Sie sind auf Seite 1von 12

Premarital Sex and the Risk of Divorce

Author(s): Joan R. Kahn and Kathryn A. London


Source: Journal of Marriage and Family, Vol. 53, No. 4 (Nov., 1991), pp. 845-855
Published by: National Council on Family Relations
Stable URL: http://www.jstor.org/stable/352992
Accessed: 13-10-2016 15:38 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

National Council on Family Relations, Wiley are collaborating with JSTOR to digitize, preserve and
extend access to Journal of Marriage and Family

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

JOAN R. KAHN University of Maryland at College Park


KATHRYN A. LONDON National Center for Health Statistics*

Premarital Sex and the


Risk of Divorce

This article examines the relationship between


ferences (e.g., the willingness to break traditional
premarital sexual activity and the long-term risk
norms) rather than to a direct causal effect.
of divorce among U.S. women married between
1965 and 1985. Simple cross-tabulations from the
During the postwar period, divorce rates have
1988 National Survey of Family Growth indicate
steadily increased to the point that one out of two
that women who were sexually active prior to
recent marriages is expected to end in divorce
marriage faced a considerably higher risk of mari(Weed, 1980). These high rates contribute to the
tal disruption than women were were virgin
rising
number of female-headed families, many of
brides. A bivariate probit model is employed to
whom live below the poverty level (Duncan et al.,
examine three possible explanations for this posi1984). Moreover, the personal and emotional
tive relationship: (a) a direct causal effect, (b) an
costs of divorce, for both parents and children,
indirect effect through intervening "high risk"
are also very high (Garfinkel and McLanahan,
behaviors (such as having a premarital birth or
1986; McLanahan and Bumpass, 1988). Because
marrying at a young age), and (c) a selectivity efof these negative consequences, it is important to
fect representing prior differences between virgins
understand the factors that place individuals (and
and nonvirgins (such as family background or atcouples) at higher risk of marital disruption.
titudes and values). After a variety of observable
characteristics are controlled, nonvirgins still face Much of the recent sociodemographic research
on divorce has focused on the impact of a variety
a much higher risk of divorce than virgins. Howof characteristics and experiences of individuals
ever, when the analysis controls for unobserved
prior
to marriage. The ages of partners at marricharacteristics affecting both the likelihood of
age, the similarity of their cultural and socioecohaving premarital sex and the likelihood of dinomic backgrounds, whether either had experivorce, the differential is no longer statistically sigenced their own parents' divorce, whether the
nificant. These results suggest that the positive rewas preceded by a pregnancy or birth, or
lationship between premarital sex and the risk marriage
of
whether the couple lived together prior to marridivorce can be attributed to prior unobserved difage have all been found to be related to the risk of
Center on Population, Gender, and Social Inequality, Department of Sociology, University of Maryland, College Park, MD
20742-1315.

*6525 Belcrest Road, Room 840, Hyattsville, MD 20782.

divorce (Balakrishnan, Rao, Lapierre-Adamcyk,


and Krotki, 1987; Bennett, Blanc, and Bloom,

1988; Bennett, Goldstein, and Abzug, 1989;


Moore and Waite, 1981; Morgan and Rindfuss,
1985; Teachman, 1983).

Journal of Marriage and the Family 53 (November 1991): 845-855 845

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

846 Journal of Marriage and the Family

TABLE 1. PROPORTION OF EVER-MARRIED WOMEN AGED


A related premarital factor
that has been virtu-

15-44 WHO WERE VIRGIN BRIDES, BY MARRIAGE

ally ignored by recent analysts


was
COHORT ANDbut
RACE, 1988
NSFG the concern of researchers earlier in this century is
Marriage White Black
whether the partners were sexually active prior to
Cohort %o Virgin n %o Virgin n
marriage (i.e., whether they were virgins) (Burgess
1965-69 .43 658 .12 196
and Wallin, 1953; Davis, 1929; Lewis and Span1970-74 .30 787 .10 286
1975-79 .22 800 .05 264
ier, 1979; Terman, 1938). Premarital sex was
1980-83 .14 501 .04 179
thought to have a harmful effect because it would
Total
.28 2,746 .08 925
make marriage "less special," and therefore, couNote: Sample limited to wo
ples would develop less respect for marriage as an
institution (Whyte, 1990). Support for this hy- ried at least five years prior
pothesis was found in several studies conducted
between the 1920s and 1950s that found that hus-

tive impact of premarital


bands and wives with no premarital sexual experistability of marriages. F
ence scored higher on scales of marital happiness

