You are on page 1of 14

ANTHONY PAIK University of Iowa

Adolescent Sexuality and the Risk of Marital


Dissolution

This research investigates whether first sexual scholars and the public (for a review, see Buhi &
intercourse during adolescence is associated Goodson, 2007). Similarly, a long-standing con-
with increased risk of first marriage dissolution cern in family research is whether adolescent
and tests whether the results are consistent with sexuality and premarital sex are linked to union
causal or selection explanations. Drawing on formation (Gaughan, 2002; Landale & Forste,
a sample of 3,793 ever-married women from 1991; Laumann, Gagnon, Michaels, & Michael,
the 2002 National Survey of Family Growth, 1994; Marini, 1985; Miller & Heaton, 1991;
this study estimated event-history models of Raley, Crissey, & Muller, 2007). In contrast,
first-marriage dissolution. Results indicate that population-based research on whether adoles-
wanted sexual debut in later adolescence cent sexual debut is associated with marital
does not directly increase the risk of marital dissolution in the United States is scant.
dissolution but is linked indirectly as a result A handful of population-based studies have
of subsequent premarital sexual outcomes. linked premarital sexual behavior to increased
Sexual debut that is not completely wanted or risk of divorce (Heaton, 2002; Kahn & London,
that occurs before age 16 is associated with 1991; Laumann et al., 1994; Teachman, 2003),
increased risk of marital dissolution. The results but they did not examine whether adolescent
suggest that the timing and context of adolescent sexual debut was particularly salient for marital
sexual debut have important implications for dissolution. There is evidence that adolescent
marital stability. sexual behavior is likely a divorce determinant.
For example, decreased relationship quality, a
Since 1988, roughly one third of U.S. female correlate of relationship instability, has been
teenagers, ages 15 – 17, reported having already linked not only to having multiple sex partners
experienced their first sexual intercourse (Abma, over the life course but also to childhood sexual
Martinez, Mosher, & Dawson, 2004). Because abuse (Browning & Laumann, 1997). Thus,
‘‘adolescent sexual debut’’ is linked to the although little is known about the relationship
number of sex partners over the life course, sex- between adolescent sexual debut and the risk of
ually transmitted diseases, and teen pregnancy subsequent marital dissolution, there are reasons
(Furstenberg, 2003; Seidman, Mosher, & Aral, to suspect that such a relationship exists.
1992), it has generated immense interest among I addressed this gap in the literature by
investigating whether adolescent sexual debut
is associated with marital dissolution among
Department of Sociology, University of Iowa, Iowa City, American women and by testing three possible
IA 52242 (anthony-paik@uiowa.edu). explanations. First, adolescent sexual debut may
This article was edited by Suzanne Bianchi. directly increase the risk of marital disruption.
Key Words: adolescence, dissolution, divorce, marriage, Sexual experiences early in the life course may
sexual behavior, premarital sex. be particularly influential in changing marital
472 Journal of Marriage and Family 73 (April 2011): 472 – 485
DOI:10.1111/j.1741-3737.2010.00819.x
Adolescent Sex and Divorce 473

and sexual attitudes, thus making marriage to one divorce is spurious, thus reflecting selection
husband less attractive. Some scholars have also on observed and unobserved characteristics.
suggested that traumatic, unwanted sexual expe- On the other hand, Heaton (2002) argued
riences during childhood and adolescence induce that the association between premarital sex
psychogenic processes repeatedly experienced and marital dissolution is causal. Premarital
throughout the life course that lead to unsta- sexual experiences, particularly with multiple
ble relationships (Finkelhor, 1988). Second, a sex partners, may induce changes in beliefs
life-course explanation suggests that adolescent and attitudes that undermine the stability of
sexual debut may be indirectly associated with marriages (Teachman, 2003). Most such studies,
marital disruption through intervening behaviors however, have not controlled for unobserved
and outcomes, such as the accumulation of pre- selection processes into premarital sex (Heaton,
marital sex partners and premarital pregnancies 2002; Laumann et al.; Teachman, 2003). One
(Browning & Laumann, 1997). Finally, asso- study rendered inconclusive results (Kahn &
ciations between adolescent sexual debut and London; Heaton, 1993).
marital dissolution may be linked to selection Second, there has been little attention to
factors (Kahn & London, 1991; Laumann et al., whether adolescent sexual debut, compared with
1994; Teachman, 2003). There may be observed delaying sex until adulthood, increases the risk of
and unobserved factors linked to marital instabil- divorce. There is some evidence that the timing
ity that also predispose some girls to adolescent of sexual debut may be important. Teachman
sexual debut. I tested for this possibility using (2003), for example, found that the age of first
two-stage models that correct for the potential sexual intercourse was negatively associated
endogeneity of adolescent sexual debut. Thus, with the risk of marital dissolution. Women
the primary goal of this research was to exam- who delayed sexual debut had lower risks of
ine whether adolescent sexual debut is linked to marital dissolution, but the implications of this
marital dissolution and to test whether observed finding were not examined. Similarly, Heaton
associations reflect selection processes, direct (2002) found that women who reported that
causation, or indirect effects, mediated by sub- their first sexual intercourse was forced had
sequent premarital sexual outcomes. increased risk of divorce, even after controlling
for premarital sex and premarital pregnancies.
Thus, although the two studies highlighted
PRIOR RESEARCH
the importance of the timing and wantedness
Only four nationally representative studies have of first intercourse for marital dissolution,
examined whether premarital sexual experiences they did not examine specifically whether
are linked to divorce (Heaton, 2002; Kahn & adolescent sexual debut mattered. Despite the
London, 1991; Laumann et al., 1994; Teachman, many contributions of prior research, there is a
2003). Nevertheless, the core finding—the asso- need for an approach that distinguishes between
ciation between premarital sex and increased selection and causation explanations and that
risks of divorce—is robust. Teachman (2003) focuses attention on whether the timing and
found that women who had sex only with unwantedness of adolescent sexuality are linked
their future husbands did not have higher risks to divorce.
of marital dissolution, which suggests that the
premarital-sex effect on divorce is related pri-
CONCEPTUAL FRAMEWORK
marily to having sex with multiple partners.
Studies of the premarital-sex effect on divorce In this research, I examined whether adoles-
are limited in two key ways, however. First, cent sexual debut is associated with marital
the nature of the association remains up for instability. I also evaluated the importance of
debate. Does the link between premarital sex three possible explanations for such an associa-
and increased risk of marital dissolution among tion. Observed associations may represent one or
women reflect selectivity or causal processes? more of the following: (1) selectivity, (2) direct
On the one hand, women predisposed to causation, or (3) indirect causation, where the
marital stability may be less likely to engage effects of adolescent sexuality are mediated by
in premarital sex (Kahn & London, 1991; subsequent premarital sexual outcomes, such as
Laumann et al., 1994). This suggests that the the accumulation of additional premarital part-
observed association between premarital sex and ners and premarital pregnancies.
474 Journal of Marriage and Family

