive method history calendar to retrospectively examine
contraceptive use and nonuse of women three years prior to date of interview. I also consider
duration of use and nonuse to be continuous and uninterrupted by periods of sexual abstinence
(see Grady et al. 2002).
Background: Historical Changes in Contraceptive Use
Trends over the last half century show worldwide changes in sexual and reproductive
health behaviors with declines in family size and increases in modern contraceptive use (United
Nation 2011a). The increases in contraception use enable couples to have their desired family
size (Darroch 2013) while, at the same time, facilitating women’s ability to regulate their own
sexual and reproductive health (Freedman and Isaacs 1993; Crossette 2005). Approximately
63% of women of reproductive age practice some form of contraception and, of that figure, 90%
use modern methods, including oral contraception (pills), condoms, injections, intrauterine
device (IUDs), and sterilization (United Nations 2011b). However, there are recent changes in
the types of contraception used by women. According to Darroch (2013), there are increases in
the proportion of women relying on sterilization as well as those using long acting reversible
7
methods (IUD, injectables and implants). Further, there are declines in oral use among women
while there is evidence of negligible change in the proportion of women relying on condoms or
other vaginal methods.
Between 1960 and 1970, there were two dominant patterns of contraception practices in
the United States. The single most important method of contraception among older couples in
their reproductive life course was sterilization while oral contraception (the pill) dramatically
changed the contraceptive behaviors of young women (Westoff, 1972). In the 1980s, favorable
opinion of the pill and condom increased among women in the United States while approval of
IUD decreased (Forrest and Fordyce, 1988). Subsequently, by the late 1980s, the use of pills
declined (Mosher, 1990), the IUD was no longer produced in the United States (Forrest 1986;
Forrest and Fordyce, 1988) and sterilization was a married couple’s primary method of
contraception (Mosher 1990). In 1991, the long-lasting and effective hormonal implant,
Norplant, was introduced in the United States following a battery of clinical tests and a lengthy
approval process (Tanfer et al. 2000). In the same year, an injectable progestin,
depotmedroxyprogesterone acetate (Depo-Provera), was approved and marketed as an effective
method of contraception (Tanfer et al. 2000). The approval and record of success of these new
methods were, in large part, based on documentation of use in developing countries (Affandi et
al. 1987; Basnayake et al. 1988; Liskin and Blackburn 1987). By mid 1990s, however, these two
new contraceptive methods had low levels of adoption (Abma et al. 1997) due to heavy reliance
on sterilization and pills, satisfaction with current contraceptive method, inadequate information
and misconceptions regarding the methods, fear of side effects and cost of methods (Tanfer et al.
2000).
8
Recent advances in contraceptive technology now allow women more autonomy
regarding their sexual and reproductive health (Lessard et al. 2012). Today, women in the
United States are exposed to new modes of delivery of hormonal contraceptives, long-acting
reversible contraception (LARCs) and other forms of contraception including IUDs which are
now available (Kavanaugh et. al 2011). Earlier assumptions of LARCs viewed them as
appropriate for women later in the reproductive life course, but this opinion is changing as the
length of time between intercourse and first birth is widening (Finer et al. 2012). Data show that
LARCs are safe, effective and acceptable for younger adults and women without children
(American College of Obstetricians and Gynecologists, 2011; Centers for Disease Control &
Prevention, 2010). Yet, the current use of LARCs remains low among U.S. women with 6%
(Kavanaugh et al. 2011).
Notwithstanding these historical changes in contraceptive use, along with recent
improvements in methods of contraception, women continue to be exposed to the risk of
unintended pregnancy. Further, many women face difficulties adhering to a strict regimen of
consistent and correct use of contraception (Frost et al. 2007b). While there are a plethora of
factors related to contraceptive use patterns researchers contend that relationship or union
context is a key variable associated with women’s overall reproductive and contraceptive
behavior (Frost et al. 2007a,b; Kost et al. 2008; Sweeney 2010).
Theoretical Perspective
Generally, studies linking relationship or union context to contraceptive behavior support
the life course perspective (Kusunoki and Upchurch 2011; Manlove et al. 2011, 2014). This
perspective posits that an individual’s life is made up of a series of transitions or life events,
which are embedded in trajectories that give them a distinct form and meaning (Elder 1985,
9
1994). Transitions are brief events that mark chronological movement from one state to another
while trajectories are more complex measures which measure broader patterns of events in
individual’s experience in specific life spheres over time (Donnelly et al. 2001).
One of the important factors that determine the shape of an individual’s life course is the
timing of life events (Giele and Elder 1998). This factor is decomposed into four sub-groups
namely, timing (when life transitions occur); sequencing (the order in which events occur);
duration (how long life events last) and prevalence (how many persons experience these
transitions) (Hagestad 1996).
Focusing on the prevalence of these transitions from one
contraceptive method to another is important for understanding contraceptive behavior dynamics
of women throughout the reproductive life course. Another salient life course principle is that
individual behaviors are informed by contexts and relationships within which a person is nested
(Bengsten and Allen 1993; South and Crowder 1999).
In other words, behaviors are not
mutually exclusive of social interactions and relationships within which individuals function.
Therefore, this theoretical approach is applicable regarding the examination of
contraceptive switching behavior among reproductive age adult women in the United States and
considers the context of union in which the behavior occurs as well as the prevalence of the
behavior. The following section uses the life course perspective as a platform for highlighting
differences in contraceptive and reproductive outcomes among women by union status.
Why is Union Status Important? Application of Theoretical Perspective
Union Formation
A key question in this dissertation is how cohabiting women differ in their contraceptive
use patterns than married or single women and underlying this issue is differentials in the
meaning of cohabitation. A life course perspective sets the stage for this question as the place of
10
cohabitation in the American family system is dependent in part on the timing and sequencing of
union formation. Due to the postponements in timing of marriage, there is more time in the early
adulthood life course for the formation of cohabiting unions (including serial cohabitation) and
other nonmarital romantic relationships (Arnett 2004), particularly during the primary
childbearing years (early twenties). The median age at first marriage has increased since 1990s
from age 24 for women and 26 years to almost 29 years for men and 27 years for women in 2011
(U.S. Census Bureau 2011) while the age at cohabitation has not changed and remains about age
22 (Manning et al. 2014). Considerable growth in the prevalence of cohabitation has been
documented, with 41% of women in their early thirties ever cohabiting prior to entering marriage
(Manning 2013). Viewed as the modal path to first marriages, two-thirds of women first married
in the last decade cohabited prior to marriage (Manning 2013). Further, of the co-residential
relationships initiated between 1997 and 2001, 32% were marriages compared to 68%
cohabitation (Kennedy and Bumpass 2008). The majority of young adults view cohabitation as
an acceptable relationship (Scott et al. 2009) and most have or will cohabit at some point in their
lives (Manning 2013). Thus, the cohort change in cohabitation suggests it is an increasingly
important context for family formation likely linked also to contraceptive and reproductive
behaviors.
While cohabitation has increased, the fundamental question is whether it has become an
acceptable union for family formation. While the fertility patterns of cohabiting, married, and
single women are well established (Kennedy and Bumpass 2008; Musick 2002, 2007; Raley
2001; Sweeney 2010), the reasons that union type differentials exist have not been directly
examined. Cohabitation is considered to be a less committed relationship and less
institutionalized than marriage and there are fewer barriers to exit this union (Nock 1995;
11
Osborne et al. 2007). Cohabitors do not enjoy the same levels of relationship quality as married
couples. Cohabitors on average report lower relationship quality and stability and less gender
equity than their married counterparts (Blackwell and Litcher 2000; Brown 2004; Brown and
Bulanda 2008; Burgoyne 2012). Further, cohabitors are less sexually exclusive than married
couples (Joyner et al. 2013; Treas and Giesen 2000; Waite and Joyner 2001). The children born
to cohabiting couples are less likely to be intended (Guzzo and Hayford 2014; Musick 2007)
indicating that this is not the preferred location for family formation. When children are born to
cohabiting unions they result in lower levels of relationship stability (Osborne et al. 2007;
Musick 2007) than marriages. Certainly cohabitation represents greater exposure to childbearing
as couples are living together with greater with greater sexual intercourse than dating couples
(Joyner et al. 2013). Further, the greater commitment of cohabiting than dating couples to one
another may mean cohabitors are moving toward a desire to have children with one another.
Sexual exclusivity in cohabiting relationships is higher than in dating couples (Forste and Tanfer
1996). Thus, a reason for reproductive behavior differentials according to union status may rest
in part on variation in the commitment and meaning of cohabiting.
Another reason prior research has focused on union status is that it can serve as a proxy
for relationship quality and commitment in the analysis of contraceptive and fertility behavior.
An underlying assumption is that the differences in the reproductive and contraceptive behaviors
of women across union status may be attributable to union quality and stability. Indeed previous
research using specialized samples finds that negative relationship qualities (i.e., partner
mistrust, perceived partner inferiority, jealousy, verbal abuse, and violence) are associated with
inconsistent condom use among unmarried adults (Gibbs et al. 2014). Howard and Wangs
(2003) also report that relationship conflict, which is a proxy for relationship instability, reduces
12
the likelihood of contraceptive use, especially for condoms. Overall, it has been inferred that
marital, cohabiting, and dating relationships are proxy indicators of quality and associated with
contraceptive behavior. Further, it is possible that the use of specific types of contraception is
based on the monogamous nature of the relationship. In summary, based on the life course
perspective; there is evidence to suggest that union status is important in understanding the
dynamism of contraceptive behavior. It is therefore expected that this association will hold for
contraceptive switching behavior. The following two sections of the dissertation examine
current contraceptive and reproductive behaviors of women and differentials by union status.
Contraceptive Behavior and Union Status
While current scholarship has examined cohabitation and childbearing processes (e.g.,
Kennedy and Bumpass 2008; Lichter 2012; Manning 2001; Musick 2007; Raley 2001; Sassler et
al. 2009; Tach and Halpern-Meekin 2012), research on contraceptive use among cohabitors is
limited. Likewise, the instability of single women’s sexual relationships and the short-term state
of cohabitation (Lindberg and Singh 2008; Goodwin et al. 2010; Kennedy and Bumpass 2008)
suggest that more research on contraceptive behavior and union status is needed. Studies
indicate that among women union status is a significant predictor of contraceptive use (Mosher
and Jones 2010; Jones et al. 2012; Sweeney 2010).
Jones and colleagues (2012) provide a recent examination of contraceptive use by union
status using the 2006-2010 NSFG while other studies rely on earlier national data sets (e.g.,
Bachrach 1987; Sweeney 2010) or more specialized samples (e.g., Frost and Darroch 2008;
Lindberg and Singh 2008). Jones and colleagues find that among married women sterilization
(30.2%) is most popular followed by the pill (19%) and condoms (15%). Among cohabiting
women contraception ranking is different with 32.2% relying on the pill, 24% sterilized and
13
15.8% using condoms. The most popular contraceptive method among single women is the pill
(46.6%), followed by the condom (22%) and other hormonal methods (12%). The variation in
contraceptive use by union status has implications for women’s reproductive behavior.
The changing patterns of reproductive behavior in the United States are also influenced
by the contracepting behaviors of women who have not yet had a child. Most reproductive years
for women are spent being sexually active and with an increase in the period between first sex
and first childbirth (Finer and Philbin 2014) the risk of unintended pregnancy simultaneously
increases (Kavanaugh et al. 2011). An examination of a recent cohort of contracepting women at
parity zero ages 15-44 years indicate that more than half (53%) are pill users, almost one in four
(23%) are condom users, approximately one-tenth (9.6%) are hormonal users and just over 2%
are sterilized (Jones et al. 2012). This finding seems to support prior research as majority of
these women are using the pill to minimize the risk of unintended pregnancy. Other research
highlight differences in contraceptive and reproductive outcomes by parity. Using a sample of
young dating adults, Manlove and colleagues (2014) find that contraceptive use is negatively
associated with having a child. In an examination of unintended pregnancy in the United States,
Finer and Zolna (2011) highlight differentials by parity. Women with at least one birth have an
unintended pregnancy rate twice the rate for women who have not yet had a child. Therefore,
this dissertation also provides a contemporary profile of women who do not have children during
the observation period and their contraceptive switching behavior which, to my knowledge, has
not been examined in prior research.
Reproductive Behavior and Union Status
Evidence suggests that cohabitation in the United States is becoming more marriage-like
(Sweeney 2010) in terms of the reproductive behavior of women. The decoupling of sex and
14
marriage (Furstenberg 2013) has contributed to increases in nonmarital childbearing, but many
children are still born into cohabitation (Guzzo 2009; Manning et al. 2007; Smock and
Greenland 2010). Recent studies indicate that the majority (60%) of all nonmarital births are to
cohabiting couples (Manlove et al. 2010; Payne et al. 2012). Overall, births to cohabiting
women have increased while births to single women have remained relatively stable (Kennedy
and Bumpass 2008; Manning et al. 2014 a). Thus, cohabitation has become a union that more
often includes children, but not at the same level as marriage (Kennedy and Bumpass 2008;
Manlove et al. 2010; Manning and Landale 1996). Research on reproductive behavior should
take into account factors, such as (in) consistent use of contraception, and also address broader
contraceptive practices displayed by women in different unions when predicting childbirths and
context of childbearing. One such contraceptive practice is method switching. The findings
from this dissertation provide current estimates of contraceptive method switching among
cohabiting, married and single women as well as generate new debate in the area of reproductive
research.
Review of Literature on Contraceptive Use and Contraceptive Method Switching
Contraceptive Use
In the United States, almost half of all pregnancies are unintended (Finer and Zolna
2011). While some unintended pregnancies are attributable to method failure, the majority occur
due to couples’ failure to use contraception and to do so consistently (Frost et al. 2007a).
Therefore, unintended pregnancy levels are higher among long-term non-contracepting women
or those who experience gaps in method use relative to continuous contraceptive users (Glei
1999). Jones and colleagues (2012) provide recent information on contraceptive use among
women of reproductive age in the United States. Overall, 62% of women are users of any form
15
of contraception. In addition, the most common method of contraception is the pill (28%)
followed by female sterilization (27%). The percentage of non-contracepting women declines
with age and never-married women are more likely not to be currently using contraception (17%)
compared to married (8%) and cohabiting (10%) women. On the other hand, almost two-thirds
(64%) of contracepting women currently use nonpermanent methods (e.g., the pill, patch,
implant, IUDs and condom) while more than two-thirds (68%) of women at risk of unintended
pregnancy who use contraceptive consistently account for only 5% of all unintended pregnancies
(Sonfield et al. 2014).
Contraceptive Method Switching
Research on contraceptive discontinuation is discussed within the framework of two
outcomes: terminating the use of all methods (via contraceptive failure or abandonment) and
switching to a different method (Grady et al. 1988; 1989). The release of the National Survey of
Family Growth (NSFG) first cycle in 1973 paved the way for researchers to explore
contraceptive discontinuity in the United States. Advantages of this survey includes month-bymonth information on contraceptive use and exposure to intercourse which are ideal for
measuring the likelihood of contraceptive discontinuation, or continuation, by given durations of
exposed method use, among women at risk of unintended pregnancy (Vaughan et al. 1980;
Grady et al. 1983; 1988; Hammerslough 1984). The 1995 NSFG was employed by Grady et al.
(2002) to examine differentials by marital status, and results indicated that the rates of switching
were high for married and unmarried women and that their decisions to switch may have been
driven by contraceptive effectiveness and health risks associated with contraceptive use
(discussed in more detail below). There has not been a contemporary update to these studies
using nationally representative data. The most recent comprehensive and nationally
16
representative study by Grady et al. (2002) is based on data collected nearly two decades ago.
However, the types and availability of contraception in the United States has changed overtime.
During the 1960s the pill and IUD were introduced and more highly effective methods later in
20th century (Hatcher et al. 2004; Center for Disease Control and Prevention 1999). Further, new
methods (Nuva Ring, Implanon and contraceptive patches) have been recently approved and
introduced in the United States (Hatcher et al. 2004). In addition, there has been a rise in
cohabitation and nonmarital fertility (Litcher et al. 2014; Manning 2013).
International data from the Demographic Health Surveys (DHS) also provide findings on
oral contraceptive discontinuation and method switching. Ali and Cleland (2010) find that a
little over one-third of women who discontinue contraceptive use because of method
dissatisfaction switch to another method within three months. This finding suggests that while
abandonment of all methods is associated with negative sexual and reproductive health outcomes
(Finer and Henshaw 2006; Moreau et al. 2007), there are women who, while at risk of
unintended pregnancy and or abortion due to contraceptive discontinuation, try to resume use of
another method in a ‘short’ time frame. Therefore, more focus needs to be placed on
contraceptive method switching as opposed to just contraceptive discontinuation in order to gain
greater knowledge and understanding of the processes involved in contraceptive behavior.
Despite growing interest in the contraceptive method switching behavior of women in
developing nations over the last three decades (e.g., Ali and Cleland 2010; Curtis et al. 2011;
Kane et al. 1988; Steele and Diamond 1999), there is a paucity of research focusing on the U.S.
There are few studies discussed below that examine contraceptive method switching based on
convenience samples in the U.S. (e.g., Davidson et al. 1997; Sangi- Haghpeykar et al. 1995;
Santelli et al. 1995; Weisman et al. 1991). For instance, in a prospective study conducted by
17
Sang-Haghpeykar and colleagues (1995), 600 women from 17 family clinics in Texas are
sampled to ascertain the characteristics of injectable contraceptive users. These Depo-Provera
recipients are followed for a year after receipt of initial injection. The participating clinics in this
study are chosen as they are thought to provide an adequate representation of racial and ethnic
diversity as well as age groups. Thirty percent of Depo-Provera recipients are 21 years or
younger and over three-quarters (77%) are unmarried. Results from this study find that women
are more likely to switch from more effective contraception (Depo-Provera) to less effective
methods (condoms and pills) because the former method is mainly used for birth spacing.
Davidson et al. (1997) also provide evidence of contraceptive method switching. The
results of this study is based on a sample of 491 women selected from three large, hospital based
family planning clinics serving poor and ethnically diverse populations in Dallas, New York City
and Pittsburgh. Women are at least 15 years, using Depo-Provera and had received contraceptive
counseling. The re-interview rate based on the follow-up interview produced a sample of 402
women. The findings of this study provide evidence of contraceptive switching from more
effective methods to less effective methods. Among the discontinuers of Depo-Provera who
report switching to another method, 55% switch to oral contraceptives while 31% use condoms.
The main reasons given for discontinuation of method are side effects rather than difficulty
returning to the clinic every 3 months.
A three year wave panel conducted by Weisman and colleagues (1991) also produces
results related to contraceptive method switching. The sample includes 308 adolescents who
report at least one instance of intercourse during a six-month follow-up period. More than threequarters of the sample are black and 9 in ten adolescents live in the inner city. The findings of
the study indicate that adolescent women who initially report use of condom switch to dual use
18
for at least part of the follow-up period. Subsequent finding indicates that consistent pill use is
negatively associated with condom use which suggests that women in this clinic based study use
condoms for pregnancy prevention purposes rather than for STIs (Sexually Transmitted
Infections) prevention.
Santelli et al. (1995) provide evidence on the likelihood of contraceptive method
switching among women at greater risk for HIV/AIDS. This study is based on a street survey
conducted among 717 women ages 17-35 in two inner city Baltimore communities as part of an
evaluation of the effects of a community-based perinatal HIV prevention program. Findings
from this study are linked to two rounds of prior community based studies. In this study, the
likelihood of switching from one method of contraception to dual method use is increased at
recent intercourse. In other words, women who use only pills or condoms at last intercourse are
likely to switch to dual methods (pills and condoms) at most recent intercourse. Overall, these
studies reveal that contraceptive method switching may improve women’s reproductive health
and sexual health by giving women the ability to space child births while on the other hand,
contraceptive method switching may produce negative consequences such as unintended
pregnancies. Notwithstanding the results of these studies, the main limitation is the use of
convenient samples of mostly young women drawn from health clinics, disadvantaged
neighborhoods, and from a higher risk of STI group of women. That is, the likelihood of
contraceptive method switching in these samples cannot be generalized to the national
population of women
Recent examination of contraceptive patterns and switching are also explored using both
specialized and nationally representative samples of women. Barber and colleagues (2011)
draws data from a population-based sample of young women, ages 18-19 years. Weekly journal
19
surveys measure contraceptive use patterns, relationship status and pregnancy outcomes. The
weekly measures of contraceptive behavior not only ensure correct recall but provide accurate
estimates for contraceptive instability and change. However, investigating method switching
using this sample limits the interpretation of results only to young adults within a specific
geographical region in the United States. Using data from the NSFG 2006-10 Manlove and
colleagues (2013) find that almost 30% of women 15-44 years switch contraception over a 12
month period with an average of almost 2 switches within the year. While the sample captures
all women of reproductive age, it examines switching based on contraceptive records during the
last year among women using only hormonal and long-acting methods.
Frost et al. (2007a, b) also produce an overview of women’s contraceptive method use
including switching behavior based on a telephone survey of a sample of nearly 2,000 women
18-44 years. Nearly one in four women switch contraception and 15% of women switch in and
out of contraceptive use due to a pregnancy or a period of sexual abstinence during the last 12
months (Frost et al. 2007b). In a related study Frost and colleagues (2007a) reveal that majority
(73.4%) of the women at risk of pregnancy who use contraception are stable users (no change in
use), 8.1% and 8.7% switch to more effective and less effective methods respectively and almost
one-tenth (9.8%) of women switch to none-use of contraception. Over 4% of women who start
with barrier/traditional methods and 2% of non-users switch to more effective methods. On the
other hand, 8.7% of hormonal users and 2% of non-users switch to less effective methods at the
end of the 12 month period. Overall, 8% of women switch to a more effective contraceptive
compared to 19% who switch to less effective methods or none of the methods (Frost et al.
2007a). These studies are limited in terms of its generalizability. Women’s overall switching
20
behavior is measured based on reports in the last 12 months and union status of women are
determined at point of interview.
In conclusion these prior studies provide a narrow lens into the understanding of
contraceptive method switching behavior. The most compelling drawback of these studies is the
inability to use findings to generalize to the national population which also restricts the overall
development of policy recommendations on reproductive health. However, these specialized
studies have paved the way for larger nationally representative studies on contraceptive method
switching.
Contraceptive Method Switching and Union Status: Married versus Unmarried Women
Studies on patterns of contraceptive method switching in the United States with a
nationally representative sample of women of reproductive age are few. The findings of these
studies focus solely on marital status (married versus unmarried women) or married women only.
Grady and colleagues (1988) estimated contraceptive discontinuation rates exclusively among
married women in the U.S. Using data from the 1982 NSFG Grady et al. (1989) produce
discontinuation rates for contraceptive methods as well as examine the probability of method
switching among married women. Findings reveal that for these women there is a high rate of
abandonment of non-permanent contraception and a switch towards sterilization and more
specially, nonuse. Although women who are nonusers are overrepresented among adopters of
sterilization, these findings partially support earlier work by Bachrach (1984), which shows a
decline in pill use but a simultaneous rise in sterilization among married women. Further, Grady
and colleagues (1989) find that there is also a general pattern of switching to nonuse which then
places married women at increased risk for unintended pregnancy.
21
The most recent systematic investigation of contraceptive method switching among a
representative sample of women in the United States is over a decade old (Grady et al. 2002).
Using data from the 1995 NSFG, the authors compare methods used by married and unmarried
women. They also estimate the proportion of women who rely on specific methods and
determine the correlates of contraceptive method switching for married and unmarried women.
This cycle of the NSFG, as is the case for previous cycles, is appropriately suited for evaluating
method switching as it uses an event-history calendar to collect month-by-month information on
contraceptive use and important life events. The authors categorize contraceptive methods into
five groups: long-term reversible methods (hormonal implant, injectable, and IUD); pill; condom
(including use in combination with other less effective methods); all other, less-effective
methods; and no method.
Grady et al. (2002) define contraceptive method switching based on three outcomes.
First, women who report different use contraceptive methods in consecutive months. Second,
use of two methods (including nonuse) separated only by a period of abstinence. Third, use of
two methods sequentially in the same month, and the woman’s use one method in the preceding
month and the other in the subsequent month. Within each marital status group (married and
unmarried), discrete-time hazards models are used to examine the socio-demographic
determinants of method switching. The results of this study indicate a significant amount of
switching among women (40% married versus 61% unmarried). Among married women the
rates of sterilization and other long acting reversible contraception are observably lower than the
rates of switching to other remaining methods (e.g., pill). One-fourth of married women who are
nonusers at the start of the observation period comprise 50% of all married women who switch to
the pill. Unmarried women are more likely to switch to nonuse compared to married women.
22
For unmarried women whose original methods of contraception are long term reversible
methods, pills and condoms, the most common destination following switching is nonuse.
Grady and colleagues (2002) provide a descriptive analysis of the percentage of women
(married and unmarried) who begin with a specific method and switch to other methods. Most
married women who begin with the pill switch to condoms while those who start with condoms
switch to more effective methods. In addition, married women who use hormonal methods
switch to more effective methods (i.e., sterilization). For unmarried women who start with the
pill, most switch to more effective methods (i.e., dual methods) and those who start with
hormonal methods switch to less effective methods. Overall, all women in the sample are more
likely to switch to no method and also to be non-users for an extended period of time.
Multivariate results suggest that women’s method switching decisions are associated with their
desire for pregnancy protection; protection against STIs (among unmarried women); and their
desire to avoid health risks related to contraceptive use. There is also a high rate of switching
(greater than 30%) for women using reversible methods and the authors suggest that this is due
in part to changes in women’s life experiences and how well their method fit their needs. Grady
and colleagues (2002) find a different story in terms of patterns of switching among married
women compared to earlier work (Grady et al. 1989). The rates of contraceptive method
switching to long-term reversible methods and sterilization are lower than all other methods.
This finding suggests that there may be a change in the pattern of contraceptive use and
switching by different cohorts of women in their reproductive life course.
Therefore, analyzing contraceptive method switching by union status (married,
cohabiting and single) among a recent cohort of reproductively age women is timely and takes
into account recent shifts in union formation, changes in fertility processes, contraceptive use
23
(Manning et al. 2014 a; Jones et al. 2012) and the accessibility of different modes of
contraception especially among young adults and women without children (Finer et al. 2012;
American College of Obstetricians and Gynecologist 2011). Prior studies examining
contraceptive method switching and method discontinuation explore differences by marital status
(married versus unmarried) and (married, cohabiting, formerly married and never-married); (see
Grady et al. 2002 and Trussell and Vaughan 1999). In addition, this dissertation also
incorporates a more nuanced approach to measuring union status. Capturing union status means
the following: 1) marriage may be first or higher order at time of observation; 2) cohabitation at
observation means persons could have been in previous marital and cohabiting unions; and
single at observation is complex as it includes women who are dating, in a nonmarital romantic
relationship, or women who have been in previous marital and cohabiting unions.