The basic selection argument is that girls developmental role in shaping the way individ-
predisposed to marital instability may be more uals view marriage, thus making nonmarital sta-
likely to become sexually active before adult- tuses less stigmatized (Kahn & London, 1991).
hood. Selection arguments have been used to Specifically, first intercourse during adolescence
explain associations between premarital sex and may change attitudes and beliefs about marriage
marital dissolution (Kahn & London, 1991; Lau- and sex—such as beliefs about the permanency
mann et al., 1994; Teachman, 2003), as well as of marriage or the attractiveness of being sin-
between nonserious contexts of sexual involve- gle in later life—that increase the likelihood
ment and lower relationship quality (Paik 2010). of marital dissolution. They may also lead to
Two arguments have been advanced to explain the acquisition of relationship skills that under-
why selectivity matters. First, the association mine marriage (Teachman, 2003). For example,
may reflect belief heterogeneity (Kahn & Lon- these experiences may make women aware of
don; Laumann et al.; Teachman, 2003). Individ- the alternatives to remaining married or pro-
uals vary in their beliefs about the appropriate- vide them with social skills that facilitate ending
ness of divorce: There are traditionalists who do marriages and starting new relationships.
not believe that divorce is an option and nontra- In contrast, women’s adolescent sexual expe-
ditionalists who do. Beliefs about marriage may riences frequently involve elements of social
also be linked to views about the appropriateness pressure and coercion. Psychogenic theories
of premarital sex, such as adolescent sexuality. of adult – child sexual contact posit that early,
Adolescents who hold traditional beliefs about traumatic sexual experiences may cause some
divorce may be more likely to avoid becoming individuals to become highly sexualized (Finkel-
sexually active before adulthood; conversely, hor, 1988), thereby undermining the attractive-
nontraditionalists may be more likely to engage ness of remaining in lifelong relationships. To
in sexual intercourse during adolescence. the extent that unwanted first intercourse is
Second, adolescent sexual debut may be asso- common among girls, these experiences, then,
ciated with marital dissolution because of sex may directly increase their likelihood of mari-
heterogeneity (Laumann et al., 1994). Specifi- tal instability (Heaton, 2002). Moreover, even
cally, some individuals may be highly interested adolescent sexual experiences that are neither
in sex. They are unlikely to remain sexually completely wanted nor unwanted—that is, those
exclusive and, therefore, likely to divorce. Such associated with mixed feelings—may increase
individuals are also more likely to select into the risk of divorce. Mixed feelings about first
first sexual intercourse before adulthood. Belief intercourse may be related to social pressure,
and sex heterogeneity both suggest that indi- being manipulated, or dissatisfaction with the
viduals predisposed to marital dissolution will experience itself. Negative experiences may also
select into adolescent sexual debut. To the extent change marital and sexual attitudes, thereby
that selection processes are operative, a positive making remaining in marriages less attractive.
association between adolescent sexual debut and In contrast, adolescent first intercourse that
the risk of marital instability would be spurious was wanted—experiences not associated with
and should disappear to the extent that selection negative or traumatic emotions—may not be
characteristics, both observed and unobserved, associated with an increased risk of marital dis-
are controlled for. solution. Taken together, these distinct causation
arguments suggest the following hypotheses:
H1 (selection hypothesis): First intercourse during
adolescence is associated with an increased risk H2a (experience hypothesis): First intercourse
of marital dissolution but reflects the selectivity of during adolescence directly increases the risk of
individuals predisposed to marital disruption into marital dissolution.
adolescent sexual debut.
H2b (coercion hypothesis): First intercourse during
adolescence that is not completely wanted is
A direct causation argument, in contrast, sug-
associated with an increased risk of marital
gests that the experience of adolescent sexual instability, but wanted sexual debut is not
debut directly influences the risk of marital dis- associated with increased hazards.
solution. If formative sexual experiences occur
particularly early in the life course, such as The third explanation suggests that the effect
during adolescence, they may play a critical of adolescent sexual debut on marital instability
Adolescent Sex and Divorce 475

is indirect, mediated by subsequent premari- Colella, 1992). It is worth highlighting that


tal sexual outcomes. Specifically, there are two the association between premarital cohabitation
potentially mediating demographic processes. and divorce appears to reflect selectivity; that
First, a life-course perspective suggests that is, individuals who cohabit before marriage
changes in attitudes about the permanency of have characteristics that predispose them to
marriage, the attractiveness of being single, divorce. One issue is that adolescent sexuality
the sexualization of girls, and the acquisition and premarital sex may be important selection
of relationship skills are not solely the prod- factors for premarital cohabitation, so a key
uct of the transition to first intercourse but are question centers on whether the effect of the
cumulative processes linked to having more sex latter is attenuated by controlling for prior
partners. Although the onset of first intercourse sexual behaviors. Finally, prior research has also
represents the life-course transition to coupled observed racial/ethnic and religious differences
sexual behavior, many individuals engage in in marital stability (Heaton, 2002), as well as the
sexual trajectories marked by multiple sexual importance of parents’ educational attainment,
relationships. This pattern of sexual behavior is number of siblings, and whether husbands were
likely associated with incremental and cumu- previously married or had children from prior
lative changes in attitudes, sexualization, and relationships (Teachman, 2003).
relationship skills. Indeed, Browning and Lau- To summarize, this research sought to make
mann (1997) found that the number of sex the following contributions. First, I examined
partners over the life course mediated the effects whether adolescent sexuality is linked to marital
of adult – child sexual contact on well-being and instability, which, to the best of my knowledge,
relationship quality. Likewise, Teachman (2003) has not been a focus in prior research. Sec-
found that women who had premarital sex part- ond, this research sought to adjudicate among
ners who were not future husbands had higher competing arguments regarding the association
rates of marital instability. Second, because between adolescent sexual debut and marital
adolescent sexual debut is linked to premari- instability. I present a conceptual framework
tal conceptions and live births, both of which that distinguishes selection effects, direct causa-
are well-known divorce determinants (Billy, tion, and indirect effects related to intervening
Landale, & McLaughlin, 1986; Teachman, premarital sexual outcomes. The relative impor-
1983), it is possible that the observed associ- tance of selectivity versus causal explanations is
ation reflects this fact (Kahn & London, 1991). key to understanding whether adolescent sexual
These factors suggest the following hypothesis: debut is a divorce determinant.

H3 (indirect-effects hypothesis): Differential sex- METHOD


ual trajectories, premarital pregnancies, and pre-
marital births mediate the positive association This research employed the women’s survey of
between first intercourse during adolescence and the 2002 National Survey of Family Growth
the increased risk of marital instability. (NSFG), a multistage probability sample of
7,643 noninstitutionalized women between the
I controlled for several previously identified ages of 15 and 44, which was collected by the
correlates of divorce. Prior research has focused National Center for Health Statistics (Lepkowski
on several key risk factors of divorce (Heaton, et al., 2006). Interviews, averaging 85 minutes,
2003; Teachman, 2002), including being raised were conducted in person using computer-
in nonintact families (Bumpass, Martin, & assisted personal interviewing (CAPI), and the
Sweet, 1991), lower educational attainment survey instrument collected detailed informa-
(South & Spitze, 1986), early age at marriage tion on marriages, cohabitations, pregnancies,
(Becker, Landes, & Michael, 1977; Lehrer, sex partners, and related items on sexual health
2008), heterogamous characteristics between and reproductive outcomes. At 80%, the overall
husbands and wives (Bratter & King, 2008; response rate for the sample was quite high.
Lehrer & Chiswick, 1993; Teachman, 2003; Unlike previous cycles, a major limitation of
Tzeng, 1992), and premarital cohabitation the 2002 NSFG female sample is that there
(Bennett, Blanc, & Bloom, 1988; DeMaris & was a key routing error in the survey that
Rao, 1992; Lillard, Brien, & Waite, 1995; produced nonrandom missing data. The CAPI
Teachman & Polonko, 1990; Thomson & program incorrectly routed approximately 500
476 Journal of Marriage and Family

women, who had reported that their marriages by 3% to 3,793 respondents for my models of
had ended, out of the key questions about their marital disruption. Table 1 presents descriptive
relationship outcomes (i.e., divorce, separation, statistics for independent variables, described
or widowhood) and the timing of marital dis- below, that were used in this research.
ruption. Almost all these cases were women
who had married only once but were currently
separated or women who had remarried but Table 1. Unweighted Descriptive Statistics of Independent
had first husbands who had previously fathered Measures, Women, 2002 NSFG (N = 3,793)
children. The vast majority of the latter cases Variables Mean or Prop. SD
were likely divorces, as widowhood is rare
at these ages; the imputation procedure esti- First intercourse (ref.: wanted, 0.22
mated that only 13 of the respondents with 18 or older)
missing data were widows. The 2002 NSFG sup- Wanted
plied imputed values, computed using multiple 13 or younger 0.01
regression imputation, for the timing of marital 14 – 15 0.05
disruption (Lepkowski et al., 2006). Researchers 16 – 17 0.10
have used these imputed values to study marital Not wanted
dissolution previously (Bratter & King, 2009; 13 or younger 0.06
Lehrer, 2008, Teachman, 2008). Analyses of the 14 – 15 0.14
16 – 17 0.22
imputed data have indicated that the results are
18 or older 0.21
similar to those produced by the 1995 NSFG as
Premarital sex (ref.: none) 0.09
well as by the men’s sample of the 2002 NSFG
Sex with future husband only 0.20
and that the imputed values had little impact on Sex with multiple partners 0.27
the results (Lehrer; Teachman, 2008). Although Premarital conception 0.23
multiple imputation is designed specifically for Premarital birth 0.21
data that are missing at random, simulations Age at marriage (in years) 23.20 4.61
have also shown that multiple imputation per- Educational attainment (in years) 13.74 2.60
forms well on data that are not missing at random Race/ethnicity (ref.: White/other) 0.63
(Buhi, Goodson, & Neilands, 2008). The possi- Latina 0.20
bility of biased coefficients, due to the imputed African American 0.14
values for data that are not missing at random, Asian 0.03
is a limitation of the data. Foreign born 0.18
This research examined the timing of mar- Religion raised (ref.: other) 0.29
ital disruption among ever-married women Catholic 0.38
(N = 4,051). Because of my interest in the Non-mainline Protestant 0.34
effects of adolescent sexual debut, I did not Parents’ average education (in 12.36 1.82
exclude women who married early, a selection years)
criterion used frequently in the study of mari- Number of siblings 3.66 1.64
tal dissolution (Heaton, 2002; Kahn & London, Nonintact family 0.26
1991; Teachman, 2003). Because many states Age at menarche 12.57 1.66
allow 16-year-olds to marry with parental con- Age difference (ref.: wife is 5+ 0.28
sent, I included only women who married at years younger)
that age or later. A question about whether first Wife: 2 years older to 0.68
sexual intercourse was wanted or unwanted was 5 years younger
limited to respondents who were at least age Wife: more than 2 years 0.05
18 at the time of the survey; consequently, sev- older
eral additional cases from the most recent birth Husband is different race 0.12
Husband was married before 0.15
cohort were excluded from the analyses. These
Husband had children before 0.17
criteria reduced the analytical sample to 3,923
Premarital cohabitation (ref.: 0.51
respondents with valid responses on age of mar-
none)
riage. I limited the use of imputed values only
Engaged before cohabitation 0.28
to the type of marital disruption as well as the Not engaged before 0.21
length of the marriages. Consequently, missing cohabitation
data on other variables reduced the sample size
Adolescent Sex and Divorce 477