Correlates of Contraceptive Method Switching
Sociodemographic Characteristics
Age
Prior studies indicate that women’s choice of contraception varies by age (Mosher et al.
2004, 2010). Jones and colleagues (2012) find that this variation continues to exist based on an
examination of current contraceptive use and changes in patterns of use in the United States.
Findings indicate that condom use decreased by nearly half among women 15-19 years from
36% of teenagers in 1995 to 20% of teenagers in 2006-10. The use of hormonal methods of
contraception increased in all age groups over the period and is noticeably higher for women
under 30 years. The use of IUDs also increased mostly among women aged 25-39 years.
Age is also a significant predictor of contraceptive method switching. Among a sample of
married women, older women who do not use any method while at risk of pregnancy are less
24
likely to switch than younger nonusers to a method (Grady et al. 2002). However, when older
women do switch contraceptive methods, they are more likely to utilize sterilization. Older
women are also more likely than younger women to switch from condom to less effective
contraceptive methods (Grady et al. 2002). Among unmarried women age is also significantly
related to women’s contraceptive switching behavior. Older women compared to younger
women are less likely to switch to condom, except among dual users (Grady et al. 2002). Older
women who use condoms or other reversible methods are less likely to switch from their method
and adopt a dual methods or the pill (Grady et al. 2002). The overall findings by Grady and
colleagues suggest that older women have lower demand for contraceptive methods that are
effective in preventing pregnancy or STIs.
Race/Ethnicity
Racial and ethnic differences in the rates of contraceptive use are well established (Finer
and Henshaw 2006; Finer and Zolna 2011; Frost et al. 200b; Jones et al. 2012). Research have
also examined race/ethnicity and women’s use of any contraception (Frost et al. 2007b; Kearney
and Levine 2009) as well as specific methods (e.g., the pill or IUD) (Frost and Darroch 2008;
Nearns 2009; Jones et al. 2012). Jones and colleagues (2012) provide recent results of changes
in contraceptive use between 1995 and 2006-2010. Among black women there is a decrease in
pill use from 24% to 18%. There are increases in hormonal method use among white (3.4% to
5.3%), black (7.8% to 13%) and U.S. born Hispanic (5.2% to 10%) women. Increases in the use
of IUDs are seen across most racial and Hispanic origin groups. Specific to the 2006-10 data, a
higher percentage of black women (37%) compared to U.S. born Hispanic (27%) and white
women (24%) use female sterilization. Overall, across all race/ethnic groups, pill use is highest
among white women (32%).
25
Prior research also finds an association between race/ethnicity and contraceptive method
switching (Grady et al. 2002). Married black women compared to married white and Hispanic
women have reduced rates of switching from less-effective reversible methods to sterilization.
Further, married black women also have higher rates of switching from pill to no method and
from no method to pill compared to whites and Hispanics (Grady et al. 2002). Hispanic married
users of reversible methods and non-users are more likely than black and white married women
to switch to a long term reversible method and to the condom (Grady et al. 2002). On the other
hand, unmarried black women compared to women of other races are less likely to switch from
any contraceptive method to either the pill or less effective reversible methods. Grady and
colleagues (2002) suggest that contraceptive method switching among black unmarried women
is likely using methods that will protect against STIs rather than a method that protects against
unintended pregnancy. Unmarried Hispanic women who are condom and pill users are less
likely than non-Hispanic women to switch to dual methods (Grady et al. 2002).
Education
The association between educational attainment and contraceptive behavior is explored
rigorously. Recent research (Jones et al. 2012) among Hispanic women indicates several
changes in contraceptive use between 1995 and 2006-10 by educational attainment. The findings
reveal that there is a decrease in pill use (21% to 14%) among women with a high school
diploma or less. Further, condom use by women’s partners decrease among Hispanic women
with more than high school education (29% to 22%). Among white women, condom use also
decrease during the period from 13% to 8.9% for women with a high school diploma or less. On
the other hand, there are increases in female sterilization among women with a high school
diploma or less (40% to 47%) and an increase in hormonal use from 2.1% to 5.2% among
26
women with more than high school education. There is an increase in hormonal use among
black women regardless of educational attainment; however the increase is highest for those with
more than a high school education (5.1% to 13%). Increases in the use of IUDs among black
women with more than a high school education (0.7% to 8.4%) is also significant.
Existing literature reveals that an association between contraceptive method switching
and education does exist. Among married women, higher levels of education are associated with
a decrease in the likelihood of women switching from the pill to less effective methods (Grady et
al. 2002). Also, for married women who use condoms, more years of education is associated
with decreased likelihood of switching to sterilization, long term reversible methods and the pill.
Nonusers with more education have increased rates of switching to sterilization, the pill and the
other reversible methods (Grady et al. 2002). These findings by Grady and colleagues suggest
married women compared to unmarried women, with more education, are likely to switch from
less effective to more effective methods. In another study, Frost and colleagues (2007b), find
that women with a high school diploma/GED and some college education have higher odds of
method switching compared to women with less than a high school education, with college
degree or higher as the reference category.
Poverty Level
Women’s contraceptive behavior is also associated with poverty level. The resources
required to obtain and sustain use of contraception suggest differences in the use of contraception
according to poverty status (Jones et al. 2013). Few nationally representative research have
examined how contraceptive choice and efficacy are associated with socioeconomic
characteristics, including poverty levels (e.g., Frost and Darroch 2008; Frost et al. 2007a; Grady
et al. 1999). In these studies, poverty status is measured using the federal eligibility criteria for
27
subsidized family planning services (Frost and Darroch 2008; Frost et al. 2007b; Trussell and
Vaughan 1999) or in other cases the income to poverty ratio is used in the analysis (e.g., Nearns
2009). Jones and colleagues (2013) report that poor women are less likely to use contraception
and more likely to have gaps in use compared to women who are better-off financially.
While extant research provides evidence of an association between poverty and
contraceptive use, studies linking poverty to contraceptive method switching are limited.
Previous research based on family clinic and inner city samples of poor women indicate that
poverty does play a role in the discontinuation of contraceptive use as well as method switching
(especially switching to more effective methods), (Davidson et al., 1997; Sangi-Haghpeykar et
al., 1995; Santelli et al., 1995). Frost and colleagues (2007b) in a more recent study find that
while poverty is not significant in predicting method switching, 25% of women below the federal
poverty line switch contraception within 12 months.
Background Characteristics
Family Structure
Family structure is one of the many background characteristics associated with
contraceptive use. Several studies using adolescent and young adult samples show an
association between family structure and contraceptive use (Brindis et al. 2000; Manning et al.
2009; Manlove et al. 2004, 2011). Specific findings reveal that adolescents leaving with two
parents are more likely to use contraceptive methods (Manlove et al. 2007; Kusunoki and
Upchurch 2011). In addition, adolescents who experience greater family structure changes have
higher risk of romantic instability as well as early sexual debut (Brown 2006; Cavanagh et al.
2008; Fomby and Cherlin 2007). Family structure is also associated with trends in the ages of
key reproductive life transitions, for example, first sex and contraceptive use (Finer and Philbin
28
2014). Prior studies find that the negative consequences of family structure manifest themselves
in adulthood. Children who experience multiple family structure transitions are more likely, as
adults, to be at a greater risk of nonmarital childbirth (Hill et al. 2001).
Earlier studies on contraceptive method switching have not included family structure in
the analyses (Grady et al. 2002; Frost et al. 2007b). However, based on prior finding, family
structure is associated with adolescent and adult contraceptive outcomes. Family structure is
expected to be a key indicator of adolescent family life that can operate as a distal covariate
predicting contraceptive method switching.
Religious Affiliation
Religious affiliation is a factor associated with sexual and contraceptive behaviors
(Goldscheider and Mosher 1991; Grady et al. 1993; Hill et al. 2013; Lammers et al. 2000; Meirer
2003). Goldscheider and Mosher (1991) using data from the NSFG 1988 examine the
importance of religious factors on contraceptive use and conclude that it is important not to
discredit religion as a factor in the contraceptive styles of women. More specifically, religious
affiliation and religiosity, measured by church attendance, taking of communion among
Catholics and attending church-related schools, are associated with contraceptive patterns
leading to low fertility within the context of a secularized society. In addition, findings reveal
that Protestants are more likely to use female sterilization than are Catholics while Catholics are
more likely to use the condom and the pill because they postpone marriage longer. Jones and
colleagues (2012) provide a contemporary description of religious affiliation and contraceptive
use in the United States. Among Baptist and Fundamentalist Protestant women, 41% use
sterilization compared to Catholic (24%) women and women not affiliated with a religion (22%).
29
Women with no religious affiliation (31%), Protestant (29%) and Catholic (28%) women use the
pill more often than Baptist and Fundamentalist (21%) women.
As it relates to contraceptive method switching and religious affiliation, Grady and
colleagues (2002) find mixed results among married and unmarried women. Among married
women with no religious affiliation and those who use condoms they are less likely to switch to
sterilization while those who are pill users have an increased likelihood of switching to
sterilization or long term reversible methods. Unmarried women with no religious affiliation are
more likely to switch to less effective contraception.
Mother’s Education
Studies focusing on the contraceptive and reproductive behavior of teenagers and young
adults tend to examine individual and family influences as possible correlates (Manlove et al.
2011; Kusunoki and Upchurch 2011; Manning et al. 2012). One such correlate is mother’s
education, which is a proxy for social class. Overall, there is consistency in the literature that
family characteristics such as mother’s educational attainment captures a dimension of
socioeconomic class status which is associated with adolescents sexual and contraceptive
behaviors. Research indicates that high levels of mother’s education is associated with later
timing of first intercourse and greater contraceptive use at first sex, and lower risk of teen
pregnancy and teen births (Manlove et al. 2000; Hogan et al. 2000). In addition, having a mother
with a higher level of education is associated with greater contraceptive use, including the use of
pill relative to condoms (Manlove et al. 2007; Kusunoki and Upchurch 2011).
The inclusion of mother’s education in the analysis of method switching and union status
is important because of its overall association with contraceptive use. In this study mother’s
education is included as an indicator of social class.
30
Fertility Characteristics
Age at First Sexual Intercourse
Most women initiate sexual intercourse as teenagers, at an average of 17 years (National
Center for Health Statistics 2013). There are several studies that examine age at first sex and
contraceptive use. These studies reveal that younger age at first sex is associated with reduced
use of contraception (Glei 1999; Kirby et al. 2005; Manning et al. 2000). Conversely, some
studies find that older age at first sex is negatively associated with condom use (DiClemente et
al. 1996; Ku, Sonenstein and Pleck 1994). However, based on recent data in the United States,
majority of sexually experienced teenagers at first sex use contraception (Martinez et al. 2011),
which is important because contraceptive use at first sex is considered an indicator of later
consistency of use (Shafii et al. 2004).
It is against this background that I consider age at first sex to be a distal indicator
measuring sexual risk which then can be associated with contraceptive method switching. Early
sexual experiences coupled with inconsistent contraceptive use may impact women’s
contraceptive method switching behavior in adulthood.
Parity
A vast body of studies on contraceptive and reproductive behaviors of women
throughout the life course have included parity in the analyses (Jacobs and Stanfors 2013; Jones
et al. 2012; Kavanaugh et al. 2011; Manlove et al. 2014). Data from the NSFG 2006-10 on
childless contracepting reproductive age women indicate that more than half (53%) are current
users of the pill, almost one-quarter (23%) are condom users, one in ten are hormonal users and
5% are sterilized (Jones et al. 2012). Frost and colleagues in their analyses of factors associated
31
with contraceptive use and nonuse find that women with one birth are more likely to switch
contraception than women at parity zero.
Decisions about contraceptive use depend on parity. The lengthening of time between
first intercourse and first child birth provides more time for women’s contraceptive switching
behavior. Among married women without children the risk of switching from a contraceptive
method to sterilization is low (Grady et al. 2002). Further, married women at parity zero rarely
switch to long-term reversible methods and they also have reduced risks of switching from no
method to the pill (Grady et al. 2002). On the other hand, Grady and colleagues report that
among unmarried women without children are less inclined than women with one or more
children to switch from condoms and pills to a long-term method (Grady et al. 2002). Frost et al.
(2007b) find an association between methods switching and parity among adult women (18-44
years) at risk of an unintended pregnancy. Women who have a child compared to women
without children are more likely to switch methods of contraception. Similar findings by Finer
and colleagues (2012) reveal that younger women who had experienced one or more births made
a switch to the use of LARC even though many may have already reached their desired fertility
goals. Overall, parity is included in the analyses because women at parity zero at the time of
observation have a higher likelihood of postponing rather than preventing a birth (Grady et al.
2002). In this dissertation, I examine parity as a possible mediating covariate in the overall
analysis of women and whether they switch contraception. In addition, specific investigation of
contraceptive method switching behavior is conducted among women who without children at
time of observation.
32
CHAPTER II: CONTEMPORARY PORTRAIT OF STABLE USERS AND
CONTRACEPTIVE METHOD SWITCHERS
Current Investigation
This research expands prior inquiries about women’s contraceptive method switching by
providing a contemporary descriptive portrait of contraceptive method switching behavior for all
women in relation to union status. The main research question of this chapter examines whether
married women differ from cohabiting and single women in terms of contraceptive method
switching and among women who are non-switchers (i.e., stable users and stable nonusers). This
group of women is investigated based whether they are stable non users and stable users. First, I
examine the overall patterns of women who switch contraception compared to those who remain
as non-switchers. Second, I provide estimates of the prevalence and predictors of contraceptive
switching and stable use. This represents an extension of prior studies from earlier time periods
(Grady et al. 1989, 2002; Hammerslough 1984; Vaughan et al.1980) by focusing on switching
behavior of women across union status (single, cohabitation and marriage) during the three year
time interval. Third, I investigate whether there are differences among stable users, stable
nonusers and switchers and if these differences vary across union status.
Prior research indicates that method switching is more prevalent among unmarried versus
married women (Grady et al. 2002), but little attention is given to cohabitation. Recent analysis
of women of reproductive age indicates that there are differences in the contraceptive behavior of
cohabiting women compared to married women (Sweeney 2010). In addition, cohabiting women
are less likely to have a child than married (Manning and Landale 1996) and more likely than
single women (Loomis and Landale 1994). Given the unstable nature of single women’s sexual
relationship (Lindberg and Singh, 2008) and short time span of cohabitation (Goodwin et al.
33
2010; Kennedy and Bumpass, 2008), I consider variations in contraceptive switching according
to union status.
Hypotheses
Given these previous findings, I propose the following hypotheses. The first two are
related to contraceptive switching relative to stable use while the remaining six examines
contraceptive switching, stable nonuse and stable use.
Hypothesis 1: Single women (compared to cohabiting and married women) will more often
engage in contraceptive method switching.
Hypothesis 2: Cohabiting women (versus married women) will experience greater odds of
contraceptive method switching.
Hypothesis 3: Single and cohabiting women (compared to married women) will more often be
stable users than stable non users of contraception.
Hypothesis 4: Single women (compared to cohabiting women) will more likely be stable users
than stable nonusers.
Hypothesis 5: Single and cohabiting women (compared to married women) will more often
switch contraception than remain as stable nonusers.
Hypothesis 6: Single women relative to cohabiting women will more often switch contraception
than remain as stable nonusers.
Hypothesis 7: Single women relative to married women are less likely to switch contraception
than remain as stable users.
Hypothesis 8: Single women relative to cohabiting women are less likely to switch
contraception than remain as stable users.
Data and Methods
The National Survey of Family Growth (NSFG) was conducted from June 2006 to June
2010. This survey comprised 12,279 non-institutionalized women, ages 15-44 years. The NSFG
is appropriate for this dissertation because of the contemporary nature of the data set which
includes detailed retrospective contraception, marriage and cohabitation histories as well as
34
socio-demographic variables that are associated with contraceptive method switching. The
NSFG contains a contraceptive method history calendar in which dates of respondent’s use of
contraception in each month during the four years preceding the interview is recorded. This
enables the examination of method switching. Marital and cohabitation start and end dates are
also provided in the NSFG. To date no other nationally representative data offers these
advantages. Event history analysis is a common technique used in prospective panel studies,
where the same sample of individuals is analyzed or retrospectively in cross-sectional studies, as
is the case with the NSFG. For the purposes of this dissertation, the dataset is restructured into
person-month files. I create a data file which contains one record for each observed month that
each woman spends using any or no form of contraception. Where there are missing items (i.e.,
refused and don’t know) for any month observed it is excluded from the analysis. The exposure
period is uninterrupted by any period of sexual abstinence but ends when there is a change in
union status, the event of switching occurs or at time of interview. The analysis essentially pools
all the person-months of risk of contraceptive method switching and estimates the effect of union
status on contraceptive method switching. For event history analyses, odds and relative risk
ratios are produced for logistic and multinomial logistic regression models respectively. These
ratios gives the risk that an event will occur given it had not previously occurred. Odds and
relative risk ratios greater than 1 indicate a higher risk of an event occurring compared to the
reference category while odds and relative risk ratios less than 1 indicate a lower risk of an event
occurring compared to the reference category.
Analytic Sample
I examine the patterns of contraceptive method switching and variations by union status.
The NSFG 2006-10 is a national non-probability sample representing the household population
35
of 12,279 non-institutionalized female respondents, ages 15-44 years. The inclusion of
respondents with valid union histories reduces the sample to 10,761. I also include only valid
responses to contraceptive method switching questions from the method history calendar
(N=9,470). The sample is further limited to respondents observed during a three-year period and
not sterilized at the start of the observation period (N= 4,674). Prior studies have used one and
two year periods to assess contraceptive method switching (Grady et al. 2002; Frost et al. 2007).
In this dissertation, a three-year observation period is used in order to capture as many cases of
women’s monthly contraceptive history. Following the merging of contraceptive method
switching file and the NSFG data file containing variables used as control covariates the sample
stands at 3,122 respondents. The final restriction includes limiting the sample to women ages
21-44 1 years with non-missing responses on age, race/ethnicity, education, poverty status,
religious affiliation, mother’s education, age at first sex and parity. The final analytic sample
consists of 2,986 respondents.
Dependent Variable
The dependent variable for this research is contraceptive method switching. I use
respondents’ retrospective reports of their contraceptive method history over the last 3 years.
Predicting method switching behavior takes two forms namely: a) dichotomous variable
(switching versus not switching); and b) categorical variable (stable nonusers, stable users and
switchers). Contraceptive method switching is measured based on women’s report of different
use of contraception in consecutive months during observation uninterrupted by any period of
abstinence. In addition, a change from nonuse of contraception to use of contraception or vice
versa is considered a switch. In this dissertation, the timing of method switching from the initial
1
The sample is restricted to women 21-44 years so that 3 years prior to interview women’s union and contraceptive
histories are taken into account between ages 18-41 years.
36
union formation is not investigated because women are not captured at first contraception use or
union formation but during a retrospective three year period which can occur at different stages
of contraception history. I only take into account the occurrence of the first switch during the
observation period.
Independent Variable
The main independent variable is union status. Respondent’s union status at the time of
interview does not tell us about the union context during contraceptive method switching.
Therefore, union status is measured as a time invariant characteristic using retrospective dates of
marital and cohabitation histories at the time of contraceptive switching. At the start of the
observation period respondents are either in a marital or cohabiting union. If there are no marital
or cohabiting dates that correspond to the commencement of the observation period, respondents
are classified as single.
Regarding the union status, being married at observation denotes any marriage (first or
higher order) at that time. Similarly, cohabitation at observation denotes that women could have
been married or in previous cohabiting unions and being single denotes that women could have
been in previous marital and cohabiting unions or have never married or cohabited.
Sociodemographic Characteristics
Several variables are included in the analyses as controls because of their potential
confounding associations between union status and contraceptive method switching. I include
the respondent’s age. Age is measured at the time of interview and is a continuous variable
measured in years. Respondent’s race/ethnicity is also included and I use a NSFG recode based
on the 1997 Office of Management and Budget (OMB) standards to create a four category
response measure: non-Hispanic white (reference category), non-Hispanic black, Hispanic, and
37
multiracial. I also include respondent’s education (measured at time of interview) in the analyses
and coded into four categories: less than high school degree, high school/GED (reference
category), some college and college degree or higher. Poverty status is based on the federal
eligibility criteria for subsidized family planning services and respondents are grouped into
dichotomous variable: 1= at/above the poverty line (federal poverty is >= 100%) and 0 = below
the poverty line (federal poverty is 0-99%).
Background Characteristics
Family type is a dichotomous variable based on a NSFG recode of intact status of
childhood family where respondents have two biological/adoptive parents from birth or
childhood family where respondents have anything other than respondents two
biological/adoptive parents from birth. Respondent’s religious affiliation while growing up is
included in the analysis. This coded into four categories: no religion (reference category),
Protestant, Catholic and other religious affiliation. Respondent’s mother’s education is taken
into account and categorized as: less than high school degree, high school/GED (reference
category), some college and college degree or higher.
Fertility Characteristics
Age at first sex is a recoded continuous variable based on the question: “Whether
respondent has ever had sexual intercourse with a male (even if before menarche)?”
Respondent’s parity at observation is included in the analyses. I use the NSFG recode, which
captures the total number of live births including multiple births, to create a dichotomous
variable: 1= have at least one child and 0= have zero births.
38
Measure of Time
I include a continuous measure of time, measured in months. The variable counts the
number of months of women’s contraceptive use during the observation period which is not
interrupted by any period of sexual abstinence. Respondents are censored (removed from
analyses) if there is change in union status during the observation period and for those
respondents who do not switch contraception, the time measure is censored at date of interview
or when the method calendar ends during the observation period.
Analytic Plan
Descriptive analyses show the percentage of women by union status and their
contraceptive outcomes during the three (3) year observation period. In order to determine the
prevalence and predictors of switching behavior I provide weighted means, proportions and
standard errors based on the analytic sample. In addition, differentials in sociodemographic,
background and fertility characteristics according to union status are presented. Finally, I show
zero-order and multinomial logistic regression models examining the effects of women’s union
status at the start of the observation period on the probability of switching contraceptive
methods. This analysis is appropriate when predicting multiple event types, for example: stable
switching; stable use; and stable nonuse.
Results
Descriptive Results
Table 2.1 shows that during a three-year observation period, among women 21-44 years,
approximately 40%, engage in contraceptive method switching. Among the remaining nonswitchers, 17% of women are stable nonusers of contraception and 43% are stable users of
contraception. Analyses showcasing differentials according to union status indicate that single
and cohabiting women are more likely to switch contraception than cohabiting and married
39
women. Overall, 53% of cohabiting women switch methods of contraception compared to 50%
of single women and 35% of married women. Married women are more likely to remain as
stable nonusers of contraception compared to single and cohabiting women. Further, married
women relative to cohabiting and single women are more likely to be stable users.
Table 2.1 presents differentials in the sociodemographic characteristics according to
union status. The average age in the sample was 33 years. Married women are significantly
older than cohabiting and single women. In terms of race and ethnicity, married and single
women have a greater proportion of non-Hispanic white compared to cohabiting women. The
modal education category (44%) is a college degree or higher. Single women are more highly
educated (college degree or higher) (47%) compared to married women (46%) and cohabiting
women (15%). Almost nine in 10 women (87%) are at or above the federal poverty line. More
married women (90%) are at or above the federal poverty line compared to single women (84%)
and cohabiting women (71%). Background characteristics also vary by union status. A greater
percentage of married women (74%) compared to cohabiting (53%) and single women (65%),
report they were raised in a two biological/adoptive parent household prior to age 18. More
cohabiting women report they were raised Catholics than married and single women. The
proportion of mothers of cohabiting women with a high school diploma is higher than proportion
of mothers of single and married women have a high school diploma. Additionally, more
mothers of single women compared to those of cohabiting and married women have less than
high school education (12%) and more have a college degree or higher (28%). The average age
at first sexual intercourse in the sample was 18 years. Cohabiting women’s age at sexual debut is
younger than those of married and single women. More than two-thirds of women in the sample
have at least a child at observation. Parity varies by union status such that married women (79%)
40
are more likely to have a child at observation compared to cohabiting and single women with
56% and 29% respectively.
Switching Contraceptive Methods: Bivariate and Multivariate Results
Table 2.2 presents the zero-order and multivariate discrete-time logistic regression
models to examine the effects of women’s union status at the start of the observation period on
the probability of switching contraceptive methods. The table shows odds ratios of contraceptive
method switching relative to not switching. A time-varying indicator (measured in months) is
included in all analyses. At the zero-order level, union status is significantly associated with
contraceptive method switching. More specifically, single and cohabiting women have higher
risks (89% and 97%) of switching contraception compared to married women. No difference is
found in the odds of switching for single and cohabiting women (results not shown). Age is a
significant predictor of contraceptive method switching at the zero order level. With each year
increase in women’s age, the risk of switching contraception is reduced by 9%. There is an
association between poverty, measured by the federal poverty line, and contraceptive switching
such that women at or above the poverty line have a 31% lower risk of switching contraception.
Women whose mothers have at least some college degree have an 18% increased risk of
switching contraception. With each year increase in age at sexual debut women’s risk of
switching contraception is reduced by 4%.
Each subsequent model in Table 2.2 adds each set of covariates: sociodemographic and
background (model 1), fertility characteristics (model 2) to the union status model predicting
method switching as well as a final model (model 3) containing all control variables. With the
inclusion of sociodemographic and background variables (model 1) the association between
union status and contraceptive method switching is not statistically significant. However, age is
41
related to contraceptive method switching. For every one year increase in a woman’s age, the
risk of switching contraception deceases by 9%. Supplemental analyses reveal that women’s
age, added separately into the model, explains the mediation effect on union status (results not
shown).
The next column shows that with the inclusion of fertility variables there is a marginally
significant association between union status and contraceptive method switching (p <.10).
Single and cohabiting women, compared to married women, have a 100% and 112% risk of
switching contraception. There is no difference in the risk of switching between single and
cohabiting women (results not shown). Age at first sexual intercourse is also marginally
significantly associated with contraceptive method switching. With each year increase in the age
at first sex women’s risk of switching is reduced by 3%.
In the final model (Model 3), union status is not statistically associated with method
switching. Age remains as the only significant predictor of method switching. Every one year
increase in the age of women decreases the risk of switching by 10%. Race and ethnicity and
parity are marginally associated with contraceptive method switching (p <.10). Non-Hispanic
black and multiracial women compared non-Hispanic white women have a 19% and 53% lower
risk of switching contraception. Women who have at least one child at observation have an 81%
higher risk of switching contraception.