Dependent Measure ‘‘At the time you first had sexual intercourse
Marital dissolution. The dependent variable with (first partner), how would you describe your
assessed whether first marriages transitioned to relationship with him?’’ Answers categories for
separation, divorce, or annulment, whichever the latter question included ‘‘married to him’’ as
came first, among ever-married women for each well as nonmarital statuses, such as being ‘‘just
friends,’’ ‘‘going with him,’’ and ‘‘engaged.’’
year they were married. First marriages, which
Women reporting no sex partners before mar-
were still current at the time of the survey, were
riage and being married to their first partner
treated as right censored. In addition, the year of
were coded as ‘‘virgins’’; those reporting one
widowhood was also treated as right censored.
partner before marriage and not being married to
their first partner were coded as having had pre-
Independent Measures marital sex with future husbands. Some women
made inconsistent reports on the two questions;
First intercourse. In the main interview, respon- they indicated having no partners before mar-
dents were asked the following: ‘‘That very first riage but reported their first sexual intercourse
time that you had sexual intercourse with a as occurring in a nonmarital status with their
man, how old were you?’’ Later, in the audio future husbands, or they reported one premarital
computer-administered, self-answered interview sex partner but reported first sexual intercourse
(ACASI) section of the survey, respondents were at marriage. I coded these inconsistent reports
asked twice about their age at ‘‘first vaginal as women who had sex premaritally with their
intercourse.’’ I took the average of the responses future husbands. I created a third category indi-
with valid data to assess age at first inter- cating women who had premarital sex with
course. I used age at marriage if respondents someone other than their future husbands and
did not have premarital sex. Also, in the ACASI also coded for whether respondents had a pre-
section, respondents were asked whether they marital conception or live birth. The final vari-
‘‘really wanted,’’ ‘‘had mixed feelings about,’’ able included the following mutually exclusive
or ‘‘really didn’t want’’ their first vaginal inter- categories for premarital sexual outcomes: none
course to happen at the time. I combined infor- (i.e., virgins at marriage), sex only with future
mation about the age and wantedness of first husbands (no premarital pregnancies), sex with
intercourse to indicate the following categories: multiple partners (no premarital pregnancies),
wanted (13 or younger, 14 – 15, 16 – 17, and premarital conception, and premarital birth.
18 or older); not completely wanted before age
18; and not completely wanted, 18 and older. Control variables. I controlled for marital dura-
It is worth highlighting that the questionnaire tion (in years) and marriage cohort, which
item for the wantedness of first sex is limited had the following categories: 1958 – 1965,
in several ways. Although the placement of this 1966 – 1975, and 1976 – 1985. I also controlled
item in the ACASI section likely reduces social for several previously mentioned factors that
desirability biases, it is based on a single retro- have been identified as consistent predictors of
spective measure, so recall bias is an issue. The marital dissolution (Heaton, 2002; Teachman,
meanings respondents attributed to the ‘‘mixed 2002, 2003). I assessed age at marriage and
feelings’’ category are somewhat ambiguous. educational attainment in years. The lack of tim-
ing information for higher degrees in the NSFG
Premarital sexual outcomes. Building on limited my ability to include this variable as
research that has highlighted the importance a time-varying covariate. Race/ethnicity distin-
of premarital sex and premarital fertility for guished among Hispanic American/Latina, non-
marital dissolution (Heaton, 2002; Kahn & Lon- Hispanic African American, non-Hispanic Asian
don 1991; Laumann et al., 1994; Teachman, American, and non-Hispanic White/other. I also
1983, 2002, 2003), I constructed a variable created a dummy variable indicating whether
that captured patterns of sexual experiences and respondents were born outside of the United
reproductive outcomes preceding marriage. I States. Two dummy variables identified respon-
used two questions to assess premarital sexual dents who were raised as Catholics or mem-
experiences: ‘‘Including your (former) husband, bers of conservative Protestant denominations
how many male sexual partners did you have (i.e., non-mainline Protestants). I calculated
before you got married (the first time)?’’ and average parental education in years; if only one
478 Journal of Marriage and Family

response was available, I used that information two-step, Heckman approach relies on a joint
only. I also included a dummy for women who normality assumption in the error terms for the
had a nonintact family at age 14—that is, they did selection and outcome equations. In later mod-
not live with both biological and/or adoptive par- els examining the effects of the wantedness of
ents—and a continuous measure for the number adolescent sexual debut on the risk of mari-
of siblings, ranging from 0 to 6 plus, with whom tal dissolution, I employed the more commonly
each woman grew up. Finally, respondents were used discrete-time logit specification, primar-
asked about their age of menarche. ily because hazard ratios are easier to interpret
In terms of husbands’ characteristics, I con- than probit coefficients. All models included
trolled for age differences, distinguishing wives weights for unequal probabilities of selection
that had partners who were 5 years older or in the sampling, stratification, and clustering in
more as well as those who were more than the sampling design and the measures for mar-
2 years younger. I included a dummy variable riage cohort and age of marriage to control for
indicating husbands of a different race/ethnicity selection biases associated with censoring.
from their wife. I did not assess religious and
educational heterogamy. The 2002 NSFG did
not collect information about the religious affili- RESULTS
ations of husbands. The measure for educational I first present descriptive results for adolescent
attainment was limited to current and sepa- sexual debut on premarital sexual outcomes and
rated husbands, so it was incomplete. Following divorce. As Table 1 shows, adolescent sexual
Teachman (2003), I included two dummies for debut is common among ever-married women
husbands who had married before and who had in the 2002 NSFG, but most of the experiences
fathered children previously. during adolescence were not completely wanted.
For example, only 1% of women reported want-
ing their first intercourse before age 14, but
Models
6% had first sexual experiences that were not
Consistent with prior research (Heaton, 2002; completely wanted during the same time frame.
Kahn & London, 1991; Laumann et al., 1994), In Table 2, the adjusted Wald tests indicate
I employed discrete-time event-history models that premarital sexual outcomes and divorce
(Allison, 1982), which rendered maximum-
likelihood estimates for the independent
variables on risks of marital dissolution. To Table 2. Premarital Sexual Outcomes and Divorce by First
estimate the discrete-time models, I transformed Sexual Intercourse Before 18 (N = 3,793)
the data into person-years (n = 29,999). Fol-
First Sexual Intercourse
lowing Kahn and London’s bivariate probit
Before 18
model, I adopted a two-step, Heckman correc-
tion for endogeneity (Hamilton & Nickerson, Variables No Yes
2003), which controlled for nonrandom selec-
Premarital sexa
tion into adolescent sexual debut related to
None 0.23 0.01∗∗∗
unobserved characteristics. First, I estimated a
Sex with future husband only 0.29 0.14∗∗∗
probit selection model on a binary specifica-
Sex with multiple partners 0.24 0.31∗∗
tion of adolescent sexual debut, which included
Premarital conception 0.15 0.29∗∗∗
age of menarche as an instrumental variable.
Premarital birth 0.10 0.26∗∗∗
Age of menarche is an important correlate of ∗∗∗
Marital dissolutionb
adolescent sexual debut (Udry, 1979) but is
Within the first 5 years 0.15 0.31
unrelated, both theoretically and empirically,
Within the first 10 years 0.27 0.47
to the risk of marital dissolution. Using the
Proportion of sample 0.42 0.58
results of selection probit model, I estimated the
inverse Mills ratio, parameterized as lambda, Note: Log-rank statistics tested for equality of survivor
which controlled for unobserved selectivity in functions.
a Tests of equality for proportions were based on adjusted
adolescent sexual debut and was included in
the discrete time probit of marital dissolution. I Wald tests. b Marital dissolution statistics were based on
employed the probit specification over the more product-limit estimation.
∗∗ p < .01. ∗∗∗ p < .001.
commonly used discrete-time logit because the
Adolescent Sex and Divorce 479