In sum, the results indicate that net of all covariates union status is not statistically related
to the odds of switching contraceptive methods. Therefore H1 is not supported as single women
share similar odds of switching contraception compared to married women. The second
hypothesis (H2) which posits that cohabiting women will experience greater odds of
contraceptive method switching relative to married women is also not supported. Further
42
analyses indicate that age of women fully mediates the association between union status and
contraceptive method switching. All other sociodemographic, background and fertility
covariates do not appear to be associated with switching in the full model, except for parity
which is marginally statistically associated with contraceptive method switching.
Stability in Use and Nonuse of Contraceptive Methods: Multivariate Discrete-Time Event
History Results
Discrete-time multinomial logistic regression models are estimated to examine women’s
contraceptive outcomes with the emphasis on union status at the start of the observation period.
Unlike the prior analyses I separate the category of women who are non-switchers into two
groups: stable users and stable nonusers. I compare the odds of stable use versus stable nonuse,
switching versus stable nonuse, and switching versus stable use. All multinomial logistic
analyses include a time-varying indicator, which is the number of months women remain in their
respective union statuses and are at risk to different contraceptive outcomes. I initially present
zero-order results for all characteristics in predicting the different contraceptive outcomes of
women (Table 2.3). I then present the association between union status and contraceptive
switching controlling for sociodemographic, background and fertility indicators in separate
models before the final model that shows the association of union status on contraceptive
switching net of all characteristics.
The first column of Table 2.3 shows that at the zero-order level when stable nonuse is
presented as a competing risk, single women relative to married women have a 154% higher risk
of stable use. Cohabiting and married women share similar risk of stable use compared to stable
nonuse. In results not shown, single and cohabiting women have similar odds of stable use than
stable nonuse. Sociodemographic, background and fertility characteristics are also related to the
43
risk of stable use relative to stable nonuse. As age increases the risk of stable use compared to
stable nonuse decreases by 12%. Non-Hispanic black and multiracial women compared to nonHispanic white women have lower risks of stable use relative to stable nonuse with 44% and
56% respectively. Women whose mothers have a college degree or higher have an 84% higher
risk of stable use compared to stable nonuse. As age increases at sexual debut, the risk of
contraceptive method switching is reduced by 6%.
Union status is also associated with contraceptive method switching relative to stable
nonuse (Table 2.3, column 2). Compared to married women, cohabiting and single women have
high risks (202% and 286%) of switching. As age increases the risk switching compared to
stable nonuse is reduced by 5%. Mother’s education is also associated with the risk of switching
compared to stable nonuse. Women whose mothers have a college degree or higher have a 78%
higher risk of switching relative to stable nonuse.
Cohabiting and single women compared to married women have 68% and 51% high risks
of switching when stable use is treated as a competing risk (Table 2.3, column 3). Age is also a
predictor of switching relative to stable use. With each year increase in age women’s risk of
switching is reduced by 8%. Non-Hispanic multiracial women compared to non-Hispanic white
women have a 48% lower risk of switching relative to stable use.
In Table 2.4 union status and sociodemographic and background characteristics are
included in the model. Union status differences in predicting contraceptive outcomes are not
entirely consistent with zero-order results. The only consistent result in this analysis indicates
that single women compared to married women have a 108% higher risk of stable use relative to
stable nonuse (Table 2.4, column 1). In addition to this significant association, religious
affiliation is also statistically associated with stable contraceptive use compared to stable nonuse.
44
Women raised as Protestants and Catholics have higher risk of 52% and 53% respectively of
stable use relative to stable nonuse.
The association between union status and contraceptive outcomes (i.e., switching
compared to stable nonuse) is mediated by sociodemographic and background characteristics
(see Table 2.4, column 2). However, age and race/ethnicity are associated with contraceptive
method switching when stable use is treated as a competing risk. With an increase in age
women’s risk of switching decreases by 10% relative to stable nonuse. Compared to nonHispanic white women, non-Hispanic multiracial women have a 64% lower risk of switching
compared to stable nonuse. Subsequent analyses reveal that when entered separately age
explains the mediation effect.
Union status is not associated with the risk of switching relative to stable use (Table 2.4,
column 3). Like results in the previous column, age and race/ethnicity are associated with these
contraceptive outcomes. As women age increases the risk of contraceptive switching relative to
stable use decreases by 8%. In addition, women who are non-Hispanic multiracial compared to
their non-Hispanic white counterparts, have a 45% lower risk of switching contraception relative
to stable use.
In Table 2.5 union status differences in predicting contraceptive outcomes are consistent
with results at the zero order level when fertility indicators are included in the model. The
effects of these associations are also larger. Single women compared to married women, have a
201% higher risk of stable use relative to stable nonuse (column 1, Table 2.5). Compared to
married women, single and cohabiting women have high risk of switching versus stable nonuse
with 221% and 391% respectively (column 2, Table 2.5). Finally, single and cohabiting women
have 69% and 63% increased risks of switching relative to stable nonuse (column 3, Table 2.5).
45
None of the fertility characteristics are statistically associated with contraceptive outcomes,
except when predicting switching versus stable nonuse. Women with at least one child at
observation have a 64% higher risk of switching relative to stable nonuse.
The full model (Table 2.6) reveals some union status differences in contraceptive
outcomes controlling for sociodemographic, background and fertility characteristics. However,
not all results are consistent with those at the zero-order level. Single women compared to
married women have higher risks of stable use and switching with 151% relative to stable
nonuse (Table 2.6, columns 1 and 2). Religious affiliation is also associated with stable use
relative to stable nonuse. Women raised as Protestants and Catholics have 52% and 53%
reduced risks of stable use compared to stable nonuse. In the full model age and race/ethnicity
are associated with switching relative to stable nonuse. With each increase in age women have a
12% lower risk of switching contraception compared to stable nonuse. Women who are nonHispanic multiracial compared to their white counterparts have a 64% decreased risk of
switching contraception relative to stable nonuse. Parity is also related to switching of
contraception compared to stable nonuse. Women with at least one child at observation have a
153% increased risk of switching contraception relative to stable nonuse. The final column of
Table 2.6 shows no significance in the association between union status and switching relative to
stable use. Further, cohabiting and single women experience similar odds of switching
compared to stable use (results not shown). As women’s age increases the risk of switching
versus stable nonuse is reduced by 9%. Non-Hispanic multiracial women compared to nonHispanic white women have 48% lower of switching contraception rather than being stable
users. Finally, women with children have a 58% increased risk of switching contraception
relative to stable use.
46
Overall, women’s contraceptive outcomes vary based on union status. Single women
more than married women have higher odds of stable use and contraceptive method switching
relative to stable nonuse. Therefore, H3 is partially supported. Single and cohabiting women
share similar odds of stable use relative to stable nonuse and so there is no evidence to support
H4. Results partially support H5 which posit that compared to married women, cohabiting and
single women will have greater odds of switching relative to stable nonuse. In this case only
single women compared to married women are more likely to switch contraception than remain
as stable nonusers. In addition, results from supplemental analyses reveal that single women and
cohabiting women share similar odds of switching relative to stable nonuse (results not shown).
This evidence therefore does not support H6. Hypotheses 7 and 8 are not supported as there are
no differences in the odds of switching versus stable use for single, cohabiting and married
women. Non-Hispanic multiracial women are less likely to switch contraception relative to
stable nonuse compared non-Hispanic white women. The likelihood of switching contraceptive
methods declines with age. Having a child increases the likelihood of contraceptive method
switching relative to stable use and nonuse.
Discussion
Among women in this sample the modal category is stable contraceptors (43%) over a 3
year period, 40% are contraceptive switchers and 17% stable non users. The descriptive results
indicate that significant variations in contraceptive behaviors exist among women across union
status. While more than one-third (35%) of married women switch contraception over the
observation period, half of single (50%) and 53% of cohabiting women switch contraception. A
little more than one-fifth (21%) of married women are stable nonusers during the observation
period compared to less than one in ten cohabiting and single women. Multivariate results
47
indicate that single and cohabiting women have a higher risk of switching (relative to not
switching) compared to married at the zero order level. In the full model race/ethnicity and
parity are marginally associated with contraceptive switching; however, it is age that mediates
the association between union status and contraceptive method switching.
Single women are much more likely than cohabiting and married women to switch
contraception than remain as stable users and nonusers of contraception. Women who are raised
as Protestants and Catholics have lower risks of stable use of contraception compared to stable
nonuse. Sociodemographic and fertility characteristics are also associated the risk of switching
contraception relative stable nonuse. Older women are more likely to be switchers relative to
being stable nonusers. Likewise, non-Hispanic multiracial women compared to non-Hispanic
white women have lower odds of contraceptive method switching relative to stable nonuse.
Women with at least one child are more likely to switch contraception when stable nonuse is
treated as a competing risk.
The results from this investigation provide a contemporary description of contraceptive
method switching and union status in the United States. The results seems to suggest that
contraceptive method switching behavior is driven more in large part by single women who are
on average younger than married women and of similar age to cohabiting women. Generally, all
women regardless of union status seem to share similar switching behavior but are more distinct
when non-switchers are dichotomized into stable users and stable nonusers. Married women
seem to be more at risk of stable nonuse. Older women are less inclined to switch contraception
and this is more typical among married and cohabiting women. Having a child increases the
likelihood of contraceptive method switching.
48
This current investigation provides evidence that supports existing research in the areas
of reproductive and contraceptive behaviors and union status. The results suggest that married
and cohabiting women are becoming more similar in terms of reproductive behaviors (see
Sweeney 2010). Further, the results support prior findings examining the association between
parity and method switching (see Frost et al. 2007b). While studies have consistently included
parity in the analyses of reproductive and contraceptive behaviors (Finer and Zolna 2011; Jacobs
and Stanfors 2013), a limitation is that not much is known about the sociodemographic profile of
women who at the time of observation have not yet had a child (i.e., parity is zero).
Additionally, little is known about contraceptive method switching behavior of women when
parity is zero (see Grady et al. 2002 for exception). This concern is addressed in the following
chapter of the dissertation.
49
CHAPTER III: CONTRACEPTIVE METHOD SWITCHING AND UNION STATUS
AMONG WOMEN WITHOUT CHILDREN
Current Investigation
With the a total fertility rate (TFR) bordering at replacement level fertility in the United
States, the typical woman wants on average two children. However, in order to achieve this goal
effective contraception must be used for approximately 30 years (Frost et al. 2008). While
women in the United States place a high value on having children (McQuillan et al. 2008), there
has been a historical shift towards postponing and limiting childbirth (Kavanaugh et al. 2011).
There are positive and negative results that emanate from this change. First, consensus among
many sociologists is that the delay in childbearing is associated in great part to increased
workforce participation and educational pursuits of women (Kirmeyer and Hamilton 2011). On
the other hand, this historical shift towards delaying childbirth has increased the amount of time
in a woman’s reproductive life course where she is at risk for unintended pregnancy (Kavanaugh
et al. 2011). Therefore, the examination of contraceptive method switching behaviors among
this specific population of women (i.e., those who have yet to have a child or women at parity
zero) is relevant.
A vast body of research on contraceptive and reproductive behaviors of women
throughout the life course has included parity in the analyses (Frost et al. 2007 b; Grady et al.
2002; Jacobs and Stanfors 2013; Jones et al. 2012; Kavanaugh et al. 2011; Manlove et al. 2014).
Data from the NSFG 2006-10 on contracepting reproductive age women, who have not yet had a
child, indicate that more than half (53%) are current users of the pill, almost one-quarter (23%)
are condom users, one in ten are hormonal users and 5% are sterilized (Jones et al. 2012). Frost
and colleagues in their analyses of factors associated with contraceptive use and nonuse find that
50
women with one birth are more likely to switch contraception than women at parity zero. In
another study, results indicate that women with at least a child are more likely to be users of
LARC than women without children at the time of observation (Kavanaugh et al. 2011). Among
a nationally representative sample of young dating adults, contraceptive use is negatively
associated with having a child (Manlove et al. 2014). Grady and colleagues (2002) find that
among unmarried women, those who are at parity zero at the start of interval, relative to women
with at least a birth, are less likely to switch to a long-term method of contraception. Similarly,
unmarried women without children at the start of the observation period are more likely than
women with a child to switch from the pill or no method to less effective method reversible
methods. Among married women, not having a child reduces the risk of switching to
sterilization. Further, married women also have decreased risk of switching from no method to
the pill. Results from chapter II of the dissertation sheds light on the need to further explore this
group of women. Approximately one-third of women have not had a child at observation and
there is variation in parity by union status. One-fifth (20%) of married women are at parity zero
compared to 43% and 71% of cohabiting and single women (see Table 2.1).
Therefore, this chapter examines the association between women’s union status and
contraceptive method switching and non switching outcomes (i.e., stable users and stable
nonusers) among women at parity zero.
Based on the prior literature related to contraceptive use and contraceptive method
switching, I propose the following hypotheses. The first 2 hypotheses (H 9-10) focus on
contraceptive method switching versus not switching and hypotheses 11 to 14 examine
contraceptive method switching relative to stable use and stable nonuse.
51
Hypotheses
Hypothesis 9:
At parity zero, single women (compared to cohabiting and married
women) will more often engage in method switching.
Hypothesis 10:
At parity zero, cohabiting women (versus married women) will experience
greater odds of method switching.
Hypothesis 11:
Single women (compared to married and cohabiting women) will more
often switch contraception than remain as stable users when parity is zero.
Hypothesis 12:
Single and cohabiting women (compared to married women) will more
likely switch contraception than remain as stable users when parity is zero.
Hypothesis 13:
Single and cohabiting women (compared to married women) are more
likely to switch contraception than remain as stable nonusers when women
are at parity zero.
Data and Methods
The National Survey of Family Growth (NSFG) was conducted from June 2006 to June
2010 and is based on a national probability sample representing the household population of the
United States, ages 15-44 years. This survey comprised 12,279 non-institutionalized women.
This NSFG is appropriate for the purpose of this investigation because of the contemporary
nature of the data set which includes detailed retrospective contraception, marriage and
cohabitation histories as well as socio-demographic variables that are associated with
contraceptive method switching. The NSFG contains a contraceptive method history calendar in
which dates of respondent’s use of contraception in each month during the four years preceding
the interview is recorded. This enables the examination of contraceptive method switching.
Analytic Sample
I examine the patterns of contraceptive method switching and variations by union status.
The NSFG 2006-10 is a national non-probability sample representing the household population
of 12,279 non-institutionalized female respondents, ages 15-44 years. The inclusion of
52
respondents with valid union histories reduces the sample to 10,761. I also include only valid
responses to contraceptive method switching questions from the method history calendar
(N=9,470). The sample is further limited to respondents observed during a three-year period and
not sterilized at start of the observation period (N= 4,674). Following the merging of
contraceptive method switching file and the NSFG data file containing variables used as control
covariates the sample stands at 3,122 respondents. The final restriction includes limiting the
sample to women ages 21-44 years with non-missing responses on age, race/ethnicity, education,
poverty status, religious affiliation, mother’s education, age at first sex and parity. The final
analytic sample consists of 728 respondents.
Dependent Variable
The dependent variable for this research is contraceptive method switching. I use
respondents’ retrospective reports of their contraceptive method history over the last 3 years.
Predicting method switching behavior takes two forms namely: a) dichotomous variable
(switching versus not switching); and b) categorical variable (stable nonusers, stable users and
switchers). Contraceptive method switching is measured based on women’s report of different
use of contraception in consecutive months during observation uninterrupted by any period of
abstinence. Contraceptive method switching is also measured by a change from nonuse of
contraception in one month to use of contraception in another month or vice versa.
Independent Variable
The main independent variable is union status. Respondent’s union status at the time of
interview does not tell us about the union context during contraceptive method switching.
Therefore, union status is measured as a time invariant characteristic using retrospective dates of
marital and cohabitation histories at the time of contraceptive switching. At the start of the
53
observation period respondents are either in a marital or cohabiting union. If there are no marital
or cohabiting dates that correspond to the commencement of the observation period, respondents
are classified as single. Importantly, it is noted that being married at observation does not denote
first marriage but any marriage at that time. Similarly, being in a cohabiting union at observation
does not indicate that women have never been married or in prior cohabiting unions and being
single does not denote never married and never cohabited.
Sociodemographic Characteristics
Several variables are included in the analyses as controls because of their potential
confounding associations between union status and contraceptive method switching. I include
the respondent’s age. Age is measured at the time of interview and is a continuous variable
measured in years. Respondent’s race/ethnicity is also included and I use a NSFG recode based
on the 1997 Office of Management and Budget (OMB) standards to create a four category
response measure: non-Hispanic white (reference category), non-Hispanic black, Hispanic, and
multiracial. I also include respondent’s education (measured at time of interview) in the analyses
and coded into four categories: less than high school degree, high school/GED (reference
category), some college and college degree or higher. Poverty status is based on the federal
eligibility criteria for subsidized family planning services and respondents are grouped into
dichotomous variable: 1= at/above the poverty line (federal poverty is >= 100%) and 0 = below
the poverty line (federal poverty is 0-99%).
Background Characteristics
Family type is a dichotomous variable based on a NSFG recode of intact status of
childhood family where respondents have two biological/adoptive parents from birth or
childhood family where respondents have anything other than respondents two
54
biological/adoptive parents from birth. Respondent’s religious affiliation while growing up is
included in the analysis. This coded into four categories: no religion (reference category),
Protestant, Catholic and other religious affiliation. Respondent’s mother’s education is taken
into account and categorized as: less than high school degree, high school/GED (reference
category), some college and college degree or higher.
Fertility Characteristic
Age at first sex is a recoded continuous variable based on the question: “Whether
respondent has ever had sexual intercourse with a male (even if before menarche)?”
Measure of Time
I include a continuous measure of time, measured in months. The variable counts the
number of months of women’s contraceptive use during the observation period which is not
interrupted by any period of sexual abstinence. Respondents are censored (removed from
analyses) if there is change in union status during the observation period and for those
respondents who do not switch contraception.
Analytic Plan
Descriptive analyses show the percentage of women at parity zero by union status and
their contraceptive outcomes (i.e., switching, stable use and stable nonuse) during the three (3)
year observation period. To determine the prevalence and predictors of switching behavior I
provide weighted means, proportions and standard errors based on the analytic sample. In
addition, differentials in sociodemographic, background and fertility characteristics according to
union status are presented. In the final step, both logistic and multinomial logistic regression
models are used to estimate discrete-time event models that examine the association between
union status and contraceptive method outcomes. In all regression models zero order results are
55
presented followed by the inclusion of sociodemographic and background variables (model 1),
fertility characteristics (model 2) and full model (model 3) with all covariates.
Results
Descriptive Results: Women at Parity Zero
During the three year observation period almost two-fifth (39%) of women who have not
yet had a child switch contraceptive methods while 17% are stable nonusers and almost 44% are
stable users (Table 3.1). Results by union status indicate that more single and cohabiting women
who do not have children are switch contraception compared to their married counterparts. For
married women at parity zero just over one-quarter (28%) switch contraception compared to
almost half (46%) of single and 56% of cohabiting women also at parity zero. Approximately
one-third (32%) of married women without children are stable nonusers compared to about 6%
of cohabiting and single women at parity zero.
The average age in the sample of women at parity zero was 30 years. On average,
married women at parity zero are older than their cohabiting and single counterparts. At parity
zero, a higher proportion of married and single women are non-Hispanic white compared to
cohabiting women. The modal education category is a college degree or higher (55%). Single
women without children are more likely to be highly educated (i.e., college degree or higher)
(61%) compared to married women (56%) and cohabiting women (28%) at parity zero. The
majority (92%) of women who do not have children at the time of observation are at or above the
federal poverty line. This high percent is also reflected in the distribution across union status
where most married (98%), cohabiting (85%) and single women (87%) are at or above the
federal poverty line. Background characteristics of women at parity zero also vary by union
status. More married (76%) and single (71%) women compared to 55% of cohabiting women
56
report being raised in a two biological/adoptive parent household prior to age 18. A greater
proportion of cohabiting women without children at observation report being raised Catholics
compared to their married and single counterparts. Also more mothers of cohabiting women
compared to mothers of single and married women have a high school diploma. Additionally, a
smaller proportion of mothers of single women compared to those of cohabiting and married
women have less than high school education (8%) and a larger proportion have a college degree
or higher (32%). The average age at first sexual intercourse in the sample was 18 years.
Cohabiting women at parity zero engage in sexual intercourse at a younger age than those of
married and single women.
Switching Contraceptive Methods: Bivariate and Multivariate Results (Women at Parity Zero)
Among women at parity zero, I present the zero order and multivariate effects of
women’s union status on the likelihood of switching contraceptive methods (Table 3.2). The
table displays odds ratios of contraceptive method switching compared to not switching. A timevarying indicator (i.e., months) is included in all analyses. At the zero-order level single and
cohabiting women have higher risks of switching contraception but the odds are not statistically
significant. There are no differences in the odds of switching for single and cohabiting women
without children (results not shown). Age and mother’s education are related to contraceptive
method switching. With each year increase in age women at parity zero have a 10% lower risk
of switching contraception.
In Table 3.2 each subsequent model includes different sets of covariates:
sociodemographic and background (model 1), and fertility characteristic (model 2) to the union
status model predicting method switching. A final model is also displayed containing all control
variables.
57
At the zero order level single and cohabiting women share similar risk of contraceptive
method switching. Additional analyses indicate that single and cohabiting women also share
similar risk of switching contraception (results not shown). Only age is statistically related
method switching. With each increase in age the risk of switching contraception is reduced by
10%.
With the inclusion of sociodemographic and background variables in model 1 for women
at parity zero, union status is not significantly related to the risk of switching contraception. In
this model age is also associated with contraceptive method switching such that a yearly increase
in age reduces contraceptive method switching by 8%.
Union status is not associated with contraceptive method switching when age at first
sexual intercourse is added to model 2. In the full model (model 3) union status remains
statistically unrelated to contraceptive method switching for women at parity zero. There are
statistical similarities in the odds of switching for cohabiting and married women as well as
single and cohabiting women, all at parity zero (results not shown). This is due mostly to the age
differences among women by union status. Age continues to be associated with method
switching such that as yearly increases in age occur, the risk of switching is lowered by 9%. In
the full model age at first intercourse becomes associated with contraceptive method switching.
The older the age at first sex, women’s risks of switching increases by 1%.
Overall, the results indicate that after controlling for key covariates, union status of
women at parity zero is not associated with contraceptive method switching. The odds of
switching contraception for single and cohabiting women are similar (results not shown).
Therefore, the findings do not support H9. There is no evidence in support of H10 as cohabiting
and married women without children share similar odds of method switching.
58
Stability in Use and Nonuse of Contraceptive Methods: Multivariate Discrete-Time Event
History Results (Women at Parity Zero)
I estimate discrete-time multinomial logistic regression models to examine contraceptive
outcomes on union status at the start of the observation period among women without children.
Previous analyses did not separate the stable category into stable users and stable nonusers of
contraception. I evaluate the odds of stable use versus stable nonuse, switching versus stable
nonuse and switching versus stable use. All multinomial logistic regression models include a
time-varying indicator (measured in months). In Table 3.3 zero-order results for all covariates in
predicting the different contraceptive outcomes are presented. I then present the association
between union status and contraceptive method switching net of sociodemographic, background
and fertility indicators before the final model that displays the association of union status on
method switching net of all characteristics.
In Table 3.3 (column 1), for women at parity zero, the first column shows that at the zeroorder level compared to married women, cohabiting and single women have high risks of stable
use relative to stable nonuse (488% and 596% respectively). Single and cohabiting women at
parity zero share similar risks of stable use relative to stable nonuse (results not shown).
Sociodemographic, background and fertility indicators are also related to women’s stable use of
contraception relative to stable nonuse. With each yearly increase in age women’s risk of stable
use of contraception is reduced by 20% relative to stable nonuse. Having less than high school
education relative to having a high school diploma increases the risk of stable use compared to
stable nonuse by 186%. Women raised as Catholics, compared to those with no religious
affiliation, have a 46% lower risk of stable use relative to stable nonuse. Women whose mothers
have a college degree or higher relative to those women whose mothers have a high school
59
diploma, have a 303% increased risk of stable use compared to stable nonuse. As the age at first
sexual intercourse increases, women’s risk of stable use relative to stable nonuse is lowered by
7%.
Union status is also associated with contraceptive method switching relative to stable
nonuse among women at parity zero (Table 3.3, column 2). At the zero order level, cohabiting
and single women compared to their married counterparts have a higher risk of switching
contraception compared to stable nonuse (1058% and 898% respectively). Additional analysis
indicate that single and cohabiting women at parity zero share similar risk of switching
contraception relative to stable nonuse (results not shown). Age, race/ethnicity, and mother’s
education are also related to contraceptive switching relative to stable nonuse of the zero order
level. With each year increase in age women’s risk of switching contraception, relative to stable
nonuse, decreases by 13%. Compared to non-Hispanic white women, Hispanic women have a
64% lower risk of switching relative to stable nonuse. Additionally, women whose mothers have
a college degree or higher (compared to women’s mothers with a high school diploma), have a
266% higher risk of switching contraception compared to stable nonuse.
Among women without children at the observation period, at the zero order level union
status is not associated with the risk of switching relative stable use (Table 3.3, column 3).
Further analysis show that single and cohabiting women at parity zero also share similar risk of
switching contraceptive methods relative to stable use (results not shown). In addition, none of
the sociodemographic, background and fertility characteristics are associated with the risk of
switching contraception relative to stable use.
Union status and sociodemographic and background characteristics are included in Table
3.4 for women at parity zero. However, unlike the results of the zero order models, union status
60
is not associated with stable use relative to stable nonuse (Table 3.4, column 1). Education is the
only covariate associated with stable use versus stable nonuse. Women with some college
education compared to those with a high school diploma, have a 108% higher risk of stable use
compared to stable nonuse of contraception.
Union status is marginally associated with the risk of switching contraception when
stable nonuse is treated as a competing risk (Table 3.4, column 2). Among women at parity zero,
single women have a 311% higher risk of switching relative to stable nonuse (p <.10). Age,
race/ethnicity and mother’s education are also associated with switching behavior. With each
year increase in age women’s risk of switching contraception decreases by 21%. Compared to
non-Hispanic white women, non-Hispanic multiracial women have a 78% lower risk of
switching contraceptive methods compared to stable nonuse. Women whose mothers have some
college education compared to women whose mothers have a high school diploma have a 35%
higher risk of switching compared to stable nonuse.