vary by whether women experienced adoles- have increased probabilities. Age of menarche is
cent sexual debut. More than half of the women an effective instrumental variable. It is a key pre-
who delayed first sexual intercourse until adult- dictor: Later age of menarche is associated with
hood were virgins at marriage or had their first lower likelihoods of adolescent sexual debut,
experience with their future husbands. Only but it is not associated with marital dissolution
15% of women experienced adolescent sexual (results are available on request). The results
debut at marriage or with their future husbands. of this probit were used to estimate the inverse
Larger proportions of ever-married women who Mills ratio, parameterized as lambda, which was
experienced adolescent sexual debut had had included in discrete-time probit of marital dis-
premarital sex with multiple partners (31% vs. solution (Model 2).
24%), premarital conceptions (29% vs. 15%), Second, in the marital-dissolution model,
and premarital births (26% vs. 10%). A higher the coefficient for adolescent sexual debut is
proportion of women who had first sexual positive and significant: Women who reported
intercourse before the age of 18 experienced adolescent sexual debut have an increased risk
marital dissolution. Nearly 31% and 47% of of marital dissolution. The coefficient for the
ever-married women who experienced adoles- inverse Mills ratio (lambda) is nonsignificant
cent sexual debut had their marriages dissolve and negative. These results do not support the
within 5 and 10 years, respectively. The corre- notion that unobserved characteristics associated
sponding percentages for women who delayed with adolescent sexual debut are linked to
sex until adulthood are far lower at 15% and increased risk of marital dissolution. To check
27%, respectively. Clearly, adolescent sexual the robustness of this result, I reestimated
debut is frequently not completely wanted and this endogeneity test using the two-stage,
is associated with both marital dissolution and residual inclusion method (2SRI), where the
premarital sexual outcomes—sex with multiple first-stage, ordinary-least-squares outcome was
partners and premarital pregnancies—which are age of first intercourse, and the second-stage
determinants of divorce. These results highlight survival model included first-stage residuals
the need for understanding the systematic rela- (Terza, Basu, & Rathouz, 2008). The results
tionships among age and wantedness of first of 2SRI model also did not support the selection
intercourse, premarital sexual experiences, and explanation (results available on request). Taken
marital dissolution. together, the results do not support the selection
To test whether selection processes account argument (Hypothesis 1), which posited that the
for the association between adolescent sexual association between adolescent sexual debut and
debut and marital dissolution, I first estimated marital dissolution was spurious and the result
a discrete-time probit model that controlled for of selectivity.
unobserved selection characteristics in Table 3. If the association between adolescent sexual
All discrete-time event-history models con- debut and marital dissolution is not due to selec-
trolled for age cohort, age at marriage, and dura- tivity, is there evidence that sexual experiences
tion dependency (i.e., marital duration). Before in adolescence lead to increased risk of mari-
modeling the remaining covariates, I tested tal dissolution? Table 4 addresses this question.
several specifications of duration dependency, Because the selectivity parameter in Table 3 is
including constant hazards, linear, quadratic, nonsignificant and hazard ratios are easier to
logarithmic, piecewise, and temporal dummies. interpret than probit coefficients, I reestimated
The Akaike information criterion (AIC) indi- the models of marital dissolution using discrete-
cated that the linear specification for duration time logits in Table 4. Model 1 shows that the
dependency was the most parsimonious (results age and context of first intercourse are associated
available on request). with the risk of marital dissolution. Wanted first
The results from Table 3 are inconsistent with intercourse during adolescence as well as not-
the selectivity explanation. First, the selection wanted sexual debut throughout the life course
probit (Model 1) indicates that Latinas, Asian are associated with increased rates of marital
Americans, the foreign born, respondents with dissolution in comparison to women reporting
more educated parents, and those with more wanted first intercourse during adulthood. Con-
siblings all have lower likelihoods of adoles- trolling for many determinants of divorce in
cent sexual debut, whereas respondents who are Model 2 attenuates the coefficients somewhat,
African American and from nonintact families but they all remain statistically significant.
480 Journal of Marriage and Family

Table 3. Probit Model of Adolescent Sexual Debut and Discrete-Time Probit Models of Marital Dissolution

Model 1 Model 2
First Sex Before 18 Dissolution
Variables b SE b SE

First sexual intercourse before 18 0.61 0.31∗


Lambda −0.30 0.19
Premarital sex (ref.: none)
Sex with future husband only 0.20 0.09∗
Sex with multiple partners 0.32 0.10∗∗
Premarital conception 0.44 0.10∗∗∗
Premarital birth 0.49 0.11∗∗∗
Age at marriage (in years) −0.05 0.01∗∗∗
Educational attainment (in years) 0.00 0.01
Race/ethnicity (ref.: White/other)
Latina −0.14 0.08∗ −0.05 0.07
African American 0.16 0.07∗ 0.11 0.06∗
Asian −0.49 0.22∗ −0.02 0.11
Foreign born −0.59 0.08∗∗∗ 0.06 0.09
Religion raised (ref.: other)
Catholic −0.02 0.08 0.08 0.06
Non-mainline Protestant −0.07 0.07 0.03 0.05
Parents’ average education (in years) −0.10 0.02∗∗∗ 0.03 0.02∗
Number of siblings −0.08 0.03∗∗ 0.00 0.02
Nonintact family 0.51 0.06∗∗∗ 0.04 0.07
Age difference (ref.: wife 5+ years younger)
Wife: 2 years older to 5 years younger 0.02 0.05
Wife: more than 2 years older 0.09 0.11
Husband is different race 0.10 0.06
Husband was married before 0.15 0.06∗∗
Husband had children before 0.14 0.06∗
Premarital cohabitation (ref.: none)
Engaged before cohabitation −0.08 0.05∗
Not engaged before cohabitation 0.06 0.05
Age of menarche −0.07 0.02∗∗∗
Constant 2.46 0.34∗∗∗ −1.57 0.43∗∗∗
Model F 25.21∗∗∗ 12.67∗∗∗
N 3,793 29,999
Note: The discrete-time probit controls for marriage cohort and a linear specification of duration dependency.
∗ p < .05. ∗∗ p < .01. ∗∗∗ p < .001 (one-tailed test).

Model 3 includes the measure for subse- premarital fertility. In contrast, women who
quent premarital sexual outcomes. The inclu- reported wanted first intercourse before age
sion of these dummy variables attenuates the 14 (hazard ratio [HR] = 1.9) and from 14 to
effect of wanted first intercourse at ages 16 – 17 15 (HR = 1.8) or did not completely want
to nonsignificance. Women who reported a their sexual debut in adolescence (HR = 1.7)
wanted sexual debut in later adolescence are or adulthood (HR = 1.4) have increased rates
not directly more likely to have increased rates of marital dissolution, even after controlling for
of marital dissolution; rather, they are indirectly subsequent premarital sexual outcomes. More-
exposed to higher risks of marital dissolution over, adolescent first intercourse that was not
because of higher probabilities of accumulat- completely wanted, in comparison to unwanted
ing premarital sex partners and experiencing experiences in adulthood, is associated with
Adolescent Sex and Divorce 481

Table 4. Discrete-Time Logit Models of Marital Dissolution on Adolescent Sexual Debut and Controls (n = 29,999)

Model 4: Only
Model 1 Model 2 Model 3 Nonimputed
Variables HR SE HR SE HR SE HR SE

Age at marriage (in years) 0.92 0.01∗∗∗ 0.90 0.01∗∗∗ 0.89 0.01∗∗∗ 0.87 0.02∗∗∗
Educational attainment (in years) 0.98 0.02 0.99 0.02 1.00 0.02
Race/ethnicity (ref.: White/other)
Latina 0.85 0.12 0.85 0.12 0.75 0.12∗
African American 1.45 0.16∗∗∗ 1.34 0.16∗∗ 1.24 0.20
Asian 0.74 0.17 0.79 0.18 0.76 0.25
Foreign born 0.81 0.11 0.87 0.12 0.77 0.13
Religion raised (other)
Catholic 1.20 0.15 1.21 0.15 1.29 0.19∗
Non-mainline Protestant 1.04 0.10 1.07 0.10 1.13 0.13
Parents’ average education (in years) 1.03 0.03 1.03 0.03 1.04 0.03
Number of siblings 0.98 0.03 0.98 0.03 0.97 0.03
Nonintact family 1.33 0.13∗∗ 1.30 0.13∗∗ 1.36 0.16∗∗
Age difference (ref.: wife 5+ years younger)
Wife: 2 years older to 5 years younger 1.08 0.11 1.07 0.11 1.04 0.15
Wife: more than 2 years older 1.25 0.33 1.22 0.32 1.03 0.39
Husband is different race 1.24 0.16 1.22 0.16 1.34 0.26
Husband was married before 1.35 0.18∗ 1.41 0.18∗∗ 1.94 0.31∗∗∗
Husband had children before 1.42 0.18∗∗ 1.33 0.18∗ 0.05 0.05∗∗
Premarital cohabitation (ref.: none)
Engaged before cohabitation 0.90 0.09 0.82 0.09∗ 0.87 0.12
Not engaged before cohabitation 1.23 0.13∗ 1.12 0.13 1.08 0.16
First intercourse (ref.: wanted, 18 or older)
Not wanted, before 18 2.84 0.42∗∗∗ 2.33 0.35∗∗∗ 1.68 0.25∗∗∗ 1.77 0.32∗∗∗
Not wanted, 18+ 1.66 0.26∗∗∗ 1.61 0.26∗∗ 1.35 0.21∗ 1.50 0.30∗
13 or younger 3.47 1.08∗∗∗ 2.84 0.94∗∗∗ 1.94 0.65∗ 1.92 0.77
14 – 15 3.08 0.57∗∗∗ 2.56 0.49∗∗∗ 1.84 0.35∗∗∗ 1.82 0.47∗
16 – 17 1.82 0.36∗∗ 1.53 0.32∗ 1.21 0.24 1.31 0.31
Premarital sex (ref: none)
Sex with future husband only 1.57 0.36∗ 1.46 0.38
Sex with multiple partners 1.91 0.51∗∗ 1.79 0.55∗
Premarital conception 2.47 0.63∗∗∗ 2.51 0.74∗∗
Premarital birth 2.70 0.73∗∗∗ 2.41 0.76∗∗
Model F 23.42∗∗∗ 13.13∗∗∗ 11.18∗∗∗ 12.52∗∗∗
Note: All models include controls for marriage cohort and a linear specification of duration dependency.

p < .05. ∗∗ p < .01. ∗∗∗ p < .001 (one-tailed test).

higher risks of divorce (p < .05). To exam- were not completely wanted are associated with
ine the robustness of these findings in Model increase risks of marital dissolution, but wanted
3, I also tested alternative specifications of the first intercourse in later adolescence is only indi-
first-intercourse variable, including linear and rectly linked to rates of divorce.
annual dummies for age of intercourse. The To summarize, I found some support for
measure presented in Table 4 was the most Hypotheses 2a, 2b, and 3. Consistent with
parsimonious and explained the most variance, Hypothesis 2a, early onset of first intercourse
according to its AIC and pseudo R 2 (results is associated with increased rates of divorce,
available on request). The findings suggest but adolescents who delayed first intercourse
that sexual debut occurring before age 16 or until age 16 do not have higher risks. Likewise,
adolescent first intercourse experiences that the results support the notion that adolescent
482 Journal of Marriage and Family