The inclusion of age at first sex does not explain the association between union status and
contraceptive outcomes (i.e., stable use versus stable nonuse and switching versus stable nonuse)
among women at parity zero as the results are consistent with those in the zero-order model
(Table 3.5).
The full model in Table 3.6 indicates that union status is not associated with
contraceptive outcomes for women who have not yet had a child, net of covariates. Age explains
the association between union status and stable use (relative to stable nonuse). For each
additional year of age the risk of stable use relative to stable nonuse decreases by 18%. In
addition, age and race/ethnicity explains away the association between union status and
contraceptive method switching when stable nonuse is treated as a competing risk. With each
61
yearly increase in age, women have a 24% lower risk of switching contraception compared to
stable nonuse. Compared to non-Hispanic white women, non-Hispanic multiracial women have
a 74% decreased risk of switching relative to stable nonuse.
In summary the findings among women at parity zero indicate that single and cohabiting
women compared to married women share similar odds of stable use and switching
contraception relative to stable nonuse as well as similar odds of switching contraceptive relative
to stable use. Subsequent analysis reveal that single women at parity zero compared to
cohabiting women at parity zero have lower risk of switching contraception relative to stable use
(results not shown). This findings runs counter to H11. Other results reveal that single and
cohabiting women who have yet to have a child, relative to their married counterparts, have
greater odds of switching compared to stable nonuse. However, this association is not significant
and therefore H12 and H13 are not supported. There is no statistical difference in the odds of
stable use compared to stable nonuse for women without children regardless of union status.
Discussion
Over one-third of women at parity zero switch contraception, and descriptive analyses
indicate women who switch contraception vary by union status. Just more than one-quarter of
married women (28%) switch contraception compared to 46% for single and 56% for cohabiting
women. There is also a wide variation in the proportion of women at parity zero who do not
switch contraception. For married women who have not yet had a child, the difference between
stable nonusers and stable users is small (34% versus 40%). Single women are more likely to be
stable users of contraception compared to cohabiting and married women.
Multivariate analysis indicates that there is no statistical association between union status
and contraceptive method switching for women at parity zero. This is explained largely by
62
women’s age. Similar multivariate findings are illustrated for women when predicting different
contraceptive outcomes. At the zero order level, childless single and cohabiting women have
high risks of stable use and contraceptive method switching relative to stable nonuse compared
to their married counterparts. However, net of all covariates, this association loses significance.
In examining both outcomes (stable use versus stable nonuse and switching versus stable nonuse) age of women explains the non-significance of the relationships while race/ethnicity also
explains, in part, the non-association between switching and stable nonuse.
63
CHAPTER IV: UNION STATUS AND CONTRACEPTIVE USE AMONG
CONTRACEPTIVE METHOD SWITCHERS
Current Investigation
The dynamics of understanding contraceptive behavior is not limited to one particular
stage in the life course nor does it entail one specific behavior that is practiced by all individuals.
From a life course perspective (Elder 1998) contraceptive behavior is a fluid process which
affects individuals differently based on context and time. Jaccard (2010) posits that there are
four components of contraceptive behavior namely, contraceptive choice, which refers to
choosing and acquiring a contraceptive method; accuracy of use, which refers to the correct use
of contraception; consistency of use, which implies the use of contraception during every
occurrence of sexual intercourse; and contraceptive method switching, which focuses on the
change of one method of contraception to another.
Contraceptive method switching research grew out of extensive work on contraceptive
discontinuation. There are at least three types of contraceptive discontinuation associated with
inconsistent contraceptive use and women’s overall reproductive health. They include
contraceptive failure, abandonment of contraceptive use (while in need of contraception) and
method switching. Contraceptive failure is based on the probability of pregnancy during the first
year of typical use of any contraception and highlights the effectiveness of the method among
persons who may fail to use methods correctly or consistently, if any at all (Trussell 2009).
Abandonment of contraception is the stopping of any contraceptive use because of contraceptive
failure or for any other reason than non-exposure or trying to get pregnant (Vaughan et al. 2008).
The use of different contraceptive methods in consecutive months is defined as method
switching (Grady et al. 2002). While contraceptive failure and abandonment are two important
outcomes to study because they lead to immediate risk of unintended pregnancy, a key limitation
64
when analyzing these outcomes is that women’s choice of new method are not taken into
account. Contraceptive method switching often results in gaps in protection and higher failure
rates as individuals learn a new method, and, if the new method is less effective, increases the
risk of unplanned pregnancy (Jaccard 2010). However, this contraceptive behavior also
positively affects the reproductive health outcomes of women (Steele and Curtis 2003). More
attention needs to be geared towards contraceptive method switching as opposed to contraceptive
discontinuation in order to gain greater knowledge and understanding of the processes involved
in contraceptive behavior.
Studies related to contraceptive method switching in the United States are rare and dated.
Most of these studies are based on convenient samples drawn from health clinics, disadvantaged
neighborhoods, and from a higher risk of STI group of women. Sang-Haghpeykar et al. (1995)
find that women are more likely to switch from more effective contraception (Depo-Provera) to
less effective methods (condoms and pills) because the former method is mainly used for birth
spacing. Similar research finds that among the discontinuers of Depo-Provera who report
switching to another method, 55% switch to oral contraceptives while 31% use condoms
(Davidson et al. 1997). Santelli et al. (1995) provide evidence on the likelihood of contraceptive
method switching among women at greater risk for HIV/AIDS. Women who use only pills or
condoms at last intercourse are likely to switch to dual methods (pills and condoms) at most
recent intercourse. While the use of convenient samples is a key limitation, another important
limitation is that there is no analyses to determine that characteristics of individuals who switch
because the population is homogenous by composition.
Using data collected in 2004, Frost and colleagues (2007b) examine patterns of
contraceptive method use. The findings show an overall profile of contraceptive switchers
65
compared to stable users, stable nonusers, and those who experience gaps in contraceptive use
regardless of being at risk. One-fifth of all married women (at time of interview) switch
contraception while for cohabiting women, almost one-third (29%) switch contraception.
Approximately 31% of young adults ages 18-24 switch contraception compared to 16% among
women 35-44 years. Differentials by race/ethnicity show that one-quarter of all non-Hispanic
whites and blacks switch contraception while approximately 20% of all Hispanic and multiracial
women switch contraception. Although this research is very useful in drawing attention to the
issue of contraceptive method switching in the United States, generalizability cannot be assumed.
This study focuses on use of any contraceptive method (whether for all of the past year or for one
or more months during that year). Finally, union status is captured at time of interview following
changes in contraceptive behaviors.
Research has also explored the patterns of contraceptive use among switchers. Frost and
colleagues (2007a) find that more than one-quarter (26%) of women switch from one method of
contraception to another during a year. More specifically, of the 6% that switch to hormonal or
long-acting methods, 4% are initial barrier/traditional users and 2% are former nonusers. Eleven
percent of women switch to barrier/traditional methods (9% are initial hormonal users and 2%
former nonusers). Of the 10% of women who switch to nonuse of contraception, 4% and 6% are
former hormonal/long-acting and barrier/traditional users respectively. Grady et al. (2002) also
provide multivariate results of contraceptive switching patterns of married and unmarried
women. Older married women who are initial nonusers are less likely than younger married
women to switch to any method. As it regards race, black women, compared to their white
counterparts, are more likely to switch from a method to sterilization. They are also more likely
to switch from the pill to no method and vice versa. Hispanic women on the other hand are less
66
likely than white women to switch to sterilization, with the exception being that they are former
nonusers. Unmarried older women compared to younger unmarried women are less likely to
switch to the condom. Black unmarried women compared white unmarried women are less
likely to switch from any method use to either the pill or less-effective reversible methods. Both
studies highlight significant variations in contraceptive switching behavior, however, they are
limited in in terms of generalizability (Frost et al. 2007 a) and focus on marital status (Grady et
al. 2002).
Frost and colleagues (2007a) classify more effective methods as pills, condoms,
injectable, patch/ring, IUD/implants while less effective methods include condoms, diaphragm,
spermicides, withdrawal and natural family planning. Trussell and Vaughn (1999) also provide a
hierarchy for contraceptive methods: sterilization, implant, injectable, IUD, pill, condom, and
other reversible. Based on these studies I classify methods of contraception used following a
switch as: most effective (i.e., sterilization and hormonal methods 2), pill, condom and least
effective (none and other methods of contraception). Hormonal methods are grouped with
sterilization because they are the most effective and do not require a daily activity for use. The
hormonal methods that are not the pill also have a relatively low rate of use (Jones et al, 2012;
Finer et al. 2012). The use of pills and condoms are analyzed separately because of their high
prevalence rate among women in the U.S. (Jones et al. 2012). More than one-third (34%) of
currently married women use the pill and condom, almost half (48%) of currently cohabiting
women use both methods and more than two-thirds (68%) of never-married, not cohabiting
women use the pill and condom. Finally, while both methods are user-dependent, use of the pill
does not need the intervention of women’s partners; in contrast, the use of condoms is dyadic
and coitus dependent. Nonuse of contraception and ‘other’ methods are grouped into one
2
Hormonal methods do not include the pill, which is treated as a contraceptive method on its own.
67
category. The ‘other’ methods include ineffective methods such as withdrawal. It must be noted
that respondents in this category are comprised mostly of those who report nonuse of
contraception.
Based on prior literature, I propose the following hypotheses in order to examine the
contraceptive switching behavior among women across union status.
Hypotheses
Hypothesis 15:
Single women compared to cohabiting and married women, are more
likely to switch to pill and condom use relative to most effective methods
of contraception.
Hypothesis 16:
Cohabiting women compared to married women are more likely to switch
to the pill and condom relative to most effective methods of contraception.
Hypothesis 17:
Single women compared cohabiting and married women are more likely to
switch to the pill and condom relative to least effective methods.
Hypothesis 18:
Cohabiting women compared married women are more likely to switch to
the pill and condom relative to least effective methods.
Data and Methods
The National Survey of Family Growth (NSFG) was conducted from June 2006 to June
2010. This survey comprised 12,279 non-institutionalized women, ages 15-44 years. This NSFG
is appropriate for this dissertation because of the contemporary nature of the data set which
includes detailed retrospective contraception, marriage and cohabitation histories as well as
socio-demographic variables that are associated with contraceptive method switching. The
NSFG contains a contraceptive method history calendar in which dates of respondent’s use of
contraception in each month during the four years preceding the interview is recorded. This
enables the examination of method switching. To date no other nationally representative data
offers these advantages.
68
Analytic Sample
I examine the patterns of contraceptive method switching and variations by union status.
The NSFG 2006-10 is a national non-probability sample representing the household population
of 12,279 non-institutionalized female respondents, ages 15-44 years. The inclusion of
respondents with valid union histories reduces the sample to 10,761. I also include only valid
responses to contraceptive method switching questions from the method history calendar
(N=9,470). The sample is further limited to respondents observed during a three-year period, not
sterilized at the start of the observation (N= 4,674). Following the merging of contraceptive
method switching file and the NSFG data file containing variables used as control covariates the
sample stands at 3,122 respondents. The sample is also limited to women ages 21-44 years with
non-missing responses on age, race/ethnicity, education, poverty status, religious affiliation,
mother’s education, age at first sex and parity. The final restriction to the sample captures only
respondents who switch contraception making the analytic sample 1,899 respondents.
Dependent Variable
The dependent variable for this research is contraceptive method switching. I use
respondents’ retrospective reports of their contraceptive method history over the last 3 years.
Predicting type of contraception used takes the form of a categorical variable (most effective sterilization and hormonal methods 3; pill; condom; and least effective (none and ‘other’
methods 4).
Independent Variable
The main independent variable is union status. Respondent’s union status at the time of
interview does not tell us about the union context during contraceptive method switching.
3
Hormonal methods include Depo-Provera, Hormonal implant, IUD, Lunelle injectable, contraceptive patch and vaginal
contraceptive ring.
4
‘Other’ methods include withdrawal, rhythm, safe period, female condom, diaphragm, foam, and cream.
69
Therefore, union status is measured as a time invariant characteristic using retrospective dates of
marital and cohabitation histories at the time of observation. At the start of the observation
period respondents are either in a marital or cohabiting union. If there are no marital or
cohabiting dates that correspond to the commencement of the observation period, respondents
are classified as single. Women in a marital union at observation do not denote first marriage
but any marriage at that time. Also, cohabiting women at observation may have been in previous
marital and cohabiting unions and single women may have been in prior marital and cohabiting
unions as well as they may have never been married and/or have never cohabited.
Sociodemographic Characteristics
Several variables are included in the analyses as controls because of their potential
confounding associations between union status and contraceptive method switching. I include
the respondent’s age. Age is measured at the time of interview and is a continuous variable
measured in years. Respondent’s race/ethnicity is also included and I use a NSFG recode based
on the 1997 Office of Management and Budget (OMB) standards to create a four category
response measure: non-Hispanic white (reference category), non-Hispanic black, Hispanic, and
multiracial. I also include respondent’s education (measured at time of interview) in the analyses
and coded into four categories: less than high school degree, high school/GED (reference
category), some college and college degree or higher. Poverty status is based on the federal
eligibility criteria for subsidized family planning services and respondents are grouped into
dichotomous variable: 1= at/above the poverty line (federal poverty is >= 100%) and 0 = below
the poverty line (federal poverty is 0-99%).
70
Background Characteristics
Family type is a dichotomous variable based on a NSFG recode of intact status of
childhood family where respondents have two biological/adoptive parents from birth or
childhood family where respondents have anything other than respondents two
biological/adoptive parents from birth. Respondent’s religious affiliation while growing up is
included in the analysis. This coded into four categories: no religion (reference category),
Protestant, Catholic and other religious affiliation. Respondent’s mother’s education is taken
into account and categorized as: less than high school degree, high school/GED (reference
category), some college and college degree or higher.
Fertility Characteristic
Age at first sex is a recoded continuous variable based on the question: “Whether
respondent has ever had sexual intercourse with a male (even if before menarche)?”
Respondent’s parity at observation is included in the analyses. I use the NSFG recode, which
captures the total number of live births including multiple births, to create a dichotomous
variable: 1= have at least one child and 0= have zero births.
Measure of Time
I include a continuous measure of time, measured in months. The variable counts the
number of months of women’s contraceptive use during the observation period which is not
interrupted by any period of sexual abstinence. Respondents are censored (removed from
analyses) if there is change in union status during the observation period and for those
respondents who do not switch contraception at time of interview.
71
Analytic Plan
In this section of the dissertation descriptive analyses are presented on the types of
contraceptive methods used among contraceptive method switchers, (most effective- sterilization
and hormonal methods); pill; condom; and least effective methods (nonuse and ‘other’ methods).
Differentials in sociodemographic, background and fertility characteristics are also presented.
Contraceptive method switching patterns are presented to show variation in the use of
contraception following a switch and also the methods that women use at the start of the
observation period. I present multinomial logistic regression to estimate discrete-time event
history models that examine the association between union status and types of contraception
used among contraceptive method switchers. Finally, supplemental multivariate analyses of
women who at the start of the observation period are pill, condom and least effective users are
shown.
Results
Descriptive Results – All Women Who Switch Contraception
Table 4.1 shows descriptive statistics for women who switch contraception over the three
year observation period. The first set of results present the method women switched to. Women
who switch to hormonal methods and sterilization each accounts for 10% of the proportion of
switchers. One-fifth (20%) of women switch to most effective methods (i.e., sterilization and
hormonal methods). A little more than one-tenth (14%) switch to the pill and 21% switch to
condoms. About 8% of women switched to ‘other’ methods (i.e., withdrawal, calendar rhythm,
natural family planning, foam, cream, and sponge) and more than one-third (36%) of women
switch to none use. Thus, almost half (45%) switch to least effective methods (i.e., none and
other methods). Analyses highlighting types of contraception used according to union status
reveal that married women (25%) are more likely to switch to most effective methods compared
72
to cohabiting women (15%) and single women (11%). More single women, compared to
married and cohabiting women, switch to pill and condom. A greater proportion of cohabiting
women (55%), compared to married (45%) and single (42%) women switch to least effective
methods.
Differentials in sociodemographic, background and fertility characteristics according to
union status among contraceptive switchers are presented in Table 4.1. The average age in the
sample was 31 years. Married and single women are more likely to be non-Hispanic white
compared to cohabiting women. The modal category of education is college degree or higher.
However, there is variation according to union status. Compared to married and single women, a
greater percent of cohabiting women have less than high school education and a low proportion
of cohabiting women have a college degree or higher. More married women (45%) have a
college degree or higher compared to single (44%) and cohabiting (10%) women. Regardless of
union status, the majority of women are at or above the federal poverty line. Married women
(72%) more than cohabiting (52%) and single women (64%), report they were raised in a two
biological/adoptive parent household prior to age 18. Most women report being raised as
Catholics (45%) and more than one-third (36%) report being raised as Protestants. This pattern
is similar across union status. Mothers of cohabiting women more than mothers of single and
married women have a high school diploma. On the other hand, more mothers of single women
compared to their married and cohabiting counterparts, have some college as well as a college
degree or higher. The average age at first sex for women in the sample is 18 years. On average,
cohabiting women engage in first sexual intercourse at younger ages than married and single
women.
73
Contraceptive Switching Patterns
Table 4.2 shows the patterns of contraceptive method switching for women who switch
contraception during the observation period. At the start of the observation period the most
common methods among switchers were as follows: more than one-third of women (36%) did
not use a contraceptive method, 27% used the pill and 19% used the condom (results not shown).
Among pill users at the start of the observation period, half (50%) switch to nonuse and more
than one-fourth (26%) switch to use of the condom. Among initial condom users the most
common option is to switch to nonuse (65%). A little more than 10% of condom users at the
start of the observation period switch to sterilization and almost 10% switch to use of the pill.
For women who start as the most effective users ( hormonal method users), 43% switch to
nonuse, 26% switch to the condom and 18% switch to use of the pill. Among women who were
not using a method at the initial point of observation, the most common methods were condom
(30%) and pill (28%).
The next series of tables (Tables 4.3-4.5) present the same contraceptive use dynamics
separately for married, cohabiting, and single women. Table 4.3 highlights variations in
switching patterns for married women. The distribution patterns of switching for married
women are similar to that for the overall sample. Over one-third (36%) of all married women
who switch contraception, switch to nonuse, less than one-fifth (19%) switch to the condom,
11% switch to use of hormonal methods, and 13% each switch to the pill and sterilization.
Switching outcomes vary greatly based on method of contraception at start of observation period.
In results not shown, among married women at the start of the observation period, 38% are
nonusers, 25% are pill users and 18% are condom users. Among married women using the pill
at the start of the observation period, half (52%) switch to nonuse, 22% switch to the condom,
74
9% switch to sterilization, and 8% switch to hormonal methods. For married women who use
the condom at the start of the observation period, two-thirds (66%) switch to nonuse followed by
16% who switch to sterilization, and 7% each switching to the pill and hormonal methods. For
married women who use hormonal methods at the start of the observation period the largest
proportion switch to nonuse (46%) followed by the condom (24%) and the pill (14%). Among
married women who at the start of the observation are not using any method of contraception,
28% of women in the sample switch to condom, one-quarter switch to the pill and 17% opt for
sterilization.
The switching patterns of cohabiting women are displayed in Table 4.4. Among
cohabiting women who switch contraception, almost half (47%) switch to nonuse, one-fifth
(20%) switch to the condom, and 12% switch to the pill. Differentials in switching patterns are
also identified based on the method of contraception used at the start of the observation period by
cohabiting women. In supplemental analyses, results reveal that most cohabiting women are
nonusers at the start of the observation period, accounting for 30% and this is followed by pill
users with 27% and condom users with 21% (results not shown). Table 4.4 shows that
cohabiting women who use the pill at the start of the observation period switch mostly to nonuse
(72%) followed by the condom (12%). For cohabiting women who are users of condoms at the
start of the observation period, more than two-thirds (69%) switch to nonuse, 15% switch to
hormonal methods and 11% switch to the pill. Further, among cohabiting women who are
nonusers at the start of the observation period, over half (51%) switch to the condom, almost
one-fifth (18%) switch to the pill and 15% switch to the pill
Table 4.5 presents differentials in contraceptive switching patterns for single women
during the observation period. About one-third (34%) of single women who switch
75
contraception, switch to nonuse of contraception. Further, over one-quarter (27%) of single
women switch to use of the condom and 17% switch to the pill. In additional analyses (results
not shown), among single contraceptive switchers, more than one-third (36%) are pill users at
start of the observation period followed by 30% who are nonusers and 19% who are condom
users. Table 4.5 indicates that single women who are pill users at the start of observation period
switch mostly to nonuse (42%) and the condom (38%). In addition, among those who start with
use of the condom, almost 60% switch to nonuse, 17% switch to use of the pill and 11% each
switch to hormonal and other methods. Among single women who switch contraception and
who are nonusers at the start of the observation period, nearly half (44%) switch to the pill, over
one-quarter (26%) switch the condom and 13% switch to hormonal methods.
Summary of Contraceptive Switching Patterns
Overall, most women switch to least effective methods of contraception (i.e. nonuse)
followed by use of the condom. The pattern of contraceptive switching is consistent across
union statuses; however, there are wide variations in switching patterns based on methods of
contraception at the start of the observation period. As expected, sterilization occurs mostly
among married women, especially those who are nonusers at the start of the observation period.
Half of all cohabiting women switch to nonuse and the largest variation in this switch is found
among those who at the start of the observation period used ‘other’ methods of contraception.
Not surprisingly, single women are least likely to switch to sterilization. In addition, among
nonusers at the start of the observation period, most married and cohabiting women switch to
condoms and most single women switch to the pill.
76
Types of Contraceptive Methods Used among Contraceptive Switchers: Multivariate DiscreteTime Event History Results
Discrete-time multinomial logistic regression models are estimated to examine women’s
methods of contraception following a switch with the emphasis on union status at the start of the
observation period. I compare the odds of switching to most effective methods (i.e., sterilization
and hormonal methods) versus least effective methods (no contraception and other methods),
switching to pill versus least effective methods, and switching to condom versus least effective
methods. All multinomial logistic analyses include a time indicator, which is the number of
months women remain in their respective union statuses and are at risk to different contraceptive
outcomes. I initially present zero-order results for all characteristics in predicting different types
of contraception used (Tables 4.6). I then present the association between union status and
methods of contraception controlling for sociodemographic and background controls (model 1)
and fertility indicators (model 2) in separate models before the final model that shows the
association of union status on contraceptive switching net of all characteristics.
The first column of Table 4.6 shows zero-order results for the risk of switching to most
effective methods relative to least effective methods of contraception. Single women compared
to married women have a 53% lower risk of switching to most effective methods relative to least
effective methods. Cohabiting women have 49% lower odds of switching to most effective
methods and this coefficient is marginally significant. In addition single and cohabiting women
share similar odds of switching to most effective methods relative to least effective methods
(results not shown). The next column shows that at the zero order level, union status is not
associated with the risk of switching to the pill relative to least effective methods (Table 4.6,
column 2). In column 3 of Table 4.6 zero order results reveal that single women compared
77
married women have an 83% higher risk of switching to condom relative to least effective
methods. Cohabiting and married women share similar risks of switching to condom relative to
least effective methods. In supplemental analysis results indicate that single women compared to
cohabiting women have a 112% higher risk of switching to condom relative to least effective
methods (results not shown). Further analyses (not shown) focusing on different sets of
contraceptive use comparisons indicate that union status is not associated with the risk of
switching to the pill relative to condom as well as the pill relative to most effect methods.
However, single women compared to married women have a 68% higher risk of switching to
condom relative to most effective methods. Compared to married women, cohabiting women
also have a 284% higher risk of switching to condom relative to most effective methods. Single
women, compared to cohabiting women, have a 127% higher risk of switching to condom rather
than most effective methods (results not shown).
Women’s age, race/ethnicity, education, and religious affiliation are also associated with
contraceptive methods used following a switch. With each year increase in age the risk of
switching to most effective methods relative to least effective methods is increased by 4%.
Compared to non-Hispanic white women, Hispanic women have a 62% lower risk of switching
to the pill relative to least effective methods. The educational attainment of women produces
two different switching pathways for condom use. Women with less than high school education,
compared to women having a high school diploma, have a 45% lower risk of switching to
condom relative to least effective methods. On the other hand, women having some college
degree compared to women with a high school diploma, have a 63% higher risk of switching to
condom relative to least effective methods. Women raised in ‘other’ religiously affiliated groups
compared to women not raised in any religiously affiliated group have a 29% lower risk of
78
switching to most effective methods than least effective methods. Women who report being
raised in ‘other’ religiously affiliated groups compared to women who have not been raised in
any religious context, have a 482% higher risk of switching to the pill relative to least effective
methods.
The inclusion of sociodemographic and background covariates is presented in Table 4.7.
Union status loses its significance and does not predict switching to most effective methods
relative to least effective methods (Table 4.7, column 1). Unlike at the zero order level, single
and married women now share similar risks of switching to most effective methods compared to
least effective methods. Women’s age is responsible for this mediating effect. With each year
increase in age women’s risk of switching to most effective methods relative to least effective
methods is increased by 3%. Further, additional analysis indicate that single and cohabiting
women share similar risks of switching to most effective methods relative to least effective
methods (results not shown).
Union status is not associated with the risks of switching to the pill relative to least
effective methods (Table 4.7, column 2). This is consistent with zero order results.
Additionally, single and cohabiting are not statistically different from each other as it concerns
the risk of switching to the pill relative to least effective methods (results not shown). However,
religious affiliation is associated with switching to the pill relative to least effective methods.
Having being raised in any religious context (Protestant, Catholic and other) compared to not
being raised in any religious context, increases the risk of women switching to the pill relative to
least effective methods by 145%, 255% and 615% respectively.
Single women compared to married women have a 72% higher risk of switching to
condom relative to least effective methods (Table 4.7, column 3). Subsequent analysis reveals
79
that single and cohabiting women share similar odds of switching to condom relative to least
effective methods (results not shown). Race/ethnicity is also associated with this contraceptive
outcome. Hispanic women compared to non-Hispanic white women have a 50% lower risk of
switching to condom relative to least effective methods. Additional analyses examining different
contraceptive use comparisons following a switch show that union status is not associated with
the risk of switching to: the pill relative to condom; the pill relative to most effective methods
and condom relative to most effective methods (results not shown).