first intercourse that is not completely wanted DISCUSSION


is associated with the risk of divorce. Finally, In this research, I examined whether first sex-
the results show that premarital sexual outcomes ual intercourse during adolescence is associated
partially mediate the associations between early with increased risk of marital dissolution. The
and not-wanted sexual debut during adoles- results show that both the age and the context
cence; wanted first intercourse during later of first intercourse matter for the risk of divorce.
adolescence is completely mediated. Women who reported that they wanted their sex-
In addition to these core findings, there are ual debut in late adolescence, ages 16 and 17,
several additional findings in Model 3 worth do not directly have an increased risk of marital
highlighting. First, some, but not all, of the con- dissolution. The onset of sexual activity at these
sistent determinants of divorce are significant ages is associated with subsequent premarital
in Model 3. Higher rates of marital dissolution sexual outcomes, such as having multiple pre-
are associated with being African American, marital partners and premarital fertility, both of
early age at marriage, being raised in a nonintact which are determinants of divorce. I observed
family, and having a husband who married or direct effects of sexual debut before age 16 and
fathered children previously. Second, there is adolescent first intercourse that was not com-
some evidence that the premarital cohabitation pletely wanted on risks of marital dissolution.
effect is related to premarital sex. In a model with The results do not support the selection expla-
controls only (not shown), the hazard ratio for nation. Thus, this research contributes to the
those who were not engaged but cohabited before literature by highlighting the importance of sex-
marriage was 1.4 (p< .01). Once I controlled ual debut in early adolescence or experiences that
for adolescent sexuality and premarital sexual are not completely wanted as factors linked, both
outcomes in Model 3, the coefficient became directly and indirectly, to marital disruption.
nonsignificant. This finding supports the notion This research makes several additional con-
that sexual histories may be a key selection factor tributions. First, it extends the existing litera-
explaining the association between premarital ture on premarital sex and marital dissolution
cohabitation and divorce. Interestingly, being by highlighting the significance of adolescent
engaged before premarital cohabitation is asso- sexual debut. Previous research has focused
ciated with a lower risk of marital dissolution, primarily on only three contrasts: virginity at
in comparison with those who did not cohabit marriage, premarital sex with future husbands,
premaritally (p < .05) and the other premarital and multiple premarital sex partners. Second,
cohabitors (p < .01). Third, premarital sexual this research tests selection and causal expla-
outcomes all increase the risk of marital dissolu- nations for linking adolescent sexual debut to
tion. In contrast to Teachman’s (2003) finding, marital dissolution. My results support the argu-
women who had sex with their future husbands ment that observed linkages between adolescent
have a greater risk of marital dissolution than do sexual debut and marital disruption is one not
women who delayed sexual debut until marriage. of selectivity but of changes in beliefs and atti-
Nevertheless, the risk is lower than those with tudes about marriage and relationships that result
multiple premarital partners (p < .10), premari- directly and indirectly from these experiences.
tal conceptions (p < .001), and premarital births First-intercourse experiences that were not com-
(p < .001). Finally, Model 4 excludes cases that pletely wanted or occurred before the age of
had imputed values for the type and timing of 16, in particular, appear to have lasting impacts
marital dissolution. As I mentioned previously, on the stability of women’s marriages, but pre-
most women who had husbands with children marital sexual outcomes mediate the effect of
from prior relationships had missing data. Con- sexual debut in later adolescence. The pro-
sequently, the coefficient for this variable is quite cesses giving rise to these associations are a key
different. In contrast, the findings for my core issue. I have argued that these formative sex-
variables are quite consistent. One difference is ual experiences directly change attitudes toward
that coefficient for wanted first intercourse at marriage and sex or lead to these changes
age 13 or less is now nonsignificant, but this indirectly through later life-course transitions,
appears to be a result of the decreased num- such as the accumulation of sexual histories or
ber of cases, which suggests that inclusion of experiencing premarital fertility. In this latter
the imputed values has little effect on the key argument, changes in attitudes and skills accu-
results. mulate as women move through different sexual
Adolescent Sex and Divorce 483

trajectories. It is possible, however, that other Of course, I cannot be certain that the counterfac-
mechanisms and processes are at play, which tual would hold—that is, whether my core results
future research could address. would still be observed if the actual data were
This research also raises several additional available. These results should be viewed with
questions for future research. Consistent with circumspection, but they are important because
Teachman (2003), I found that adolescent sex- they focus future research on this topic. Sec-
uality and premarital sexual outcomes may be ond, the 2002 NSFG lacked information for
important for explaining the well-known pre- critical variables, such as the timing of educa-
marital cohabitation effect. I also found that tional attainment as well as information about the
women who were engaged before cohabiting characteristics of husbands. These are data limi-
premaritally had lower risks of marital dissolu- tations that are common in other studies as well.
tion than those who did not cohabit premaritally Third, the observed direct effects will need to be
and women who were not engaged when they replicated in future research. The cross-sectional
began cohabiting. Future research should exam- design of this study is not ideal for either making
ine both findings. I did not replicate Teachman’s causal claims or rejecting selection explana-
(2003) finding that women who had premarital tions. Moreover, it is possible that selection or
sex only with their future husbands were not at indirect effects would have been important if
greater risk of marital dissolution in compari- I had access to more extensive measures. For
son to women who were virgins at marriage. example, the results of the two-step selection
I note, however, that the observed association model rely on specifying adequate instrumental
was not particularly strong—close to nonsignif- variables. Nevertheless, I have attempted to con-
icance—and that women who had sex only with trol for selection effects given the data available.
their future husbands had lower risks of divorce Although I did not observe a significant selec-
than other categories of premarital sexual out- tion effect, future research will need to confirm
comes. Indeed, the discrepancy could simply this result. Finally, the findings of this research
be the result of a false-positive observed here, depend on the item about the wantedness of first
linked to sampling variability. Finally, I did intercourse. Recall bias may contaminate this
not test whether the lower likelihood of divorce measure, and respondents likely attributed multi-
among women who were virgins at marriage ple meanings to the ‘‘mixed feelings’’ category.
reflected selectivity or causal explanations; this More important, it is possible that the associa-
question remains unresolved. tion between not completely wanting adolescent
It is important to highlight the limitations sexual debut and marital dissolution is spurious.
of this study. Chief among them is the use of For example, personality characteristics linked
imputed values primarily for the timing of mari- to attachment styles may be associated with the
tal dissolution. As I discussed above, nearly one presence of mixed feelings at both first inter-
third of ever-married women reporting marital course and divorce (reviewer comment). Future
disruption had values imputed for whether the research should examine these issues.
marriage ended as a result of widowhood, sepa- This research shows that adolescent sexu-
ration, or divorce or the timing of marital disrup- ality/premarital sex is associated with marital
tion. The former case is unlikely to be important, dissolution, but few definitive conclusions about
because widowhood is rare at these ages. In the nature of this association should be drawn.
contrast, the latter may introduce biases in the Whether this association is causal or spurious
coefficient estimates. The NSFG used regression is a question that should be revisited. This
imputation, which some have suggested only research does show that women who experi-
increases the standard errors (Lehrer, 2008). enced wanted sexual debut in later adolescence
Given the nonrandom nature of this missing have an increased risk of marital dissolution, but
data, I believe this interpretation is incorrect; that this is the result of engaging in premarital sex-
is, the imputations may produce biased results. ual behaviors associated with divorce. Increased
As such, studies using these data need to account rates of divorce are linked to first intercourse
for whether the imputed data affected their experiences that are wanted but occur before
results (see Teachman, 2008). My strategy was to the age of 16 and those experiences that are
reanalyze the results only with the nonimputed not completely wanted throughout adolescence.
data, and I found that excluding the imputed To the extent that women engage in adoles-
observations had no impact on my core findings. cent sexual debut, premarital sex, and premarital
484 Journal of Marriage and Family