In Table 4.8 union status and fertility characteristics are included in the models to predict
different switching outcomes. Union status is not associated with the risk of switching to most
effective methods relative to least effective methods (Table 4.8, column 1). In results not shown
single and cohabiting women have similar odds of switching to most effective methods to least
effective methods. Women with at least a child at observation have a 101% increased risk of
switching to most effective methods relative to least effective methods. Parity seems to fully
mediate the relationship between single women and switching to most effective methods relative
to least effective methods. This may be due to the fact that only one-third of single women have
a child at observation and therefore switching to most effective methods may not be appropriate
given fertility intentions.
As is the case at the zero order level, union status is not associated with switching to the
pill compared to least effective methods when fertility characteristics are included in the model
(column 2, Table 4.8). In column 3 of Table 4.8 union status is not associated with the risk of
switching to condom relative to least effective methods. Both fertility covariates, age at first sex
and parity, partially mediate the association union status and switching to condom relative to
least effective methods. In subsequent analyses results indicate that union status is not
80
associated with women’s risk of switching to the pill compared to condom as well as switching
to the pill relative to most effective methods. However, single women compared to married
women have a 150% higher risk of switching to condom relative to most effective methods while
cohabiting women compared to married women have a 40% higher risk of switching to condom
relative to most effective methods (p<.10) (results not shown).
The full model (Table 4.9) reveals that union status is only related to risk of switching to
condom relative to least effective methods. Compared to married women, single women have a
63% higher risk of switching to condom relative to least effective methods (Table 4.9, column
3). In subsequent analysis single women compared to cohabiting women have an 84% higher
risk of switching to condom relative to least effective methods (results not shown).
Race/ethnicity is also related to risk of switching to condom. Further analyses (not shown)
examining different sets of contraceptive use comparisons shown that union status is not
associated with switching to the pill compared to condom, switching to pill compared to most
effective methods and switching to condom relative to most effective methods.
While union status is not associated with the risk of switching to most effective methods
and the pill when switching to least effective methods is treated as a competing risk, there are
two covariates associated with these switching outcomes (column 1 and 2, Table 4.9). First,
women with at least one child at observation have an 88% risk of switching to most effective
methods relative to least effective methods. Second, women raised in a religious context, albeit
Protestant, Catholic and other, compared to women not raised in any religious context have high
risks of switching to the pill relative to least effective methods. Hispanic women compared to
their non-Hispanic white counterparts, have a 48% lower risk of switching to condom relative to
least effective methods.
81
Based on the findings only one of four hypotheses is supported, albeit partially.
Hypothesis 17 posits that single women compared to married and cohabiting women are more
likely to switch to the pill and condom relative to least effective methods. Single women,
compared to married and cohabiting women, do have higher odds of switching to the pill relative
to least effective methods but for the condom relative to least effective methods.
Summary of Method Switching Patterns Based on Any Contraception Used at Start of
Observation
The multivariate models provide new evidence of contraceptive switching by focusing on
switching to specific methods. Prior studies have considered factors associated with switching to
specific methods (Grady et al. 2002; Manlove et al. 2013), but have not focused on union status.
Overall, among contraceptive switchers most (45%) switch to least effective methods (nonuse
and other methods), one-fifth switch to most effective methods (sterilization and hormonal
methods), 21% switch to condom and 13% switch to the pill. Further, the types of contraception
used following a switch varies according to union status. Married women compared their single
and cohabiting counterparts are more likely to switch to most effective methods. Cohabiting
women are more likely to switch to least effective methods compared to single and married
women. On the other hand, single women compared to married and cohabiting women, are more
likely to switch to the pill and condom.
Descriptive findings are similar to those at the zero order level. Compared to married
women, single women have lower risk of switching to most effective methods relative to least
effective methods. Compared to married women, single women have high risk of switching to
condom relative to least effective methods. Cohabiting women compared to married women
have a higher risk of switching to condom relative to most effective methods.
82
The results of multivariate analyses indicate that single women compared to their married
counterparts have high risk of switching to condom relative to most effective methods but this
association is marginally significant. In addition, compared to married women, single women
have a high risk of switching to condom rather than least effective methods.
Women’s sociodemographic, background and fertility characteristics are also related to
contraceptive methods used following a switching. Women with at least one child at time of
observation have higher risk of switching to most effective methods rather than least effective
methods. Parity is also associated with women’s risk of switching to the pill relative to condom.
Women with at least one child at observation have a high risk of switching to the pill relative to
condom.
Religious affiliation is associated with women’s risk of switching to the pill relative to
least effective methods and most effective methods. Women who report being raised as
Protestants, Catholics and other religious context (compared to those not raised in any religious
context) have higher risks of switching to the pill compared to least effective methods as well as
most effective methods. Women raised at Protestants compared to women not raised in any
religious context, have higher risk of switching to the pill compared to condom.
Women raised in a two biological/adoptive parent household have lower risk of switching
to the pill compared to condom. Compared to women with a high school diploma, women with
less than a high school diploma have a lower risk of switching to the pill relative to most
effective methods. Hispanic women compared their non-Hispanic white counterparts have a
lower risk of switching to condom relative to least effective methods.
83
Users of the Pill at the Start of the Observation Period Who Switch Contraception: Multivariate
Discrete-Time Event History Results
I present supplemental multivariate analyses of women who at the start of the observation
period used one of the three most common types of contraception (i.e., pill, condom and least
effective methods). Discrete-time multinomial logistic regression models are estimated to
examine women’s methods of contraception following a switch with the emphasis on union
status and contraceptive methods used (i.e. pill) at the start of the observation period. Among
initial pill users I compare the odds of switching to most effective methods (sterilization and
hormonal) versus least effective methods (‘other’ and no method) and switching to condom
relative to least effective methods. All multinomial logistic analyses include a time indicator. I
present the final model that shows the association of union status on contraceptive used
following a switch net of all characteristics. As shown above (Table 4.2), among women who
started using the pill three-fifths switched to the least effective methods, one-quarter switched to
condoms, and 15% switched to more effective methods.
In column 1 of Table 4.10 the results show that net of all characteristics, among initial
pill users, cohabiting women compared to married women have a 76% lower risk of switching to
most effective methods relative to least effective methods. This union status difference is also
significant at the zero order level (results not shown). In subsequent analysis results indicate
that among initial pill users cohabiting and single share similar risks of switching to most
effective methods relative to least effective methods (results not shown). The next of columns
show there are no statistically significant union status differentials in the odds of switching to
condoms rather than the least effective methods.
84
Background and fertility characteristics are related to the risk of switching from the pill to
most effective methods relative to least effective methods. Women raised as Protestants and
Catholics compared to those not raised in any religious context have a 73% and 54% lower risk
of switching to most effective methods compared to least effective methods. For every increase
in the age at first sex, women’s risk of switching to most effective methods relative to least
effective methods is reduced by 9%. With the inclusion of sociodemographic, background and
fertility indicators, union status is not associated with the risk of switching to condom relative to
least effective methods (Table 4.10, column 2).
Users of Condom at the Start of the Observation Period Who Switch Contraception: Multivariate
Discrete-Time Event History Results
Table 4.2 shows that among initial condom users about 70% switch to least effective,
approximately 20% switch to the most effective methods and 10% switch to the pill. In the full
model (Table 4.11, columns 1 and 2) the results are consistent with those at the zero order level.
Union status is not associated with the risk of switching to most effective methods relative to
least effective methods as well as switching to the pill relative to least effective methods. Thus,
the patterns of switching from condoms to other methods are similar for single, cohabiting, and
married women. Race/ethnicity and poverty status are also related to the risk of switching to the
pill relative to least effective methods. Compared to non-Hispanic white women, non-Hispanic
black and multiracial women have a 93% and 95% lower risk of switching to the pill compared
to least effective methods. Women at or above the poverty federal line have a 74% lower risk of
switching to the pill relative to least effective methods.
85
Users of Least Effective Methods at the Start of the Observation Period Who Switch
Contraception: Multivariate Discrete-Time Event History Results
Nearly equal proportions of women who started off using the least effective methods
switched to the pill, most effective methods and least effective methods (results not shown).
Table 4.12 (column 1) presents findings based on the full model. Single women are significantly
different from married women when predicting switching to most effective methods relative to
condom. This finding differs from the zero order results in which single and married women
share similar risks of switching to most effective methods relative to condom (results not
shown). When variables are entered separately in the model, age, education, mother’s education
and age at first sex are found to explain the suppression effect. In column 2 of Table 4.12 the
results indicate that with the inclusion of sociodemographic, background and fertility
characteristics the association between union status and the risk of switching to the pill relative
to condom is not significant. This result mirrors that at the zero order level. Further, additional
analysis shows that single and cohabiting women are not statistically different in the odds of
switching to the pill relative to condom (results not shown).
Two sociodemographic variables are also associated with women’s risk of switching to
most effective methods relative to condom. Compared to non-Hispanic white women, nonHispanic multiracial women have a 170% increased risk of switching to most effective methods
compared to condom. Women with less than a high school education compared to women with a
high school diploma have a 191% higher risk of switching to most effective methods relative to
condom.
86
Summary of Method Switching Patterns Based on Specific Contraception Used at Start of
Observation
Results indicate that union status differs in some extent depending on the originating
contraceptive method. Cohabiting and married women who start out as pill and least effective
method users appear to differ in their specific switching patterns. Among initial pill users at the
start of the observation period, cohabiting women compared to married women have lower odds
of switching to most effective methods relative to least effective methods. For least effective
methods users at start of the observation period cohabiting women compared to their married
counterparts have a lower risk of switching to most effective methods relative to condom. In
contrast, single and married women are comparable in their specific switching outcomes.
Overall these findings provide further support for research that distinguishes cohabiting and
single women as well as work that accounts for the contraceptive method at start of the period.
Prior work by Grady and colleagues (2002) examines the effectives of women’s characteristics
on the risk of switching from a specific origin method to another method of contraption.
However, consideration was not given to cohabitation and only marital status (married and
unmarried women). Manlove and colleagues’ (2013) estimates of contraceptive method
switching based on specific contraception used at start of observation period are different. The
reasons for differences are based partly on classification of contraception (for example, hormonal
methods include the pill and nonuse does not include ‘other’ methods) and they analyze multiple
switches within only a one year time frame and not first switch within a three-year period as is
the case for this research.
87
CHAPTER V: CONCLUSIONS
Trends over the last half century show worldwide changes in sexual and reproductive
health behaviors with increases in modern contraceptive use (United Nation 2011a). With
specific reference to the United States, it was not until 1960 that modern contraceptive methods
(example, pill and IUD) became available (Hatcher et al. 2004) and this was followed by the
availability of other effective methods during the course of the 20th century (CDC 1999). Today,
women have more options for and access to contraception (Institute of Medicine 2011) and with
recent advances in contraceptive technology; women have more independence regarding their
sexual and reproductive health (Lessard et al 2012). Notwithstanding these improvements, in the
United States almost half of all pregnancies are unintended (Trussell and Wynn 2008).
The sexual and reproductive health of all women in the United States continues to be a
serious public health concern. Government initiatives, such as Healthy People 2020, highlight
the importance of effective and consistent use of contraception to prevent unintended
pregnancies (U.S. Department of Health and Human Services 2012). While a lot of attention has
focused on young adults because of their high rates of unintended pregnancy (Finer and
Henshaw 2006), concerns about women 30 years and older are important as they contribute to
one-third of all unintended pregnancies (Finer and Henshaw 2006) and even older women (35
years or more) are at risk as they have the greatest proportion of pregnancies ending in abortion
(D’Angelo et al. 2004).
From the life course perspective (Elder 1994), understanding contraceptive and
reproductive behaviors within the context of relationships (or unions) are important. Family (and
union) formation has undergone a major transformation over the last decade with delays in
marriage and modest increases in the proportion of never married population given as a possible
88
explanation (Manning and Brown 2014). The increase in cohabitation is another explanation
posited by many family scholars. This union type is now normative as Manning (2013) finds
that two-thirds of women who recently married had lived in a cohabiting unions prior to
marriage. Single-hood among women is also important as the impermanency of both marriage
and cohabitation means that women are single at different stages in her lifetime (Lindberg and
Singh 2008). Prior studies find that reproductive and contraceptive behaviors among cohabiting
women are becoming more comparable to those of married women (Sweeney 2010) while the
reproductive behaviors of cohabiting women relative to single women are very distinct (Musick
2007; Manning et al. 2014a).
This dissertation focuses on a particular type of contraceptive behavior (contraceptive
method switching) that has been not recently investigated and provides an update to research by
considering changes in the behavior of married as well as cohabiting and single women.
Understanding contraceptive method switching is important because the type of switching has
negatively and positively reproductive implications for women throughout their life course.
Key Findings
Data from the National Survey of Family Growth (2006-10) is used to expand prior
research on contraceptive method switching by investigating variations in this behavior for
women across union status. The NSFG is applicable for this dissertation because it contains
detailed union (marital and cohabiting) and contraceptive histories for a nationally representative
sample of adult women between the ages of 21 and 44 (N= 2,986). There are some limitations to
the use of this data set.
In chapter II, I examine whether married women differ from cohabiting and single
women in terms of contraceptive method switching and stable use outcomes (i.e., stable nonusers
89
and stable users). I find that contraceptive method switching does differ by union status. Almost
two-fifths of women switch contraception. More than one-third of married women switch
contraception compared to more than half for all cohabiting and single women. Specifically,
single women are more likely than married to switch contraception relative to stable nonuse.
Compared to married women, single women are also more likely to remain as stable users
relative to stable nonusers. Grady et al. (2002) found that considerable amount of switching took
place among married (two-fifths) and unmarried women (three-fifths) within a two-year period.
The difference between results of this investigation with prior research may be due to datedness
of prior study, the operationalization and measurement of method switching and duration of
observation (see appendix for detailed discussion of Grady et al. 2002).
In chapter III, I focus on the contraceptive switching behavior of women parity zero and
investigate whether there are differences across union status. Almost two-fifth of women
without children switch contraception. An overall description of the sample indicates that
cohabiting women compared to married and single women are more likely to switch
contraception relative to not switching. At the bivariate level results indicate that single and
cohabiting women at parity zero differ from their married counterparts in terms of contraceptive
outcomes (switching, stable use and stable nonuse). However, multivariate analyses reveal that
women at parity zero regardless union status at observation share similar odds of contraceptive
method switching.
Chapter IV investigates types of contraception used among switchers across union status
regardless of method type used at the start of the observation. The majority of contraceptive
method switchers switch to least effective methods (none and other methods) regardless of initial
type of contraception used. Married women are more likely to switch to most effective methods
90
(sterilization and hormonal methods), cohabiting women tend to switch to least effective
methods and single women are more inclined to switch to the pill and condom. Specific findings
from this investigation corroborate those from prior research. For example, as expected
sterilization occurs mostly among married women, especially those who are nonusers at the start
of the observation period (see Grady et al. 2002; Jones et al. 2012). Multivariate analyses
indicate that single women more than married women have greater odds of switching to condom
relative to least effective methods of contraception. In addition, findings also reveal that union
status does differ to some extent based on the originating contraceptive method. Cohabiting and
married women who are initial pill users as well as least effective users have distinct specific
switching patterns.
Contributions
There are four main major contributions of the current study. First, prior research on
contraceptive method switching typically examines method switching using small, clinical and
disadvantaged samples to establish switching patterns. Generalization of results to the U.S.
population of reproductive age women is not applicable. Further, these studies tend to focus on
specific types of contraceptive method. Also union status is taken into account (at interview
date) after contraceptive method switching has occurred. Therefore, the assumption that
contraceptive method switching occurs within a particular union status is erroneous. The final
limitation to existing research is that uses nationally representative data, is that it is dated and
uses a 1995 cohort of reproductive age women (see Grady et al. 2002). This dissertation presents
current data with a recent cohort of reproductive age women to examine the prevalence of
contraceptive method switching.
91
Second, prior research on contraceptive method switching examines behavior among married
women as the context of childbearing was mostly highlighted through marital unions (Grady et
al. 2002). With the growth in unmarried couples living together and the increase in non-marital
childbearing especially among cohabiting women relative to single women (Brown 2005;
Manning et al. 2014a), understanding women’s contraception biography via contraceptive
switching becomes absolutely relevant. I investigate contraceptive method switching by
examining union status (married, cohabiting and single) at the start of the observation period
which allows accurate analyses of women who switch methods. In this study single women at
time of observation are measured and operationalized differently than in prior studies and
include women who may be dating, in a non-romantic relationship, former married or cohabitors.
Third, research on contraceptive behavior of women has consistently included parity as a
possible confounding characteristic influencing both contraceptive outcomes as well as
independent variables (Jacob and Stanfors 2013; Kavanaugh et al. 2011). Specific studies on
contraceptive method switching also have examined the possible mediating effects of parity
(Frost et al. 2007a, b; Grady et al. 2002). However, to my knowledge, no research has examined
contraceptive method switching behavior according to parity. This is of particular interest
because there has been a shift in toward delays in childbirth which has led to an extension in
women’s reproductive life course for being at risk to unintended pregnancy (Kavanaugh et al.
2011).
Fourth, measurement is a key issue in quantitative analysis of contraceptive switching. Prior
studies have conceptualized and operationalized contraceptive method switching in various
ways. Manlove and colleagues’ (2013) measure contraceptive method switching within a oneyear period by creating a cross-tabulation of prior month’s method with current method and
92
creating 12 matrices and then calculating the expected probability that a woman would be using
each method in each month in month 12, dependent on the method used in month 1. In another
study, contraceptive method switching is based on a series of questions about contraceptive
method used in the past year. The method type at the start of the year as well as the method type
used at the end of the year are collected (Frost et al. 2007 a). Grady and colleagues (2002) used
the contraceptive method history calendar to measure contraceptive switching over a two-year
period. Switching occurred if there was use of different methods in consecutive months, use of
two methods separated by a period of abstinence and use of two methods sequentially in the
same month- one method used in the prior month and the other in the subsequent months. For
this dissertation I use statistically rigorous techniques to evaluate the main research questions.
First, I construct the analytic sample based on the appending, reshaping, expanding and merging
of several files created from the original NSFG data 2006-10. For Chapters II, III and IV I use
discrete-time event history data to examine the risk of switching versus remaining as nonswitchers (i.e., stable nonusers and stable users) as well as the risk of switching to most effective,
the pill, condom and least effective methods.
Limitations and Future Research
This dissertation provides an exploratory and descriptive framework on contraceptive
method switching. The up-to-date analysis of women’s contraceptive method switching
behavior across union status provides a solid platform for future research. This dissertation has
some limitations. First, due to retrospective nature of the contraceptive method history calendar,
I cannot ascertain reasons for contraceptive method switching. Reasons for contraceptive
switching and method discontinuation can only be determined based on questions relating to
93
contraceptive use during the past 12 months asked at the time of interview. Studies that pursue
reasons for switching will provide important evidence about switching.
Second, this dissertation cannot examine the mediating effects of relationship
dynamics/quality on contraceptive method switching and union status. Recent studies, especially
those using young adult samples, have noted the importance of these relationship dynamics as
having both a direct and mediating effect on contraceptive use (Manning et al. 2009; Gibbs et al.
2014; Manlove et al. 2014). Perhaps diary data as collected by the Barber and colleagues would
be best to capture quality at the time of contraceptive switching.
Third, the analysis based on the contraceptive method calendar does not account for
experiences of women who have successfully used contraception for longer period (that is,
women whose first month of contraceptive use occurred before the start of the observation period
and who continued using contraception until the end of the observation period).
Fourth, while the NSFG 2006-10 provides marital and cohabitation dates for men, there
is no corresponding contraceptive method history calendar. Therefore, equivalent analyses for
men cannot be investigated. Future attention to men is warranted.
In this dissertation parity was included as a key control variable and separate analyses
were conducted for women without children. For this research exposure intervals were right
censored by end of observation and a change in union status. However, while women may
switch to nonuse, having censored cases due to pregnancy (measured as time-varying) may help
explain the reasons for contraceptive method switching. This approach could be considered for
future research.
Sixth, the data are cross-sectional and prevent time specific indicators of education,
religious affiliation, and health insurance. The use of education attainment serves as a crude
94
proximate measure in the analyses. The NSFG 2006-10 does not include a time-varying measure
of education and so I am unable to capture changes in education which may be consequential for
switching behavior. Religious affiliation is another crude proximate measure used a control
variable in the analyses. However, this measure is based on childhood experiences and though
majority of women were affiliated with religious groups, this by no means indicates that women
are currently religious or were religious during the time of observation. An indicator of religious
affiliation or religiosity three years prior to the interview was not available. Therefore, results
specifically related to religious affiliation must be cautiously interpreted.
In the United States, health insurance is intrinsically linked to access to health care
(Institute of Medicine, 2004). Research reveals that non insurance is a barrier to prescription
contraceptive methods (Culwell and Feinglass 2007; Frost and Darroch, 2008) which are an
effective and ideal method of birth control due to their greater protection against unintended
pregnancy and reversibility of fertility for future pregnancy (Nearns 2009). Uninsured women
are less likely compared to those with private insurance to use oral contraceptives (Shortridge
and Miller 2007). Further, Frost et al. (2007b) report that women without health insurance are
less likely to report switching contraceptives than those with private insurance. However, the
NSFG does not provide a time-varying variable on health insurance which could be useful in
predicting reasons switching. However, as a proxy for economic status, the federal poverty line
is used to capture women who may be disadvantaged and it is assumed that this may be linked to
access to health insurance.
A related issue is as with all retrospective data, there is the methodological issue of recall
of events. However, starting with the NSFG cycle 6 (2002) and now in subsequent surveys, the
female interview uses a life history calendar as a recall aid for the pregnancy and contraceptive
95
history portions of the interview (see Groves et al. 2005). Despite these methodological
strategies, I recognize that the recall of contraceptive methods may be problematic.
Research on race/ethnicity and contraceptive use are well established. Generally, blacks
are less likely than all other racial ethnic groups to practice contraception (Mosher and Jones
2010). Most studies examining racial and ethnic differences in contraceptive decisions have
placed attention on the individual’s use of any contraception and few studies have explored
contraceptive effectiveness (Shih et al. 2011; Frost et al. 2007 a, b; Culwell and Feinglass 2007).
While race/ethnicity is used only as a control covariate in this dissertation, its association with
contraceptive method switching and type of contraception used among initial condom and pill
users warrants further investigation. The dissertation provides evidence that non-Hispanic black
and multiracial women are less likely to switch contraception relative to stable use compared to
their non-Hispanic white counterparts. In addition, compared to non-Hispanic white women,
Hispanic women are less likely to switch to effective methods relative to least effective methods.
Therefore, investigation into the association between contraceptive method switching and
race/ethnicity including nativity may help to broaden our understanding of this particular
behavior and help inform policies to improve contraceptive and reproductive health among
particular racial/ethnic groups.
This dissertation also provides evidence that age mediates the relationship between
contraceptive method switching and union status. I find that switching behavior occurring
mostly among single and younger women. Research characterizes the life course stage of young
or emerging adults as demographically inconsistent (Arnett 2012) and one in which individuals
are sexually active but not in committed relationships; there are high rates of unintended and non
marital childbearing as well as inconsistent contraceptive use (Scott et al. 2011; Kavanaugh et al.
96
2013). Overall, there are increases in contraceptive use, however, many young adults either fail
to use effective methods or some do not use any form of contraception (Mosher and Jones 2010).
Therefore, the contraceptive switching behavior of young unmarried women during this stage of
the life course may be of interest to scholars and a subject for future study.
Overall, a large proportion of reproductive age women switch contraception at least once
during a three year period. Additionally, findings from the dissertation indicate that there is a lot
of movement in terms of switching patterns across union status. Prior studies have also
examined multiple switching behaviors among specialized samples of women. Barber and
colleagues (2011) have used weekly journal data collected from a U.S. sample to ascertain
information on contraceptive histories and contraceptive use patterns including multiple
switching. On the other hand, Manlove, Welti and Wildsmith (2013) have used nationally
representative data to examine multiple switching behaviors among women using hormonal and
LARC. This dissertation provides the framework for extending the discussion on contraceptive
method switching by using future research to explore multiple switching behaviors of women
across union status. Understanding the volatility of contraceptive use among reproductive age
women will better shape our understanding of this particular behavior.
The correct and consistent use of contraception is important in reducing unintended
pregnancies as well as STIs (Frost and Darroch 2008). The use of dual contraceptive methods is
viewed as a better approach to combat these challenges (Cates and Steiner 2002). However, scant
national data exist on factors associated with dual contraceptive use among adolescents and
young women (Tyler et al. 2014). Therefore, focusing on contraceptive method switching but
including dual methods as one possible type of contraception method among women of
reproductive age is an extension to the dissertation that can be considered for future research.
97
The findings reveal the proportion of women who switch contraception based on union
status at start of observation. This provides an important starting point but does not represent a
complete portrait of union status and switching behavior. In future research union status can be
measured as a time-varying characteristic in order to ascertain how changes in union predict
contraceptive method switching. More specifically, an examination of women who transition
from cohabitation to marriage and contraceptive method switching behavior is useful for further
expanding our knowledge of contraceptive and reproductive behavior of women in the United
States.
Finally, this dissertation provides an analysis of only one recent time period. In order to
assess method switching overtime and whether cohort differences predict this behavior,
upcoming research in the area of contraceptive method switching may want to focus on the use
1995, 2002 and 2006-10 NSFG data. This analysis would provide insights into contraceptive
switching over time.
Summary
Almost all sexually active women have used a form of contraception during their
lifetime. However, it is the consistency of use which creates the most challenge for women.
According to Sweeney (2010) future work should begin to investigate the association between
union status and a broader range of contraceptive practices. This dissertation serves to extend
this discussion by investigating contraceptive method switching, which despite its critical role in
reproductive regulation, has received scant attention from researchers. The results of this project
further supports the point that contraceptive behavior cannot be examined based on point
estimates, for example, whether contraception is used at ‘last month’ or ‘year’. It also extends the
98
discussion of measuring contraceptive behavior by providing possible explanations for
inconsistent contraceptive use.
I argue in this dissertation that contraceptive method switching among women of
reproductive age differs across union status. Results from chapters II indicates that contraceptive
switching behavior is predominantly driven by single women. In chapter III, women at parity
zero all share similar risk in terms of contraceptive switching behavior. Women’s contraceptive
use varies considerably across union status for women who switch contraception (chapter IV).
Regardless of contraceptive method use at start of the observation period, married women are
more likely to switch to most effective methods of contraception, cohabiting women switch to
least effective methods while single women mostly switch to the pill and condom.
Also noteworthy, are significant sociodemographic characteristics that are associated
with contraceptive switching behavior of women. Contraceptive method switching does not
seem to be a behavior that is practiced by older women. Further, having a child is related to an
increased likelihood of contraceptive switching. Non-Hispanic multiracial women are more
likely to remain as stable nonusers of conception than practice contraceptive method switching.
Having at least one child at time of observation increases the odds of switching contraception
relative to stable nonuse. Similar results also hold for women at parity zero.
Among contraceptive method switchers, parity increases the risk of switching to most
effective methods relative to least effective methods. Religious affiliation is associated with the
risk of switching to the pill relative to least effective methods such that women who were raised
in any religious context compared to women not raised in a religious context, have higher odds
of switching to the pill relative to least effective methods. Hispanic women compared to non-
99
Hispanic white women have lower odds of switching to condom relative to least effective
methods.
The results of this dissertation point to the growth in the similarity in reproductive and
contraceptive behaviors of married and cohabiting women. It also highlights the point that
attention must be given to single women and their reproductive and contraceptive wellbeing.
Given the instability of single women’s sexual relationships and the high probability of
switching among this group of women, programs and services aimed at improving their
reproductive and contraceptive health may be greater than those for married and cohabiting
women.
Overall, this dissertation indicates that union status has an important bearing on
contraceptive switching behavior among all women of reproductive age in the United States. The
fact that more than half of all cohabiting women and almost half of single and married women
who switch, switch to least effective methods (none and other methods) may further help to
explain the likelihood of unintended pregnancy especially among non-marital women in the
United States. On the other hand, contraceptive method switching behavior can also be used to
explain the intentionality of births especially among cohabitors. Although most are more likely
to switch to least effective methods, this may be due to their desire to have or ambivalence about
having a child. In addition, the results that single women are more likely to switch to condoms
relative to least effective methods may suggest that single women are more concerned about
STIs than unintended pregnancy.
In the final analysis, this dissertation illustrates that dynamic nature of contraceptive
behavior and the importance of distinguishing unmarried single and cohabiting women in the
analysis of contraceptive switching behavior.
100
REFERENCES
Abma, J., et al. (1997). Fertility, family planning, and women’s health: new data from the 1995
National Survey of Family Growth. Vital and Health Statistics, Series 23, No. 19.
Abma J., et al. (2004). Teenagers in the United States: sexual activity, contraceptive use, and
childbearing, 2002, Vital and Health Statistics, Vol. 23, No. 24.
Affandi, B., et al. (1987). Insertion and removal of NORPLANT contraceptive implants by
physicians and nonphysicians in an Indonesian clinic. Studies in Family Planning,
18(5):302-306.
Ali, M. M., & Cleland, J. (2010). Contraceptive switching after method-related
discontinuation: levels and differentials. Studies in Family Planning, 41(2):129-133.
American College of Obstetricians and Gynecologists. (2011). ACOG Practice Bulletin No. 121:
Long-acting reversible contraception: Implants and intrauterine devices. Obstetrics and
Gynecology, 118:184-96.
Arnett, J. J. (2004). Emerging adulthood: The winding road from the late teens through the
twenties. New York: Oxford University Press.
Arnett, J. J. (2012). New horizons in research on emerging and young adulthood. Early
Adulthood in a Family Context. National Symposium on Family Issues, 2, 231–244.
Bachrach, C. (1984). Contraceptive practice among American women, 1973-82. Family
Planning Perspectives, 16(6):253-259.
Bachrach, C. (1987). Cohabitation and reproductive behavior in the U.S. Demography,
24(4):623-637.
Bankole, A., & Westoff, C. F. (1998). The consistency and validity of reproductive attitudes:
Evidence from Morocco. Journal of Biosocial Sciences 30(4): 439–455.
101
Barber, J., Kusunoki, Y., & Gatny, H. (2011). Design and implementation of an online weekly
journal to study unintended pregnancies. Vienna Yearbook of Population Research,
(Vol.9), pp. 327-334.
Barden-O’Fallon, J., & Speizer, I. (2011). What differentiates method stoppers from switchers
among Honduran women. International Perspectives on Sexual and Reproductive Health,
37(1):16-23.
Basnayake, S., Thapa, S., & Balogh, S. (1988). Evaluation of safety, efficacy, and accessibility
of NORPLANT implants in Sri Lanka. Studies in Family Planning, 19(1):39-47.
Bengston, V. L., & Allen, K. R. (1993). The life course perspective applied to families over time.
In P. G. Boss, W. J. Doherty, R. LaRossa, W. R. Scham & S. K. Steinmetz (Eds.), Source
of families, theories and methods: A contextual approach (pp. 469-499). New York, NY:
Plenum.
Blackwell, D. L., & Lichter, D.T. (2000). Mate selection among married and cohabiting couples.
Journal of Family Issues, 21(3):275-302.
Bongaarts, J. (1978). A framework for analyzing the proximate determinants of fertility.
Population and Development Review, 4(1):105-132.
Billy, J. O. G., Brewster, K. L., & Grady, W. R. (1994). Contextual effects on the sexual
behavior of adolescent women. Journal of Marriage and Family, 56(2), 387–404.
Brown, S. L. (2004). Moving from cohabitation to marriage: effects on relationship
quality. Social Science Research, 33:1-19.
Brindis, C., Pagliaro, S., & and Davis. L. (2000). Protection as prevention: contraception for
sexually active teens. Washington, DC: National Campaign to Prevent Teen Pregnancy.
102
Brown, S. L. (2005). How cohabitation is reshaping American families. Contexts 4:33-37.
Brown, S. L. (2006). Family structure transitions and adolescent wellbeing. Demography,
43(3):447-461.
Brown, S. L., & Bulanda, J. R. (2006). Relationship violence in early adulthood: A comparison
of daters, cohabitors, and marrieds. Social Science Research, 37:73-87
Burgoyne, S. (2012). Relationship quality among married and cohabiting couples. (FP-12-12).
National Center for Family and Marriage Research.
http://ncfmr.bgsu.edu/pdf/family_profiles/file114312.pdf
Cates, W., Jr., & Steiner, M. (2002). Dual protection against unintended pregnancy and sexually
transmitted infections: What is the best contraceptive approach? Sexually Transmitted
Diseases, pp. 168-174.
Cavanagh, S. E., Crissey, S. R. & Raley, R. K. (2008). Family Structure History and Adolescent
Romance. Journal of Marriage and Family, 70: 698–714.
Center for Disease Control and Prevention (CDC). (2010). U.S. medical eligibility criteria for
contraceptive use. MMWR Recommendation Report, 59(RR-4):1-86.
Crossette, B. (2005). Reproductive health and the millennium development goals: The missing
link. Studies in Family Planning, 36(1):71-79.
Culwell, K. R., & Feinglass, J. (2007). The association between health insurance with use of
prescription contraceptives. Perspectives of Sexual and Reproductive Health, 39(4):226230.
Curtis, S., Evens, E., & Sambisa, W. (2011). Contraceptive discontinuation and unintended
pregnancy: an imperfect relationship. International Perspectives on Sexual and
Reproductive Health, 37(2):58-66.
103
D’Angelo, D., Gilbert, B., Rochat, R., Santelli, J., & Herold, J. (2004). Differences between
mistimed and unwanted pregnancies among women who have live births. Perspective on
Sexual and Reproductive Health, 36(5):192-7.
Daniels, K., Mosher, W. D., & Jones, J. (2013). Contraceptive methods women have ever used:
United States, 1982-2010. National Health Statistics Report (62), 1-16.
Darroch, J. E. (2013). Trends in contraceptive use. Contraception, 87(3):259-263.
Davidson, A. R., Kalmuss, D., Cushman, L. F., Romero, D., Heartwell, S., & Rulin, M. (1997).
Injectable contraceptive discontinuation and subsequent unintended pregnancy among
low-income women. American Journal of Public Health, 87(9):1532-1534.
DiClemente, R., Lodico, M., Grinstead, G., Harper, R., et al. (1996). African-American
adolescents residing in high-risk urban environments do use condoms: correlates and
predictors of condom use among adolescents in public housing developments. Pediatrics,
98(2):269-278.
Donnelly, D. A., Burgess, E., Anderson, S., Davis, R., & Dillard, J. (2001). Involuntary celibacy:
A life course analysis. The Journal of Sex Research, 38:159-169.
Eisenberg, D. L., Allsworth, J. E., Zhao, Q., & Peipert, J. F. (2012). Correlates of dual-method
contraceptive use: an analysis of the National Survey of Family Growth (2006-2008).
Infectious Diseases in Obstetrics and Gynecology, 1-6.
Elder, (Jr.), G. H. (1994). Time, human agency, and social change: perspective on the life course.
Social Psychology Quarterly 57(1): 4-15.
Elder, (Jr.), G. H. (1985). Perspective on the life course. In G. H. Elder (Jr.) (Ed.), Life Course
Dynamics: Trajectories and Transitions, 1968-1980. Ithaca: Cornell University Press, pp.
23-49.
104
Elder, (Jr.), G. H. (1998). The Life Course as Developmental Theory. Child Development, 69:1–
12.
Ersek, J. L. et al. (2011). Satisfaction and discontinuation of contraception by contraceptive
method among university women. Maternal and Child Health Journal, 15(4):497-506.
Finer, L., & Henshaw, S. (2006). Disparities in rates of unintended pregnancy in the United
States, 1994 and 2001. Perspectives on Sexual and Reproductive Health, 38(2):90-96.
Finer, L. B., Jerman, J., & Kavanaugh, M. (2012). Changes in use of long-acting contraceptive
methods in the United States, 2007-2009. Fertility and Sterility, 98(4):893-897.
Finer, L., & Philbin, J. (2014). Trends in ages at key reproductive transitions in the United
States, 1951-2010. Women’s Health Issues, 24(3):e271-e279.
Finer, L., & Zolna, M. (2011). Unintended pregnancy in the United States: incidence and
disparities, 2006. Contraception, 84(5):478-85.
Fomby, P., & Cherlin, A. J. (2007). Family instability and child well-being. American
Sociological Review, 72, 181–204.
Forrest, J. D. (1986). The end of IUD marketing in the United States: what does it mean for
American women? Family Planning Perspectives, 18(2):52-55 & 57.
Forrest, J. D., & Fordyce, R. (1988). U.S. Women’s contraceptive attitudes and practice: How
have they changed in the 1980s? Family Planning Perspective, 20(3):112-118.
Forste, R., & Tanfer, K. (1996). Sexual exclusivity among dating, cohabiting, and married
women. Journal of Marriage and Family, 58(1):33-47.
Freedman, L. P., & Isaacs, S. L. (1993). Human rights and reproductive choice. Studies in
Family Planning, 24(1):18-30.
105
Frost, J. J. (2011). The state of hormonal contraception today: Overview of unintended
pregnancy. American Journal of Obstetrics and Gynecology, 205(4):S1-S3.
Frost, J. J., & Darroch, J. E. (2008). Factors associated with contraceptive choice and
inconsistent method use, United States, 2004. Perspectives on Sexual and Reproductive
Health, 40(2):94-104.
Frost, J. J., Darroch, J. E., & Ramez, L. (2008). Improving contraceptive use in the United States.
Issues Brief (Alan Guttmacher Institute), 1:1-8.
Frost, J. J., Singh, S., & Finer, L. (2007a). U.S. women’s one-year contraceptive use patterns,
2004. Perspective on Sexual and Reproductive Health, 39(1):48-55.
Frost, J. J., Singh, S., & Finer, L. (2007b). Factors associated with contraceptive use and
nonuse, United States, 2004. Perspective on Sexual and Reproductive Health, 39(1):9099.
Furstenberg, (Jr.), F. F. (2013). Transitions into adulthood. The Annals of the American
Academy of Political and Social Sciences, 646:28-41.
Gibbs, L., Manning, W. D., Longmore, M., & Giordano, P. C. (2014). Qualities of Romantic
Relationships and Consistent Condom Use among Dating Young Adults. In
Contraceptives: Predictors of Use, Role of Cultural Attitudes & Practices and Levels of
Effectiveness, edited by Louis Bourgois and Samuel Cauchois. Hauppauge, New York:
Nova Science Publishers, Inc., pp, 157-182.
Giele, J. Z., & Elder, (Jr.), G. H. (Eds). (1998). Methods of Life Course Research: Qualitative
and Quantitative Approaches. Thousand Oaks: Sage Publications.
Glei, D. (1999). Measuring contraceptive use patterns among teenage and adult women. Family
Planning Perspectives, 31:73–80
106
Goodwin, P. Y., Mosher, W. D., & Chandra, A. (2010). Marriage and cohabitation in the United
States: a statistical portrait based on Cycle 6 (2002) of the National Survey of Family
Growth. National Center for Health Statistics. Vital and Health Statistics 23 (28).
Goldscheider, C., & Mosher, W. D. (1991). Patterns of contraceptive use in the United States:
the importance of religious factors. Studies in Family Planning, 22:102-115
Grady, W. R., Billy, J. O. G., & Klepinger, D. H. (2002). Contraceptive Method Switching in the
United States. Perspectives on Sexual and Reproductive Health, 34(3):135-145.
Grady, W. R., Hayward, M. D., & Florey, F. A. (1988). Contraceptive discontinuation among
married women in the United States. Studies in Family Planning, 19(4):227-235.
Grady, W. R., Hayward, M. D., Billy, J. O. G., & Florey, F. A. (1989). Contraceptive switching
among currently married women in the United States. Journal of Biosocial Science,
11(Supplement):117-132.
Grady, W. R., Hirsch, M. B., Keen, N., & Vaughan, B. (1983). Contraceptive failure and
continuation among married women in the United States, 1970-75. Studies in Family
Planning, 14(1):9-19.
Grady, W. R., Klepinger, D. H., & Billy, J. O. G. (1993). The influence of community
characteristics on the practice of effective contraception. Family Planning Perspectives,
25(1):4-11.
Grady, W. R., Klepinger, D. H., & Nelson-Wally, A. (1999). Contraceptive characteristics: the
perceptions and priorities of men and women, Family Planning Perspectives, 31(4):168175.
Groves, R. M., Benson, G., Mosher, W. D. et al. (2005). Plan and operation of cycle 6 of the
National Survey of Family Growth. Vital Health Statistics, August (42):1-86.
107
Guzzo, K. B. (2009). Marital intentions and the stability of first cohabitations. Journal of Family
Issues, 30:179-205.
Guzzo, K. B., & Hayford, S. R. (2014). Fertility and the stability of cohabiting unions: Variation
by intendedness. Journal of Family Issues, 35:547–576.
Hagestad, G. (1996). On-time, off-time, out of time? Reflections on continuity and discontinuity
from an illness process. In V. Bengtson (Ed.), Adulthood and aging: Research on
continuitiesand discontinuities (pp. 204–222). New York: Springer.
Hammerslough, C. R. (1984). Characteristics of women who stop using contraceptive. Family
Planning Perspectives, 16(1):14-18.
Hatcher, R, A. et al. (1994). Contraceptive technology. 16th edition. New York Irvington
Publishers.
Hatcher, R. A., Trussell, J., Nelson, A., Cates, W., Stewart, F., & Kowal, D. (2007).
Contraceptive technology. 19th revised edition. New York (NY), p.41.
Hill, M. S., Yeung, W. J., & Duncan, G. J. (2001). Childhood family structure and young adult
behaviors. Journal of Population Economics, 14, 271-299.
Hogan, D. P., Sun, R., & Cornwell, G. T. (2000). Sexual and fertility behaviors of American
females aged 15-19 years: 1985, 1990, and 1995. American Journal of Public Health, 90
(9):1421-1425.
Howard, D. and Wang, M. (2003). Risk profiles of adolescent girls who were victims of dating
violence. Adolescence 38(149):1-14.
Jacobs, J. & Stanfors, M. (2013). Racial and Ethnic Differences in U.S. Women's Choice Of
Reversible Contraceptives, 1995–2010. Perspectives on Sexual and Reproductive Health,
45:139–147.
108
Jaccard, J., & Levitz (2010). Counseling adolescents about contraception: towards the
development of an evidence-based protocol for contraceptive counselors. Journal of
Adolescent Health, 52(4): s6-s13.
Jones, R. K., Frohwirth, L., & Moore, A. (2013). More than poverty: disruptive events among
women having abortions in the USA. Journal of Family Planning and Reproductive
Health Care, 39:36-43
Jones, J., Mosher, W., & Daniels, K. (2012). Current contraceptive use in the United States,
2006-2010, and changes in patterns of use since 1995. National Health Statistics Reports,
No. 60, 1-25.
Joyner, K., Manning, W. D., & Bogle, R. H. (2013). The stability and qualities of same-sex and
different-sex couples in young adulthood. A Working Paper. Center for Family and
Demographic Research, Working Paper Series.
Kane, T., Gaminiratne, K., & Stephens, E. (1988). Contraceptive method switching in Sri Lanka:
Patterns and implications. International Family Planning Perspectives, 14(2):68-75.
Kavanaugh, M. L., Jerman, J., Hubacher, D., et al. (2011). Characteristics of women in the
United States who use long-acting reversible contraceptive methods. Obstetrics and
Gynecology, 117:1349-57.
Kavanaugh, M. L., Jerman, J., Eithier, K., & Moskosky, S. (2013). Meeting the contraceptive
needs of teens and young adults: youth-friendly and long-acting reversible contraceptive
services in the U.S. family planning facilities. Journal of Adolescent Health, 52(3):284292.
Kearney, M. S., & Levine, P. (2009). Subsidized Contraception, Fertility, and Sexual
Behavior. Review of Economics and Statistics, 91(1): 137–51.
109
Kennedy, S., & Bumpass, L. (2008). Cohabitation and children’s living arrangements: New
estimates from the United States. Demographic Research, 19, 1663-1692.
Kirby, D., Lepore, G., & Ryan, J. (2005). Sexual risk and protective factors: factors affecting
teen sexual behavior, pregnancy, childbearing, and sexually transmitted disease. The
National Campaign to Prevent Teen Pregnancy, Washington, DC.
Kirmeyer, S. E., & Hamilton, B. E. (2011). Childbearing differences among three generations of
U.S. women. National Center for Health Statistics Data Brief, No. 68, pp. 1-8.
Kost, K., Singh, S., Vaughan, B., Trussell, J., & Bankole, A. (2008). Estimates of contraceptive
failure from the 2002 National Survey of Family Growth. Contraception, 77:10-21.
Ku, L., Sonenstein, F. L., & Pleck, J .H. (1994). The dynamics of young men’s condom use
during and across relationships. Family Planning Perspectives, 26(6):246–251.
Kusunoki, Y., & Upchurch, D. M. (2011). Contraceptive method choice among youth in the
United States: The importance of relationship context. Demography, 48(4), 1451-1472.
Lammers, C., Ireland, M., Resnick, M., & Blum, R. (2000). Influences on adolescent’s decision
to postpone onset of sexual intercourse: A survival analysis of virginity among youths
aged 13 to 18 years. Journal of Adolescent Health, 26(1):42-48.
Lessard, L., Karasek, D., Ma, S., Darney, P., Deardorff, J., Lahiff, M., Grossman, D., & Greene
Foster, D. (2012). Contraceptive features preferred by women at high risk of unintended
pregnancy. Perspectives on Sexual and Reproductive Health, 44(3):194-200.
Litcher, D. T., Sassler, S., & Turner, R. (2014). Cohabitation, post-conception unions, and rise in
nonmarital fertility. Social Science Research, 47:134-147.
110
Lichter, D. T. (2012). Childbearing among cohabiting women: Race, pregnancy, and union
transition. In A. Booth et al. (eds.), Early adulthood in a family context, National
Symposium on Family Issues 2, 209-219.
Lindberg, L., & Singh, S. (2008). Sexual behavior of single adult American women. Perspectives
on Sexual and Reproductive Health, 40(1):27-33.
Liskin, L., & Blackburn, R. (1987). Hormonal contraception: new long-acting methods.
Population Reports, Series K, No.3.
Loomis, L. S., & Landale, N. S. (1994). Nonmarital cohabitation and childbearing among black
and white American women. Journal of Marriage and Family, 56, 949-962.
Manlove, J., Welti, K., Wildsmith, E. and Barry, M. (2014). Relationship Types and
Contraceptive Use Within Young Adult Dating Relationships. Perspectives on Sexual
and Reproductive Health, 46: 41–50.
Manlove, J., Ryan S., Wildsmith, E., & Franzetta, K. (2010). The relationship context of
nonmarital childbearing in the U.S. Demographic Research, 23:615-653.
Manlove, J., Welti, K., Barry, M., Peterson, K., Schelar, E. & Wildsmith, E. (2011). Relationship
Characteristics and Contraceptive Use Among Young Adults. Perspectives on Sexual and
Reproductive Health, 43: 119–128.
Manlove, J., Welti, K. & Wildsmith, E. (2013). Patterns of Method Switching and Continuation
among Women Using Hormonal and Long-Acting Methods. Child Trends.
http://www.caiglobal.org/caistage/images/pdfs/ntc/Grantee2013/breakout2-1jmanlove.pdf
Manning, W. D., & Brown, S. L. (2014). American Families: Demographic Trends and Social
Class. The Wiley-Blackwell Companion to the Sociology of Families, pp. 43-60.
111
Manning, W. D., Brown, S. L., & Payne, K. K. (2014a). Two decades of stability and change
in age at first union formation. Journal of Marriage and Family, 76: 247–260.
Manning, W. D., Brown, S. L., Lamidi, E., & Payne, K. K. (2014b). Trends in births to single
and cohabiting mothers, 1980-2009. (FP-14-05). National Center for Family & Marriage
Research. http://www.bgsu.edu/content/dam/BGSU/college-of-arts-andsciences/NCFMR/documents/FP/FP-14-05_TrendsInBirths.pdf
Manning, W. D. (2013). Trends in cohabitation: Over twenty years of change, 1987-2010.
(FP-13-12). National Center for Family & Marriage Research.
http://ncfmr.bgsu.edu/pdf/family_profiles/file130944.pdf
Manning, W. D. (2001). Childbearing in cohabiting unions: Racial and ethnic differences.
Family Planning Perspectives, 33(5), 217–223.
Manning, W. D., & Cohen, J. A. (2012). Premarital cohabitation and marital dissolution: An
examination of recent marriages. Journal of Marriage and Family, 74(2):377-387.
Manning, W. D., Flanigan, C. M., Giordano, P. C., & Longmore, M. A. (2009). Adolescent
dating relationships and consistency of condom use. Perspectives on Sexual and
Reproductive Health, 41(3), 181–190.
Manning, W. D., & Landale, N. (1996). Racial and Ethnic Differences in the Role of
Cohabitation in Premarital Childbearing. Journal of Marriage and the Family, 58(1):
63-77.
Manning, W. D., Longmore, M. A., & Giordano, P. C. (2000). The relationship context of
contraceptive use at first intercourse. Family Planning Perspectives, 32:166-75.
112
Manning, W. D., Longmore, M. A., & Giordano, P. C. (2007). The changing institution of
marriage: Adolescents’ expectation to cohabit and marry. Journal of Marriage and
Family, 69(3):559-575.
Martinez, G., Copen, C. E., & Abma, J. C. (2011). Teenagers in the United States: Sexual
activity, contraceptive use, and childbearing, 2006-2010 Nation Survey of Family
Growth. National Center for Health Statistics. Vital and Health Statistics, 23 (31), 1–35.
McQuillan, J., Greil, A., Shreffler, K., & Tichenor, V. (2008). The importance of motherhood
among women in the contemporary U.S. Gender & Society, 22, 477–496.
Meier, A. M. (2003). Adolescents’ transition to first intercourse: Religiosity, and attitudes about
sex. Social Forces, 81(3):1031-1052.
Monea, E., & Thomas, A. (2011). Unintended pregnancy and taxpayer spending. Perspectives on
Sexual and Reproductive Health, 43(2):88-93.
Moreau, C., Cleland, K., & Trussell, J. (2007). Contraceptive discontinuation attributed to
method dissatisfaction in the United States. Contraception, 76:267-272.
Mosher, W. D. (1990). Contraceptive practice in the United States, 1982-1988. Family Planning
Perspectives 22(5):198-205.
Mosher, W. D., & Jones, J. (2010). Use of contraception in the United States: 1982-2008. Vital
Health Statistics 23, 29:1-44.
Musick, K. (2002). Planned and unplanned childbearing among unmarried women. Journal of
Marriage and Family, 64(4):915-929.
Musick, K. (2007). Cohabition, nonmarital childbearing and the marriage process. Demographic
Research, 16(9):249-286.
National Center for Health Statistics. (2013). http://www.cdc.gov/nchs/
113
Nearns, J. (2009). Health insurance coverage and prescription contraceptive use among young
women at risk for unintended pregnancy. Contraception, 79(2):105-110.
Nock, S. L. (1995). A comparison of marriages and cohabiting relationships. Journal of Family
Issues, 16:53-76.
Osborne, C., Manning, W. D. and Smock, P. J. (2007). Married and cohabiting parents’
relationship stability: A focus on race and ethnicity. Journal of Marriage and Family,
69: 1345–1366.
Payne, K. K., Manning, W. D., & Brown, S. L. (2012). Unmarried births to cohabiting and single
mothers, 2005-2010. On the road to adulthood: Sequencing of family experiences. FP-1206, National Center for Family and Marriage Research.
http://ncfmr.bgsu.edu/pdf/family_profiles/file109171.pdf
Population Reference Bureau. (2012). World population data sheet.
http://www.prb.org/DataFinder/Geography/Data.aspx?loc=312
Ramstrom, K., Baron, A., Crane, L., & Shlay, J. (2002). Predictors of contraceptive
discontinuation in a sexually transmitted disease clinic population. Perspectives on
Sexual and Reproductive Health, 34(3):146-52.
Raley, R. K. (2001). Increasing fertility in cohabiting unions: Evidence for the second
demographic transition in the United States? Demography, 38, 59–66.
Rocca, C., Harper, C., & Raine-Bennett, T. (2013). Young women’s perception of the benefits of
childbearing: associations with contraceptive use and pregnancy. Perspectives on Sexual
and Reproductive Health, 45(1):23-32.
114
Sangi-Haghpeykar, H., Poindexter(3rd), A. N., Moseley, D. C., Bateman, L., & Reid, E. D.
(1995). Characteristics of injectable contraceptive users in a low-income population in
Texas. Family Planning Perspectives, 27(5):208-211 & 225.
Santelli, J. S., Davis, M., Celentano, D. D., Davis Crump, A., & Burwell, L. G. (1995).
Combined Use of Condoms with Other Contraceptive Methods Among Inner-City
Baltimore Women. Family Planning Perspectives, 27(2):74-78.