fertility, information about whether these deci- DeMaris, A., & Rao, V. K. (1992). Premarital cohabi-
sions increase the risk of marital dissolution may tation and subsequent marital stability in the United
facilitate informed choices. In contrast, adoles- States: A reassessment. Journal of Marriage and
cent sexual debut that is not completely wanted the Family, 54, 178 – 190.
is both directly and indirectly linked to mari- Finkelhor, D. (1988). The trauma of child sexual
abuse: Two models. In G. E. Wyatt & G. Johnson
tal dissolution. Because these decisions are not Powell (Eds.), Lasting effects of child sexual abuse
completely voluntary, a choice framework does (pp. 61 – 82). Newbury Park, CA: Sage.
not apply, which highlights the need for greater Furstenberg, F. F. (2003). Teenage childbearing as a
attention to the consequences of unwanted sex- public issue and a private concern. Annual Review
ual experiences among girls. of Sociology, 29, 23 – 39.
Gaughan, M. (2002). The substitution hypothesis: The
impact of premarital liaisons and human capital on
NOTE marital timing. Journal of Marriage and Family,
I thank Jennifer Glass for her comments and suggestions. 64, 407 – 419.
Hamilton, B. H., & Nickerson, J. A. (2003). Cor-
recting for endogeneity in strategic management
REFERENCES research. Strategic Organization, 1, 51 – 78.
Abma, J. C., Martinez, G. M., Mosher, W. D., & Heaton, T. B. (1993). Feedback: Comment on ‘‘Pre-
Dawson, B.S. (2004). Teenagers in the United marital sex and the risk of divorce.’’ Journal of
States: Sexual activity, contraceptive use, and Marriage and the Family, 55, 240 – 241.
childbearing, 2002 (Vital Health Statistics, Series Heaton, T. B. (2002). Factors contributing to increas-
23, No. 24). Hyattsville, MD: National Center for ing marital stability in the United States. Journal
Health Statistics. of Family Issues, 23, 392 – 409.
Allison, P. D. (1982). Discrete-time methods for Kahn, J. R., & London, K. A. (1991). Premarital sex
the analysis of event histories. Sociological and the risk of divorce. Journal of Marriage and
Methodology, 13, 61 – 89. the Family, 53, 845 – 855.
Becker, G. S., Landes, E. M., & Michael, R. T. Landale, N. S., & Forste, R. (1991). Patterns of entry
(1977). An economic analysis of marital instability. into cohabitation and marriage among mainland
Journal of Political Economy, 85, 1141 – 1187. Puerto Rican women. Demography, 28, 587 – 607.
Bennett, W. G., Blanc, A. K., & Bloom, D. E. (1988). Laumann, E. O., Gagnon, J. H., Michael, R. T., &
Commitment and the modern union: Assessing Michaels, S. (1994). The social organization of
the link between premarital cohabitation and sub- sexuality: Sexual practices in the United States.
sequent marital stability. American Sociological Chicago: University of Chicago Press.
Review, 53, 127 – 138. Lehrer, E. L. (2008). Age at marriage and marital
Billy, J. O. G., Landale, N. S., & McLaughlin, S. D. instability: Revisiting the Becker-Landes-Michael
(1986). The effect of marital status at first birth hypothesis. Journal of Population Economics, 21,
on marital dissolution among adolescent mothers. 463 – 484.
Demography, 23, 329 – 350. Lehrer, E. L., & Chiswick, C. U. (1993). Religion as
Bratter, J. L., & King, R. B. (2008). ‘‘But will a determinant of marital instability. Demography,
it last?’’: Marital instability among interracial 30, 385 – 494.
and same-race couples. Family Relations, 57, Lepkowski, J., Mosher, W., Davis, K., Groves, R. M.,
160 – 171. van Hoewyk, J., & Willem, J. (2006). National
Browning, C. R., & Laumann, E. O. (1997). Sexual Survey of Family Growth, Cycle 6: Sample design,
contact between children and adults: A life course weighting, imputation, and variance estimation
perspective. American Sociological Review, 62, (Vital and Health Statistics, Series 2, No. 142)
540 – 560. Hyattsville, MD: National Center for Health
Buhi, E. R., & Goodson, P. (2007). Predictors of Statistics.
adolescent sexual behavior and intention: A theory- Lillard, L. A., Brien, M. J., & Waite, L. J. (1995).
guided systematic review. Journal of Adolescent Premarital cohabitation and subsequent marital
Health, 40, 4 – 21. dissolution: A matter of self-selection? Demog-
Buhi, E. R., Goodson, P., & Neilands, T. B. (2008). raphy, 32, 437 – 457.
Out of sight, not out of mind: Strategies for Marini, M. M. (1985). Determinants of the timing
handling missing data. American Journal of Health of adult role entry. Social Science Research, 14,
Behavior, 32, 83 – 92. 309 – 350.
Bumpass, L. L., Martin, T. C., & Sweet, J. A. (1991). Miller, B. C., & Heaton, T. B. (1991). Age at first
The impact of family background and early marital sexual intercourse and the timing of marriage and
factors on marital disruption. Journal of Family childbirth. Journal of Marriage and the Family,
Issues, 12, 22 – 42. 53, 719 – 732.
Adolescent Sex and Divorce 485

Paik, A. (2010). ‘‘Hookups,” dating, and relation- dissolution among women. Journal of Marriage
ship quality: Does the type of sexual involve- and Family, 65, 444 – 455.
ment matter? Social Science Research, 39, Teachman, J. (2008). Complex life course patterns
739 – 753. and the risk of divorce in second marriages. Journal
Raley, R. K., Crissey, S., & Muller, C. (2007). Of sex of Marriage and Family, 70, 294 – 305.
and romance: Late adolescent relationships and Teachman, J. D., & Polonko, K. A. (1990). Cohabi-
young adult union formation. Journal of Marriage tation and marital stability in the United States.
and Family, 69, 1210 – 1226. Social Forces, 69, 207 – 220.
Seidman, S. N., Mosher, W. D., & Aral, S. O. (1992). Terza, J. V., Basu, A., & Rathouz, P. J. (2008). Two-
Women with multiple sexual partners: United stage residual inclusion estimation: Addressing
States, 1988. American Journal of Public Health, endogeneity in health econometric modeling.
82, 1388 – 1394. Journal of Health Economics, 27, 531 – 543.
South, S. J., & Spitze, G. (1986). Divorce deter- Thomson, E., & Colella, U. (1992). Cohabitation and
minants. American Sociological Review, 51, marital stability: Quality or commitment? Journal
583 – 590. of Marriage and the Family, 54, 259 – 267.
Teachman, J. (1983). Early marriage, premarital Tzeng, M. S. (1992). The effects of socioeconomic
fertility, and marital dissolution. Journal of Family heterogamy and changes on marital dissolution
Issues, 4, 105 – 126. for first marriages. Journal of Marriage and the
Teachman, J. (2002). Stability across cohorts in Family, 54, 609 – 619.
divorce risk factors. Demography, 39, 331 – 351. Udry, J. R. (1979). Age at menarche, at first
Teachman, J. (2003). Premarital sex, premarital intercourse, and at first pregnancy. Journal of
cohabitation, and the risk of subsequent marital Biosocial Science, 11, 433 – 441.

You might also like