Sassler, S., Miller, A., & Favinger, S. F. (2009). Planned parenthood? Fertility intentions and
experiences among cohabiting couples. Journal Family Issues, 30, 206–232
Scott, M. E., Shelar, E., Manlove, J., & Cui, C. (2009). Young adult attitudes about relationships
and marriage: Times have changed but expectations remain high (No. 2009-30).
Washington, DC: Child Trends Research Brief.
Scott, M. E., Wildsmith, E., Welti, K., Ryan, S., Schelar, E. and Steward-Streng, N. R. (2011).
Risky Adolescent Sexual Behaviors and Reproductive Health in Young Adulthood.
Perspectives on Sexual and Reproductive Health, 43: 110–118.
Shafii, T., Stovel, K., Davis, R., & Holmes, K. (2004). Is condom use habit forming? Condom
use at sexual debut and subsequent condom use. Sexually Transmitted Diseases,
31(6):366-372.
Shih, G., Turok, D. K., & Parker, W. J. (2011). Vasectomy: the other (better) form of
sterilization. Contraception 83(4):310-315.
Smock, P. J. and Greenland, F. R. (2010). Diversity in Pathways to Parenthood: Patterns,
Implications, and Emerging Research Directions. Journal of Marriage and Family,
72: 576–593.
115
Sonfield, A., Hasstedt, K., & Gold, R.B. (2014). Moving forward, family planning in the era of
health reform. Guttmacher Institute, New York. .
South, S. J., & Crowder, K. D. (1999). Neighborhood effects on family formation: Concentrated
poverty and beyond. American Sociological Review, 64, 113-132.
Steele, F., & Curtis, S. (2003). Appropriate methods for analyzing the effect of method choice on
contraceptive discontinuation. Demography, 40(1):1-22.
Steele, F., & Diamond, I. (1999). Contraceptive Switching in Bangladesh. Studies in Family
Planning, 30(4):315-28.
Stuart, J. E. et al. (2013). Factors associated with 12-month discrimination among contraceptive
pill, patch, and ring users. Obstetrics and Gynecology, 121(2):330-336.
Sweeney, M. M. (2010). The reproductive context of cohabitation in the United States: recent
change and variation in contraceptive use. Journal of Marriage and Family, 72(5):11551170.
Tach, L. M. and Halpern-Meekin, S. (2012), Marital quality and divorce decisions: How do
premarital cohabitation and nonmarital Childbearing Matter? Family Relations, 61: 571–
585.
Tanfer, K., Wierzbicki, S., & Payn, B. (2000). Why are U.S. women not using long-acting
contraceptives? Family Planning Perspectives, 32(4):176-183 & 191.
Treas, J., & Giesen, D. (2000). Sexual infidelity among married and cohabiting Americans.
Journal of Marriage and Family, 62: 48–60.
Trussell, J. (2009). Understanding contraceptive failure. Contraception and Sexual Health,
23(2):199-209.
116
Trussell, J. (2010). Update on the cost-effectiveness of contraceptives in the United States.
Contraception, 82(4):391.
Trussell, J. (2011). Contraceptive failure in the United States. Contraception, 83(5):397-404.
Trussell, J., Henry, N., Hassan, F., Prezioso, A., Law, A., & Filonenko, A. (2013). Burden of
unintended pregnancy in the United States: potential savings with increased use of longacting reversible contraception. Contraception, 87(2):154-161.
Trussell, J., & Vaughan, B. (1999). Contraceptive failure, method-related discontinuation and
resumption of use: results from the 1995 National Survey of Family Growth. Family
Planning Perspectives, 31(2):64-72 & 93.
Trussell, J., & Wynn, L. L. (2008). Reducing unintended pregnancy in the United States.
Contraception, 77(1):1–5.
Tyler, C. P., et al. (2014). Dual use of condoms with other contraceptive methods among
adolescents and young women in the United States. Journal of Adolescent Health,
54(2):169-175.
United Nations Department of Economic and Social Affairs, Population Division. (2011a).
Total fertility by major area, region and country, 1950-2010. World Population Prospects,
CD-ROM Edition.
United Nations Department of Economic and Social Affairs, Population Division. (2011b).
World Contraceptive Use, 2011.
www.un.org/esa/population/.../contraceptive2011/wallchart_front.pdf
United States Census Bureau. (2011).
http://factfinder2.census.gov/faces/tableservices/jsf/pages/productview.xhtml?src=bkmk
117
United States Department of Health and Human Services, Healthy People 2020. (2012).
www.healthypeople.gov/2020
Vaughan, B., Trussell, J., Kos,t K., Singh, S., & Jones, R. (2008). Discontinuation and
resumption of contraceptive use: results from the 2002 National Survey of Family
Growth. Contraception, 78(4):271-283.
Vaughan, B., Trussell, J., Menken, J., Jones. E. F., & Grady, W. R. (1980). Contraceptive
efficacy among married women aged 15-44 years. Vital and Health Statistics, Series 23,
No. 5. Washington, DC: National Center for Health Statistics.
Waite, L. J., & Joyner, K. (2001). Emotional satisfaction and physical pleasure in sexual unions:
time horizon, sexual behavior, and sexual exclusivity. Journal of Marriage and Family,
63:247–264.
Weisman, C. S., Plichta, S., Nathanson, C. A., Ensminger, M., & Robinson, J. C. (1991).
Consistency of condom use for disease prevention among adolescent users of oral
contraceptives. Family Planning Perspectives, 23(2):71-74.
Westoff, C. F. (1972). The modernization of U.S. contraceptive practice. Family Planning
Perspectives, 4(3):9-12.
Wildsmith, E., Guzzo, K. B., & Hayford, S. R. (2010). Repeat Unintended, Unwanted and
Seriously Mistimed Childbearing in the United States. Perspectives on Sexual and
Reproductive Health, 42(1):14–22.
118
TABLES
Table 2.1 Means (and standard errors) and Percentages of Women by Union Status (N= 2,986)
All Women
Characteristics
Contraceptive outcomes a,b,c
Switchers
Non-Switchers
Stable nonusers
Stable users
Sociodemographic
Age
Race/Ethnicity
Mean/%
S.E
Married
Range
39.83
60.17
17.31
42.86
32.77
Mean/%
S.E
Cohabiting
Range
35.45
64.53
20.82
43.71
0.25
21-44
34.93
Mean/%
S.E
Single
Range
52.96
47.03
9.89
37.14
0.29
21-44
27.25
Mean/%
0.33
21-44
27.33
16.82
17.81
14.64
14.17
Non-Hispanic White
62.91
62.11
57.07
67.77
Non-Hispanic Black
9.26
7.81
21.02
9.96
Non-Hispanic Multiracial
11.00
12.24
7.26
8.08
9.32
9.34
16.21
6.71
High school/GED
19.79
19.63
34.23
15.09
Some college
27.02
24.92
34.38
31.60
College degree or higher
43.85
46.08
15.16
46.57
87.46
12.54
90.00
10.00
71.72
28.28
84.18
15.82
70.37
73.65
53.35
65.21
29.63
26.35
46.65
34.79
Federal Poverty Line (FPL)
At/Above FPL
Below FPL
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household
Religious Affiliation
Range
50.16
49.83
7.84
41.99
Hispanic
Education
Less than high school
S.E
0.21
21-44
119
All Women
Characteristics
Mean/%
S.E
Married
Range
Mean/%
S.E
Cohabiting
Range
Mean/%
S.E
Single
Range
Mean/%
None
8.23
7.09
16.11
9.32
Protestants
34.31
35.68
25.39
32.85
Catholics
45.31
44.60
53.39
44.84
12.13
12.62
5.09
12.98
Less than high school
22.45
25.56
21.48
12.06
High school/GED
33.51
34.64
37.06
28.32
Some college
21.77
19.23
19.25
31.47
College degree or higher
22.25
20.55
22.18
28.12
Other religious affiliation
Mother's Education
Fertility
Age at first sex
Parity - 1 or more children
N (unweighted)
18.46
0.16
67.33
10-40
18.81
0.22
79.43
2,986
10-40
16.52
0.54
56.62
1,927
10-40
17.98
S.E
Range
0.12
10-40
29.46
386
657
Note: All results are weighted. Ns are unweighted. Figures may not add to 100% due to rounding errors. Significant subgroup differences (p <0.05) are denoted by subscripts a,b,c.
Subscript a: differences between married and cohabiting women; subscript b: differences between cohabiting and single women; and subscript c: differences between married and
single women. Source: 2006-2010 National Survey of Family Growth (Female data file);period of observation is 3 years.
120
Table 2.2 Zero Order and Multivariate Discrete-Time Event History Models Predicting Contraceptive Method Switching (N= 2,986)
Odds
Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Zero-Order
S.E
Odds
Ratio
1.97
1.89
*
*
0.27
0.23
1.09
0.98
0.91
**
0.01
0.91
0.11
Model 1
*
S.E
Odds
Ratio
0.21
0.22
2.00
2.12
Model 2
†
†
S.E
Odds
Ratio
0.34
0.42
1.24
1.23
Model 3
S.E
0.30
0.34
0.01
0.90
*
0.01
1.02
0.26
0.98
0.05
0.88
0.08
0.81
†
0.05
0.09
0.47
†
0.08
Race/Ethnicity
Hispanic
1.02
0.24
Non-Hispanic White (ref.)
Non-Hispanic Black
0.82
†
Non-Hispanic Multiracial
0.50
0.18
0.50
0.81
0.24
0.73
0.22
0.71
0.19
Some college
1.04
0.22
1.13
0.27
1.17
0.27
College degree or higher
0.84
0.12
1.09
0.20
1.20
0.23
†
Education
Less than high school
High school/GED (ref.)
Federal Poverty Line (FPL)
At/Above FPL
0.69
*
0.04
0.78
0.12
0.81
0.14
0.86
*
0.02
0.98
0.04
0.98
0.05
0.28
0.95
0.21
0.96
0.18
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
1.04
121
Odds
Ratio
Catholics
1.09
Other religious affiliation
1.23
Zero-Order
†
Model 1
0.18
1.10
0.14
1.08
0.10
0.06
1.35
0.10
1.33
0.11
0.02
1.09
0.20
1.03
0.17
0.02
1.00
0.14
0.96
0.17
0.15
1.11
0.09
1.10
0.12
1.01
0.01
S.E
Odds
Ratio
Model 3
S.E
S.E
Odds
Ratio
Model 2
Odds
Ratio
S.E
Mother's Education
Less than high school
0.98
High school/GED (ref.)
Some college
1.18
College degree or higher
1.22
*
Fertility
Age at first sex
0.96
*
0.01
0.97
†
0.01
Parity - 1 or more children
0.96
0.04
1.30
0.17
1.81
†
0.27
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time is included and
continuous; 60, 403 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
122
Table 2.3 Zero Order Discrete-Time Event History Models Predicting Contraceptive Outcomes (N= 2,986)
Stable Users
(vs. Stable Nonusers)
Odds Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line (FPL)
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Switchers
(vs. Stable Nonusers)
S.E.
Odds Ratio
Switchers
(vs. Stable Users)
S.E.
Odds Ratio
S.E.
1.79
2.54
†
*
0.59
0.90
3.02
3.86
***
***
0.79
1.26
1.68
1.51
*
*
0.43
0.24
0.88
*
0.01
0.95
*
0.01
0.92
*
0.01
0.23
0.97
0.21
1.02
0.13
0.15
0.56
0.85
0.18
0.30
0.99
0.52
0.80
0.20
0.98
0.28
0.81
0.19
1.34
0.38
1.43
0.43
0.93
0.19
0.13
1.04
0.26
0.68
0.95
0.17
1.15
0.24
0.82
0.12
0.61
0.67
0.21
0.22
0.49
0.52
0.18
0.19
1.25
1.29
0.28
0.29
1.00
0.56
0.44
0.71
*
*
†
†
†
†
0.19
*
†
0.20
0.14
0.14
123
Stable Users
(vs. Stable Nonusers)
1.38
0.58
Switchers
(vs. Stable Nonusers)
1.14
0.62
Switchers
(vs. Stable Users)
1.21
0.35
Other religious affiliation
Mother's Education
Less than high school
1.14
0.29
1.25
0.32
0.91
0.17
High school/GED (ref.)
Some college
1.59
†
0.41
1.52
0.41
1.04
0.21
College degree or higher
1.84
*
0.57
1.78
*
0.50
1.03
0.19
Fertility
Age at first sex
0.94
*
0.02
0.97
0.02
0.97
†
0.01
Parity - 1 or more children
0.96
0.17
0.98
0.18
0.96
0.13
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 60,403 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
124
Table 2.4 Multivariate Discrete-Time Event History Using Union Status and Sociodemographic and Background Characteristics to Predict
Contraceptive Outcomes (N= 2,986)
Stable Users
(vs. Stable Nonusers)
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line (FPL)
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Switchers
(vs. Stable Nonusers)
Switchers
(vs. Stable Users)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
1.75
2.08
0.65
0.76
1.66
1.75
†
0.58
0.59
0.94
0.84
0.28
0.14
0.97
0.01
0.90
***
0.01
0.92
1.20
0.39
1.16
0.40
0.96
0.64
0.65
0.25
0.23
0.65
0.36
0.20
0.12
1.01
0.55
0.88
0.30
0.67
0.22
0.76
0.20
1.41
1.26
0.44
0.40
1.45
1.28
0.43
0.42
1.03
1.01
0.21
0.20
0.98
0.27
0.76
0.18
0.77
0.18
1.19
0.28
1.12
0.23
0.94
0.14
*
**
***
0.01
0.24
*
0.22
0.14
125
Protestants
0.48
*
0.17
0.54
†
0.18
1.12
0.27
Catholics
0.47
*
0.17
0.61
0.20
1.30
0.32
Other religious affiliation
1.18
0.66
1.56
0.77
1.31
0.37
Mother's Education
Less than high school
1.35
0.41
1.35
0.37
0.99
0.22
High school/GED (ref.)
Some college
1.35
0.37
1.25
0.32
0.92
0.18
College degree or higher
1.45
0.41
1.47
0.46
1.01
0.20
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 60,403 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
Table 2.5 Multivariate Discrete-Time Event History Using Union Status and Fertility Characteristics to Predict Contraceptive
Outcomes (N= 2,986)
Stable Users
(vs. Stable Nonusers)
Switchers
(vs. Stable Nonusers)
Switchers
(vs. Stable Users)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
Union Status at start of observation
Married (ref.)
Cohabiting
1.89
†
0.65
3.21
***
0.95
1.69
*
0.46
Single
3.01
**
1.18
4.91
***
1.91
1.63
**
0.29
Fertility
Age at first sex
0.98
0.02
0.97
0.02
0.98
0.01
Parity - 1 or more children
1.38
0.29
1.64
*
0.36
1.18
0.19
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 60,403 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file) ; period of observation is 3 years.
126
Table 2.6 Multivariate Discrete-Time Event History Using Union Status and All Characteristics to Predict Contraceptive
Outcomes (N= 2,986)
Stable Users
(vs. Stable Nonusers)
Odds Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line (FPL)
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Other religious affiliation
Switchers
(vs. Stable Nonusers)
S.E.
Odds Ratio
0.71
0.94
1.99
2.51
0.97
0.01
0.88
1.22
0.39
1.13
0.62
0.65
0.24
0.23
0.59
0.34
0.83
0.29
1.43
1.36
Switchers
(vs. Stable Users)
S.E.
Odds Ratio
S.E.
†
*
0.72
0.89
1.05
0.99
0.33
0.20
***
0.01
0.91
0.38
0.93
0.19
0.12
0.94
0.52
0.62
0.21
0.74
0.19
0.44
0.42
1.53
1.49
0.47
0.47
1.06
1.08
0.22
0.23
1.04
0.28
0.83
0.21
0.80
0.19
1.26
0.29
1.17
0.24
0.92
0.14
0.17
0.16
0.70
0.54
0.60
1.61
0.17
0.19
0.80
1.12
1.28
1.26
0.26
0.32
0.36
1.89
2.51
0.48
0.47
1.27
†
*
*
*
**
†
***
0.01
0.23
*
0.22
0.13
127
Mother's Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
1.32
0.41
1.25
0.35
0.95
0.21
1.29
1.47
0.35
0.42
1.16
1.48
0.29
0.48
0.89
1.00
0.17
0.20
Fertility
Age at first sex
0.96
0.02
0.99
0.02
1.02
0.02
Parity - 1 or more children
1.59
†
0.40
2.53
**
0.66
1.58
*
0.28
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 60,403 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
128
Table 3.1 Means (and standard errors) and Percentages of Women at Parity Zero by Union Status (N= 798)
All Women
Characteristics
Contraceptive outcomes a,b,c
Switchers
Non-Switchers
Stable nonusers
Stable users
Sociodemographic
Age
Mean/%
S.E
Married
Range
38.95
61.04
17.42
43.62
29.70
Mean/%
S.E
Cohabiting
Range
27.90
72.09
31.96
40.13
0.22
21-44
33.15
Mean/%
S.E
Single
Range
55.90
44.09
5.29
38.80
0.44
21-44
26.17
Mean/%
0.37
21-44
27.02
Hispanic
Non-Hispanic White
9.16
75.60
8.63
75.23
9.95
73.33
9.50
76.47
Non-Hispanic Black
6.34
6.44
10.11
5.40
Non-Hispanic Multiracial
8.89
9.68
6.60
8.61
Less than high school
3.06
3.05
9.53
1.62
High school/GED
Some college
14.32
27.41
18.25
22.83
27.01
35.75
7.54
30.14
College degree or higher
55.19
55.84
27.69
60.68
At/Above FPL
91.79
97.94
85.30
87.05
Below FPL
8.21
2.05
14.69
12.94
71.72
75.91
55.11
71.20
28.28
24.08
44.88
28.79
9.50
7.02
14.39
10.90
28.66
26.61
20.32
32.59
Education
Federal Poverty Line (FPL)
Religious Affiliation
None
Protestants
Range
46.29
53.70
5.50
48.20
Race/Ethnicity
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household
S.E
0.32
21-44
129
All Women
Characteristics
Mean/%
S.E
Married
Range
Mean/%
S.E
Cohabiting
Range
Mean/%
S.E
Single
Range
Mean/%
Catholics
48.31
53.98
59.98
40.01
Other religious affiliation
13.51
12.37
5.28
16.49
Less than high school
High school/GED
10.23
35.57
11.51
41.66
12.84
40.25
8.34
28.41
Some college
24.06
18.50
19.00
30.78
College degree or higher
30.12
28.30
27.90
32.45
S.E
Range
0.08
346
10-40
Mother's Education
Fertility
Age at first sex
N (unweighted)
18.27
0.16
798
10-40
19.63
0.12
330
10-40
16.69
0.37
122
10-40
18.48
Note: All results are weighted. Ns are unweighted. Figures may not add to 100% due to rounding errors. Significant subgroup differences (p <0.05) are denoted by subscripts a,b,c.
Subscript a: differences between married and cohabiting women; subscript b: differences between cohabiting and single women; and subscript c: differences between married and
single women. Source: 2006-2010 National Survey of Family Growth (Female data file) ; period of observation is 3 years.
130
Table 3.2 Zero Order and Multivariate Discrete-Time Event History Models Predicting Contraceptive Method Switching among Women at Parity
Zero (N= 798)
Odds
Ratio
Zero-Order
S.E
Odds
Ratio
1.53
0.87
2.12
1.42
0.01
0.92
0.72
0.10
Non-Hispanic Black
0.89
Non-Hispanic Multiracial
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
3.14
2.32
0.90
*
Model 1
*
S.E
Odds
Ratio
1.09
0.66
3.06
2.30
Model 2
S.E
Odds
Ratio
1.54
0.86
2.14
1.42
Model 3
S.E
1.11
0.66
0.01
0.91
*
0.01
0.60
0.24
0.60
0.24
0.30
1.35
0.62
1.38
0.64
0.73
0.64
0.64
0.33
0.63
0.32
0.56
0.35
0.57
0.26
0.56
0.26
1.49
1.12
0.58
0.59
1.41
1.18
0.59
0.57
1.41
1.16
0.59
0.56
0.92
0.45
1.43
0.56
1.43
0.58
0.80
0.29
0.90
0.39
0.89
0.38
1.05
1.29
0.91
0.80
0.97
1.47
0.58
0.56
0.95
1.45
0.57
0.54
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
131
Odds
Ratio
Other religious affiliation
Mother's Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Zero-Order
Model 1
1.92
1.22
1.97
0.94
1.88
0.89
1.08
0.53
1.15
0.58
1.13
0.56
0.23
0.11
0.93
1.07
0.22
0.10
0.94
1.05
0.22
0.09
1.27
1.31
†
S.E
Odds
Ratio
Model 3
S.E
S.E
Odds
Ratio
Model 2
Odds
Ratio
S.E
Fertility
Age at first sex
0.97
0.01
0.99
0.02
1.01
*
0.01
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time is included and
continuous; 18,809 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
132
Table 3.3 Zero Order Discrete-Time Event History Models Predicting Contraceptive Outcomes among Women at Parity Zero (N= 798)
Stable Users
(vs. Stable Nonusers)
Odds Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Switchers
(vs. Stable Nonusers)
S.E.
Odds Ratio
Switchers
(vs. Stable Users)
S.E.
Odds Ratio
S.E.
5.88
6.96
**
*
0.75
3.05
11.58
9.98
*
*
4.84
4.48
1.96
1.43
1.06
0.62
0.80
*
0.02
0.77
*
0.02
0.96
0.02
0.34
0.19
0.36
*
0.89
1.04
0.38
0.44
0.32
0.20
0.20
0.51
0.35
0.30
0.39
1.15
1.09
0.17
0.79
2.86
**
0.16
1.12
0.64
0.39
0.24
2.27
2.51
†
0.45
0.87
2.51
2.04
0.92
1.30
1.10
0.81
0.47
0.36
0.50
0.41
0.54
0.60
1.08
0.43
0.83
0.36
0.70
0.18
0.84
0.39
0.44
0.07
1.11
0.86
1.01
0.49
1.03
1.59
0.89
1.08
1.07
0.54
*
133
Stable Users
(vs. Stable Nonusers)
1.84
1.07
Switchers
(vs. Stable Nonusers)
3.10
1.53
Switchers
(vs. Stable Users)
1.68
1.17
Other religious affiliation
Mother's Education
Less than high school
0.65
0.16
0.86
0.30
1.32
0.77
High school/GED (ref.)
Some college
3.32
2.14
3.02
1.37
0.91
0.23
College degree or higher
4.03
*
0.79
3.66
*
0.56
0.90
0.08
Fertility
Age at first sex
0.93
*
0.01
0.92
†
0.02
0.99
0.02
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 18,809 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
134
Table 3.4 Multivariate Discrete-Time Event History Using Union Status and Sociodemographic and Background Characteristics to Predict
Contraceptive Outcomes among Women at Parity Zero (N= 798)
Stable Users
(vs. Stable Nonusers)
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Switchers
(vs. Stable Nonusers)
Switchers
(vs. Stable Users)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
2.02
3.31
0.61
1.68
3.84
4.11
†
3.14
2.09
1.89
1.24
0.97
0.58
0.82
0.02
0.79
*
0.02
0.97
0.02
0.25
0.17
0.21
†
0.10
0.83
0.47
0.68
0.30
0.17
0.20
0.96
0.26
*
0.54
0.07
1.41
0.85
0.44
0.48
†
0.62
0.40
0.23
5.80
†
3.02
2.35
2.08
1.50
*
0.36
0.89
2.60
1.53
0.93
1.05
1.25
1.02
0.54
0.43
1.36
1.31
1.82
2.13
1.34
0.28
1.55
0.92
1.32
0.79
0.85
0.40
135
Protestants
0.70
0.32
0.76
0.48
1.07
0.71
Catholics
0.48
†
0.12
0.84
0.16
1.75
0.77
Other religious affiliation
1.59
1.54
3.06
†
1.03
1.92
1.26
Mother's Education
Less than high school
0.54
0.18
0.73
0.30
1.35
0.76
High school/GED (ref.)
Some college
1.58
0.48
1.35
*
0.09
0.85
0.23
College degree or higher
2.56
0.58
2.36
†
0.71
0.92
0.06
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 18,809 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
Table 3.5 Multivariate Discrete-Time Event History Using Union Status and Fertility Characteristics to Predict Contraceptive
Outcomes among Women at Parity Zero (N= 798)
Stable Users
(vs. Stable Nonusers)
Switchers
(vs. Stable Nonusers)
Switchers
(vs. Stable Users)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
Union Status at start of observation
Married (ref.)
Cohabiting
5.11
*
1.06
10.21
*
3.92
1.99
1.12
Single
6.68
*
3.00
9.62
*
4.15
1.43
0.63
Fertility
Age at first sex
0.95
0.02
0.95
0.03
1.01
0.02
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 18,809 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
136
Table 3.6 Multivariate Discrete-Time Event History Using Union Status and All Characteristics to Predict Contraceptive
Outcomes among Women at Parity Zero (N= 798)
Stable Users
(vs. Stable Nonusers)
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Other religious affiliation
Switchers
(vs. Stable Nonusers)
Switchers
(vs. Stable Users)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
2.02
3.30
0.60
1.72
3.87
4.10
3.17
2.19
1.91
1.24
0.99
0.59
0.02
0.76
*
0.02
0.96
0.02
0.25
0.17
0.21
†
0.10
0.82
0.48
0.67
0.30
0.14
0.19
0.98
0.26
*
0.50
0.07
1.45
0.84
0.48
0.47
0.82
*
5.79
†
3.06
2.34
†
0.63
0.40
0.22
2.07
1.51
†
0.39
0.90
2.60
1.51
†
0.85
1.00
1.25
1.00
0.54
0.43
1.36
1.31
1.82
2.15
1.33
0.29
1.56
0.92
1.31
0.79
0.84
0.39
0.71
0.48
1.60
0.35
0.12
1.67
0.75
0.83
2.93
0.47
0.14
1.15
1.05
1.72
1.82
0.71
0.75
1.20
137
Mother's Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
0.54
1.58
2.57
†
0.19
0.72
0.50
0.64
1.35
2.34
†
†
0.31
1.33
0.74
0.11
0.71
0.85
0.90
0.23
0.05
Fertility
Age at first sex
0.99
0.03
1.01
0.02
1.01
0.01
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 18,809 person-months; Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
138
Table 4.1 Means (and standard errors) and Percentages of Women who Switch Contraception by Union Status (N= 1,899)
All Women
Characteristics
Contraceptive Methods Used After
Switch a,b,c
Sterilization1
Mean/%
S.E
Married
Range
Mean/%
S.E
Cohabiting
Range
Mean/%
S.E
Single
Range
Mean/%
9.95
13.15
4.66
3.26
Hormonal
10.26
10.63
9.50
9.52
Pill
Condom
Other3
None
Sociodemographic
Age
Race/Ethnicity
14.36
21.32
7.86
36.27
13.04
19.45
8.23
35.50
12.60
20.23
5.25
47.76
18.68
26.89
7.88
33.78
2
30.74
0.21
21-44
33.02
0.34
21-44
26.78
0.48
21-44
26.69
Hispanic
17.69
18.79
16.17
15.58
Non-Hispanic White
66.20
67.13
56.52
67.65
Non-Hispanic Black
8.78
6.19
19.11
11.15
Non-Hispanic Multiracial
7.31
7.88
8.19
5.59
8.91
8.26
16.76
7.49
High school/GED
21.01
20.02
40.25
16.05
Some college
29.02
26.83
33.07
32.79
College degree or higher
41.04
44.86
9.90
43.65
84.47
15.21
87.67
12.32
71.37
28.62
83.24
16.75
68.08
72.21
52.10
63.86
31.98
27.78
47.89
36.13
Education
Less than high school
Federal Poverty Line (FPL)
At/Above FPL
Below FPL
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household
S.E
0.39
Range
21-44
139
All Women
Characteristics
Mean/%
S.E
Married
Range
Mean/%
S.E
Cohabiting
Range
Mean/%
S.E
Single
Range
Mean/%
S.E
Range
0.43
10-40
Religious Affiliation
None
7.94
7.01
13.00
8.29
Protestants
35.54
34.97
32.31
34.34
Catholics
Other religious affiliation
44.71
12.79
44.10
13.90
51.37
3.29
43.65
13.71
Less than high school
21.37
22.74
28.36
15.36
High school/GED
32.00
33.20
41.15
25.56
Some college
23.06
21.37
13.84
30.75
College degree or higher
23.54
22.68
16.63
28.31
Mother's Education
Fertility
Age at first sex
18.10
Parity - 1 or more children
N (unweighted)
68.04
0.10
10-40
18.55
0.24
83.81
1,899
10-40
16.41
0.65
54.21
1,150
10-40
17.66
34.90
262
487
Note: All results are weighted. Ns are unweighted. Figures may not add to 100% due to rounding errors. Significant subgroup differences (p <0.05) are denoted by subscripts a,b,c.
Subscript a: differences between married and cohabiting women; subscript b: differences between cohabiting and single women; and subscript c: differences between married and
single women. Subscript 1 and 2 denotes most effective methods. Hormonal methods include: Depo-Provera, Hormonal implant, IUD, Lunelle injectable, contraceptive patch and
vaginal contraceptive ring. Subscript 3 and none category denotes least effective method. ‘Other’ methods include withdrawal, rhythm, safe period, female condom, diaphragm,
foam, and cream. Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation is 3 years.
140
Table 4.2 Percentage of Women Using Selected Contraceptive Methods Who Switch Contraception
(N= 1,899)
Methods of Contraception following Switch
Methods of Contraception at
start of observation
Pill
Condom
Hormonal
Other
None
Total
Pill
Condom
26.46
Hormonal
8.65
9.10
Other
7.53
5.54
6.40
10.26
18.19
6.84
28.50
26.71
10.53
29.83
7.38
15.56
11.71
Sterilization
5.85
10.62
5.24
8.41
14.40
14.36
21.32
10.26
7.86
9.95
None
51.51
64.49
43.45
66.84
Total
100.00
100.00
100.00
100.00
100.00
36.27
100.00
Note: All percentages are weighted. N is unweighted. Figures may not add to 100% due to rounding errors. Source: 2006-2010 National Survey
of Family Growth (Female data file); period of observation is 3 years.
Table 4.3 Percentage of Married Women Using Selected Contraceptive Methods Who Switch Contraception
(N= 1,150)
Methods of Contraception following Switch
Methods of Contraception at
start of observation
Pill
Condom
Hormonal
Other
None
Total
Pill
Condom
22.30
Hormonal
8.99
7.28
Other
6.85
3.81
8.78
7.21
13.97
7.77
25.26
24.31
8.4
28.41
9.32
16.23
13.18
Sterilization
9.23
15.84
6.70
9.55
16.93
13.04
19.45
10.63
8.23
13.15
None
52.63
65.86
46.24
64.96
Total
100.00
100.00
100.00
100.00
100.00
35.50
100.00
Note: All percentages are weighted. N is unweighted. Figures may not add to 100% due to rounding errors. Source: 2006-2010 National Survey
of Family Growth (Female data file); period of observation is 3 years.
141
Table 4.4 Percentage of Cohabiting Women Using Selected Contraceptive Methods Who Switch Contraception
(N= 262)
Methods of Contraception at
start of observation
Pill
Condom
Hormonal
Other
None
Total
Pill
Condom
11.68
Hormonal
5.35
14.82
Other
10.36
1.08
0.00
11.47
40.64
17.80
18.09
6.09
7.58
50.97
11.70
15.50
7.05
Sterilization
0.00
3.34
7.05
6.31
8.39
12.6
20.23
9.50
5.25
4.66
None
72.60
69.29
46.13
83.15
Total
100.00
100.00
100.00
100.00
100.00
47.76
100.00
Note: All percentages are weighted. N is unweighted. Figures may not add to 100% due to rounding errors. Source: 2006-2010 National Survey
of Family Growth (Female data file); period of observation is 3 years.
Table 4.5 Percentage of Single Women Using Selected Contraceptive Methods Who Switch Contraception (N= 487)
Methods of Contraception following Switch
Methods of Contraception at
start of observation
Pill
Condom
Hormonal
Other
None
Total
Pill
Condom
38.94
Hormonal
9.03
11.18
Other
7.95
11.90
2.48
17.45
19.49
5.89
44.18
44.25
21.31
26.12
2.90
13.24
8.48
Sterilization
1.24
0.60
0.00
4.84
7.98
18.68
26.89
9.52
7.88
3.26
None
42.85
58.88
33.78
65.07
Total
100.00
100.00
100.00
100.00
100.00
33.78
100.00
Note: All percentages are weighted. N is unweighted. Figures may not add to 100% due to rounding errors. Source: 2006-2010 National Survey
of Family Growth (Female data file);period of observation is 3 years.
142
Table 4.6 Zero Order Discrete-Time Event History Models Predicting Contraceptive Method Use among Women Who Switch Contraception (N=
1,899)
Most Effective1
(vs. Least Effective2)
Odds Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Pill
(vs. Least Effective)
Condom
(vs. Least Effective)
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
0.19
0.20
0.51
0.47
†
*
0.10
0.11
0.82
1.33
0.48
0.47
0.86
1.83
1.04
**
0.02
0.98
0.03
0.98
0.82
0.09
0.68
0.13
0.38
0.84
0.95
0.14
0.14
0.85
0.41
0.37
0.16
0.59
0.62
1.44
0.68
0.68
0.27
0.55
*
0.03
1.29
0.80
0.40
0.29
0.90
1.23
0.28
0.58
1.63
1.60
*
0.11
0.29
0.98
0.23
1.15
0.07
1.38
0.18
1.04
0.08
0.85
0.23
1.15
0.27
0.63
0.19
2.37
0.56
0.71
0.23
†
*
0.02
*
0.07
0.11
0.25
143
Most Effective1
(vs. Least Effective2)
0.74
0.15
0.71
*
0.02
Pill
(vs. Least Effective)
2.77
†
0.78
5.82
*
2.42
Condom
(vs. Least Effective)
0.83
0.38
0.62
0.53
Catholics
Other religious affiliation
Mother's Education
Less than high school
0.88
0.25
0.71
0.43
0.34
†
0.10
High school/GED (ref.)
Some college
0.67
0.23
0.86
0.31
1.00
0.43
College degree or higher
0.74
0.18
1.26
0.79
1.14
0.26
Fertility
Age at first sex
0.93
0.03
0.97
0.07
0.98
0.02
Parity - 1 or more children
2.61
†
0.64
0.90
0.07
0.59
0.11
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 25,666 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period of observation
is 3 years.
144
Table 4.7 Multivariate Discrete-Time Event History Using Union Status and Sociodemographic and Background Characteristics to Predict
Contraceptive Method Use among Women Who Switch Contraception (N= 1,899)
Most Effective
(vs. Least Effective)
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Pill
(vs. Least Effective)
Condom
(vs. Least Effective)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
0.48
0.57
0.18
0.15
0.89
1.16
0.34
0.71
0.93
1.72
0.04
0.14
0.01
0.98
0.04
0.99
0.56
0.19
0.73
0.23
0.50
0.92
0.84
0.26
0.17
0.93
0.31
0.46
0.12
0.74
0.74
0.22
0.32
1.77
0.60
0.81
0.29
0.76
0.13
1.21
0.58
0.40
0.21
0.87
1.16
0.15
0.41
1.37
1.10
1.03
0.21
1.00
0.14
0.95
0.26
1.06
0.14
0.70
0.16
1.09
0.15
1.03
*
†
*
0.02
**
†
0.02
0.12
0.17
145
Protestants
0.66
0.18
2.45
*
0.51
0.75
0.17
Catholics
0.84
0.14
3.55
*
0.70
1.18
0.57
Other religious affiliation
0.71
0.12
7.15
*
2.32
0.63
0.48
Mother's Education
Less than high school
0.87
0.24
0.82
0.38
0.44
0.16
High school/GED (ref.)
Some college
0.79
0.29
0.78
0.22
0.90
0.35
College degree or higher
0.94
0.26
1.11
0.49
1.11
0.13
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 25,666 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period
of observation is 3 years.
Table 4.8 Multivariate Discrete-Time Event History Using Union Status and Fertility Characteristics to Predict Contraceptive
Method Use among Women Who Switch Contraception (N= 1,899)
Most Effective
(vs. Least Effective)
Pill
(vs. Least Effective)
Condom
(vs. Least Effective)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
S.E.
Union Status at start of observation
Married (ref.)
Cohabiting
0.52
0.13
0.76
0.56
0.73
0.13
Single
0.59
0.13
1.27
0.69
1.49
†
0.17
Fertility
Age at first sex
0.92
0.04
0.98
0.08
0.98
0.02
Parity – 1 or more children
2.01
*
0.29
0.96
0.32
0.67
0.17
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 25,666 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period
of observation is 3 years.
146
Table 4.9 Multivariate Discrete-Time Event History Using Union Status and All Characteristics to Predict Contraceptive
Method Use among Women Who Switch Contraception (N= 1,899)
Most Effective
(vs. Least Effective)
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Other religious affiliation
Mother's Education
Pill
(vs. Least Effective)
Condom
(vs. Least Effective)
Odds Ratio
S.E.
Odds Ratio
S.E.
Odds Ratio
0.56
0.73
0.20
0.15
0.96
1.28
0.44
0.93
0.88
1.63
0.01
0.98
0.05
1.00
0.59
0.18
0.76
0.28
0.52
0.83
0.89
0.27
0.13
0.87
0.32
0.40
0.11
0.75
0.73
0.20
0.31
1.68
0.50
0.79
0.26
0.77
0.13
1.26
0.71
0.38
0.21
0.89
1.33
0.19
0.70
1.35
1.07
1.13
0.31
1.03
0.14
0.94
0.25
1.15
0.21
0.75
0.12
1.10
0.12
0.68
0.85
0.91
0.19
0.16
0.13
2.44
3.44
8.14
0.51
0.91
2.04
0.75
1.18
0.63
0.18
0.59
0.48
1.03
†
†
*
*
*
S.E.
*
0.04
0.16
0.02
*
†
0.03
0.13
0.15
147
Less than high school
0.90
0.24
0.86
0.47
0.44
0.16
High school/GED (ref.)
Some college
0.78
0.31
0.77
0.20
0.91
0.37
College degree or higher
0.97
0.28
1.09
0.42
1.10
0.13
Fertility
Age at first sex
0.92
0.04
0.95
0.10
0.99
0.01
Parity – 1 or more children
1.88
*
0.25
1.24
0.26
0.87
0.20
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 25,666 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period
of observation is 3 years.
148
Table 4.10 Multivariate Discrete-Time Event History Models Predicting Contraceptive Method Use among Initial Pill Users Who Switch
Contraception (N= 479)
Most Effective
(vs. Least Effective)
Odds Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Other religious affiliation
Condom
(vs. Least Effective)
S.E.
Odds Ratio
S.E.
0.07
0.45
0.27
2.05
0.19
0.87
1.03
0.03
1.05
0.02
1.10
1.29
0.70
0.33
1.01
1.43
0.37
1.27
0.19
2.62
1.51
1.44
0.58
0.48
1.77
1.18
0.85
0.52
1.23
0.44
0.35
0.16
0.74
0.75
2.65
1.30
0.44
0.99
0.47
0.02
0.07
0.93
0.63
2.56
2.02
0.72
3.03
1.73
0.24
0.52
0.27
0.45
0.53
*
**
*
†
†
0.26
0.74
0.86
149
Most Effective
(vs. Least Effective)
Condom
(vs. Least Effective)
Mother's Education
Less than high school
1.11
0.71
0.14
**
0.02
High school/GED (ref.)
Some college
0.82
0.37
0.63
†
0.09
College degree or higher
1.62
0.97
0.88
0.18
Fertility
Age at first sex
0.91
*
0.01
0.92
0.06
Parity – 1 or more children
1.01
0.43
0.51
0.15
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 7,784 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period
of observation is 3 years.
150
Table 4.11 Multivariate Discrete-Time Event History Models Predicting Contraceptive Method Use among Initial Condom Users Who Switch
Contraception (N=395)
Most Effective
(vs. Least Effective)
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Other religious affiliation
Mother's Education
Pill
(vs. Least Effective)
Odds Ratio
S.E.
Odds Ratio
S.E.
1.32
1.09
1.14
0.22
0.46
0.69
0.14
0.37
1.06
0.07
0.93
0.05
0.51
0.44
0.05
0.06
0.14
1.36
0.17
0.27
0.07
0.05
0.08
41.86
138.82
0.87
0.44
0.59
0.43
13.81
31.17
22.00
51.92
0.73
0.48
0.26
0.35
0.17
0.38
0.23
2.48
1.00
0.36
1.95
1.03
0.22
6.15
12.40
29.73
8.83
33.18
48.09
0.17
†
*
**
*
0.04
0.01
0.07
151
Most Effective
(vs. Least Effective)
0.16
*
0.04
Pill
(vs. Least Effective)
1.77
0.70
Less than high school
High school/GED (ref.)
Some college
0.98
0.64
0.66
0.37
College degree or higher
1.25
0.76
0.30
†
0.11
Fertility
Age at first sex
1.04
0.01
0.91
0.14
Parity – 1 or more children
2.16
0.83
0.07
0.13
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 5,170 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period
of observation is 3 years.
152
Table 4.12 Multivariate Discrete-Time Event History Models Predicting Contraceptive Method Use among Initial Least Effective Users Who
Switch Contraception (N= 827)
Most Effective
(vs. Condom)
Odds Ratio
Union Status at start of observation
Married (ref.)
Cohabiting
Single
Sociodemographic
Age
Race/Ethnicity
Hispanic
Non-Hispanic White (ref.)
Non-Hispanic Black
Non-Hispanic Multiracial
Education
Less than high school
High school/GED (ref.)
Some college
College degree or higher
Federal Poverty Line
At/Above FPL
Below FPL (ref.)
Background
Family Type during Childhood
Two Bio/Adoptive Parent
Household
Non Two Bio/Adoptive Parent
Household (ref.)
Religious Affiliation
None (ref.)
Protestants
Catholics
Other religious affiliation
Mother's Education
Pill
(vs. Condom)
S.E.
Odds Ratio
S.E.
0.09
0.20
0.56
2.45
0.13
0.91
1.03
0.03
0.99
0.02
0.95
0.09
1.72
2.13
0.28
0.73
**
0.97
2.70
*
0.14
0.37
0.97
0.76
0.38
0.58
2.91
**
0.12
0.80
0.48
1.04
0.38
†
0.32
0.12
0.83
1.57
0.10
0.68
1.57
0.76
0.81
0.39
1.98
0.50
1.39
0.20
1.90
2.70
3.83
0.92
1.69
4.12
4.92
7.62
11.35
2.96
7.65
11.70
153
Most Effective
(vs. Condom)
1.34
0.77
Pill
(vs. Condom)
Less than high school
0.59
0.22
High school/GED (ref.)
Some college
0.72
1.00
0.83
0.55
College degree or higher
0.63
0.52
0.85
0.05
Fertility
Age at first sex
0.90
†
0.02
0.97
0.03
Parity – 1 or more children
2.84
2.08
2.43
1.24
†p<.10; *p<.05; **p<0.01; ***p<0.001. Note: Reference category in parentheses. All analyses are weighted with SVY commands in STATA; measure of time
is included and continuous; 10,071 person-months. Subscript 1 denotes most effective methods which include hormonal methods (excluding the pill) and sterilization.
Subscript 2 denotes least effective methods and includes ‘other’ methods and none. Source: 2006-2010 National Survey of Family Growth (Female data file); period
of observation is 3 years.
154
APPENDIX A
Main articles on contraceptive method switching in the United States used in the dissertation
1. Contraceptive Method Switching in the United States (Grady et al. 2002)
Data and Methods:
-
National Survey of Family Growth 1995 (N=10847).
-
Contraceptive methods: hormonal implants, injectable, IUD, pill, condoms (including
condom use in combination with other less effective methods, less effective methods
and no method.
Switching Event:
-
Use different methods in consecutive months.
-
Use of two methods (including nonuse) separated only by a period of abstinence.
-
Use of two methods sequentially in the same month and the respondent use on of the
methods in the prior month and the other in the subsequent month.
-
Dual use is only captured among unmarried women in the sample.
Exposure Period:
-
Observation of method use truncated at 10 months prior to interview because NSFG.
data does not allow authors to identify month of conception for women who were
pregnant at the time of interview.
-
Beyond the 24 month period, months are censored.
-
Exposure intervals are right censored by end of observation, stopping use to conceive,
and change in marital status or infertility.
155
2. Injectable Contraceptive Discontinuation and Subsequent Unintended Pregnancy among
Low-Income Women (Davidson et al. 1997)
Data and Methods:
-
Sample based on interviews of women from three large hospital-based family clinics
serving poor and ethnically diverse populations in New York, Dallas and Pittsburg
(N= 491).
-
First round of interviews conducted between June 1993 and October 1994 at family
clinics who are initiating Depo-Provera use.
-
Second round of interviews conducted at 1 year post initiation.
Switching Event:
-
Switching is defined as women who discontinued Depo-Provera use at second round
of interviews and started using another method of contraception.
3. Combined Use of Condoms with Other Contraceptive Methods among Inner-City
Baltimore Women (Santelli et al. 1995)
Data and Methods:
-
Street Survey in two inner-city Baltimore communities as part of an evaluation of
HIV prevention program (N=717).
-
Two rounds of data collected in 1991 and three rounds in 1992.
-
Contraceptive methods: pill, condom, implant, diaphragm, IUD, sponge, spermicide,
condom (with any other method), condom with pill.
156
Switching Event:
-
Respondents are asked to name the method of pregnancy or STD prevention they
used the last time they had intercourse. They are also asked whether any additional
methods are used.
-
This question is asked in subsequent rounds of interview to ascertain whether
contraceptive method switching occurs.
Exposure Period:
-
Women are interviewed and re-interviewed during a one year period. Sample
restrictions include: currently pregnant women, those pregnant the last time they had
intercourse, women who had been surgically sterilized.
4. Characteristics of Injectable Contraceptive Users in a Low-Income Population in Texas
(Sangi-Haghpeykar et al. 1995)
Data and Methods:
-
Sample based on interviews from women at 17 family planning clinics in Texas who
expressed interest in using the Depo-Provera for the first time (N= 600).
-
First round of interviews conducted between October 1993 and September 1994.
-
Participants who visit the clinics for contraceptive methods are counseled on the
efficacy of the method they choose and then told about the survey.
-
They complete a self-administered questionnaire and then are followed for one year
after receiving Depo-Provera.
157
Switching Event:
-
Switching is measured at follow up if women have not returned to the clinic for their
Depo-Provera injection and are using a different method of contraception.
5. Consistency of Condom Use for Disease Prevention among Adolescent Users of Oral
Contraceptives (Weisman et al. 1991)
Data and Methods:
-
Three-wave panel study of adolescent women’s contraceptive decision-making
-
430 unmarried, non pregnant adolescents (11-18 years) receive baseline interview
-
Follow-up interviews are conducted at three month intervals and at the end of the six
month follow-up period information on STDs and pregnancies is collected.
Switching Event:
-
Change in type of contraception used at baseline and contraceptive methods used at
any period during the six month follow-up period.
158
APPENDIX B
Data Construction for Event History Analysis
Data from the 2006-2010 NSFG are used to construct the variables and produce descriptive and
multivariate analyses. There are 12,279 women interviewed in this survey. An event history file is
created that represents the person-months of exposure starting with three years prior to interview. The
statistical software used for the data construction is STATA version 12. In the first stage of data
construction the original NSFG 2006-10 data in its wide data format, each individual is a case, is
examined for duplicate records of century month dates among married and cohabiting women.
Results indicate that there are 37 duplicate files for cohabiting women. In an effort not to delete
these cases, only the first cohabitation dates for each case are kept. The original sample size is
maintained (N= 12,279). Two files are generated based on marital and cohabitation histories. I
reshape the original data set from wide to long format such that each individual has a record for
each marriage. From the original file, century month dates of marriage formation and dissolution
are kept in the data with the respondents’ case identification numbers. The century month
variable names (mardat01-06 and mardis01-06) are renamed. The data is then reshaped in long
format (persons months = 73,674). Subsequently, cases are deleted from this file if century
month marital start and end dates are missing (denoted with ‘.’). The sample size now stands at
6,412. The dataset is reshaped a second time based on the cohabitation start and end dates
(cmcohstx-4 and cmstpcohx-4) such that each cohabitation is a record. The result from the
manipulation of data to the long format indicates that there are now 49,116 century month cases.
Cases with missing cohabitation start and end dates are also dropped from the analyses and the
total number of cases is 4,349. Both sub-datasets (i.e., reshaped married and cohabitation files)
are then appended to form a new union file (N=10,761). In order to establish the time period for
analyses, a start and end period is generated by subtracting 36 (months) from the century month
159
at interview. This file in its wide format is merged with the previously constructed union file to
create a union-window file. This new merged file is then expanded using the ‘month’ variable to
create an ‘expand union window file’. The ‘expand’ command is useful for this analyses as I can
replicate current observations in memory. A total of 558,864 person month observations are
created. I generate a century month (cm_month) variable in this file which is the addition of
(month + window start – 1). It should be noted that ‘month’ and ‘window start’ dates are already
century month variables and month is the timing variable used for subsequent event history
analyses.
The second stage of data construction involves the reshaping of original data to
accommodate the analyses of contraceptive methods used by women for consecutive months
during the three-year window. First, the contraceptive method history file includes responses to
questions of contraceptive use and nonuse each month and allows respondents to give a
maximum of four opportunities to state the same. This dissertation examines the use and nonuse
of contraception based on first mention records. This strategy is favored because using the first
mention dates captures the majority of respondents and there is a sharp decrease in sample size
as date records move from 2nd mention to 4th mention records. Research using NSFG 2006-2008
data by Eisenberg and colleagues (2012) indicate that the overall rate of dual use is low in the
United States. Therefore, the use of dual methods per month is not examined. Subsequently, all
cases with second, third and fourth mention dates in the contraceptive history file are dropped
from the dataset. I rename the variable names of first mention dates for each month. The data is
then reshaped to long format (person months = 589,392). I also rename the corresponding
century month dates for each first mention record. For example, ‘cmmhcalx1-48’ are variable
names for century month dates covering the data collection period of 4 years. All are renamed to
160
‘cm_month’. The next stage in this process is to delete cases where century month dates and
contraceptive methods are missing. Also, if responses to the contraceptive method history for
any month are ‘refused’ or ‘did not know’, these cases are deleted as well. This leaves the total
number of century month cases in the contraceptive method file at 423,869.
Stage three of the data construction involves merging of the expand union-widow file and
the contraceptive method file (person months= 558,864). The types of contraception that
respondents indicate they use for each month is then recoded into six (6) categories namely: pill,
condom, hormonal methods (Depo-Provera, injectables, hormonal inplant, IUD, coil, loop,
Lunelle injectable, contraceptive patch, vaginal contraceptive ring), other methods (withdrawal,
rhythm or safe method by calendar, diaphragm, female condom, foam, jelly, suppository and
sponge), sterilization (partner’s vasectomy and female sterilization) and none.
In stage four I construct a file that captures respondents within the three-year window. I
use century month union start and end dates and window start and end dates to create a
dichotomous variable called ‘event_in’. I only keep the records of respondents who are assigned
a value of 1 (i.e., records fall within the window period).
Stage five involves the construction of the data set to account for contraceptive method
switching and the file developed at stage four is used as the framework. The first step is to
generate a variable that keeps the records of respondents up to the occurrence of first
sterilization. For step two I remove all records for women who are sterilized in the first month of
the contraceptive method calendar. In step three I create a switching variable to account for the
first switch within the three year observation period. Switching by definition occurs when the
method type in the first month of the method calendar is different from the method in any of the
months following the start of the observation period. I also create century month dates for each
161
of the six method types (see stage three). Therefore, I am able to view each monthly record of an
individual and the type of method they start with and the method of contraception used following
the switch, if they switch during the observation period. In the final step of stage five I create the
main independent variable – union status at the start of the observation period. I use century
month union start and end dates as well as window start and end dates to create this categorical
variable (married, cohabiting and single) which is further recoded into 3 dummy variables. The
sample size after data manipulation is 3,532 women.
For stage six I use the original NSFG 2006-10 female file (N=12,279) to construct a new
file (predictor file) that includes control variables for subsequent analyses. The control variables
include: age, race/ethnicity, education, poverty status, family type, religious affiliation, mother’s
education, age at first sexual intercourse and parity at observation. Most of these variables are
categorical but recorded into dummy variables. For example, respondents’ education is originally
based on an 11 type category. I recode this into a smaller categorical variable (less than high
school, high school/GED, some college and college degree or higher). Each category is further
recoded into a dichotomous dummy variable. In the case of mother’s education, the value of 95
(no mother figure identified) is deleted. For religious affiliation during childhood, values 9 and
10 (refused and don’t know) are deleted. Extreme values reported for age at first sex are also
removed from the analysis, these include ages <=10.
In stage seven I merge the method switching file (long format) in stage five with the
previously constructed predictor file (wide format) with the control variables. This particular file
is merged in order to retain the long format for multivariate discrete-time event history analyses
(N=3,122).