and satisfaction (i.e., the studies' indicators of prior to marriage could


marital "success") than did couples in which atgreater knowledge about t
least one spouse had had premarital relations allowing incompatible cou
lationships. Through this
(Burgess and Wallin, 1953).
In this study, we update and expand this pastprocess, nonvirgins may
marriages than virgins. Si
research by examining the relationship between

virginity status at marriage and the risk of divorcepotentially

greater experie

among a national sample of recent marriage co-ships with other partners


virgins may be able to ma
horts. Since the early 1960s, there has been a
tal choices, which could
dramatic increase in premarital sexual activity,
quality
and stability. At
especially among teenagers (Hofferth, Kahn, and
if physical intimacy is m
Baldwin, 1987; Pratt and Eglash, 1990). Whereas
compatibility prior to ma
only 40% of the 1938-1940 female birth cohort
couples with premarital s
had had premarital intercourse prior to age 20,
average make worse mar
this number rose to almost 70%0 for women born
experience with other par
between 1959 and 1961 (Hofferth, Kahn, and
Baldwin, 1987). Given this trend, in combination virgins with higher expe
with the rising age at marriage, it is not surprising

within marriage and thus

disappointment (Whyte, 1990). On the other

that the proportion of virgin brides had declined


during the past several decades: more than 40% of hand, it may also provide them with more realistic

white women marrying between 1960 and 1965


were virgins as compared with less than 15% of
those marrying between 1980 and 1985 (see Table
1). A similar trend is evident for black brides,
though throughout the period, a much smaller
proportion of black brides were virgins. This rise
in premarital sexual activity represents a substantial change in courtship behavior among American couples, certainly when compared with earlier
decades (Whyte, 1990). Not only do they achieve
greater physical intimacy prior to marriage, but
recent couples are also more likely to enter marriage with premarital sexual experience with partners other than their spouse.
What are the implications of these trends for
subsequent marital quality and stability? Argu-

expectations than one might have without any

premarital experience.

Perhaps because virginity at marriage has become the exception rather than the rule among re-

cent marriage cohorts or because of lack of data


on premarital sex, few contemporary studies of
marital stability have considered its impact. In a
study of 450 women in the Detroit metropolitan
area, Whyte (1990) found a positive though insignificant relationship betwen premarital sex and
both marital problems and marital disruption. Recent national data, however, suggest a significant
positive relationship, at least for whites (see Table
2). Among white women first married between
1965 and 1985, virgin brides were less likely to
have dissolved their marriages through separation
ments could be made for both a positive and nega- or divorce than women who had not been virgins

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

Premarital Sex and Risk of Divorce 847

at marriage (25% vs. 35%).


tivelyAlthough
related to there
the age
is at
no
overall difference between
black
and
non(Udry
and virgins
Cliquet,
1982).
N
virgins in the percentage
of that
firstcouples
marriages
tha
shown
marryi
had been dissolved (46%
vs. higher
47%), we
do of
find
a
much
risks
sepa
pattern similar to the those
white marrying
pattern for
more
re
later.
Thi
cent black marriage cohorts.'
The
fact
that
this re
tributed
to
lower
levels
of
lationship holds for alllack
white
the more
of and
experience
inrecen
the

black marriage cohorts


during
this period level
indilower
socioeconomic
cates that the overall differential
(Moore and Waite,
between
1981;
virgi
M
and nonvirgin brides in
marital
stability
does not
1985;
Teachman,
1983).
Th
simply reflect increasing
trends over
time
in both
of-wedlock
child
at the
tim
nonvirginity and divorce.
But does
it reflect a di
evidence
of premarital
se
rect causal link or is there
some other
explanabeen shown
to place
a cou
tion? In this article, we
examine
the relationshi
divorce,
possibly
because t
in greater detail and consider
the relevance
of sev
fewer advantages
to the
tr
eral competing explanations.
In particular,
we
because her
husband may
n
consider the possibility
(Morgan
that the
and positive
Rindfuss,
rela19
tionship is due to the impact
of other
"interven
The positive
relationship
ing" behaviors that are
positively
related
bot
sex
and divorce
mayto
also
r
premarital sex and the
risk of
divorce.
We also
tivity.
That
is, women
wh
consider the possibility
that
virgins
from
sex
may
differdiffer
from virg
nonvirgins in significant
ways
that influence
influence mariways
that
mari
tal stability (i.e., a "selectivity"
argument).
In the
1990). One
way is through
sections that follow, we ues.
first
describe
these
It briefly
is likely
that remain
explanations; we then develop
a model
that tradit
allows
riage implies
a more
us to distinguish among
the explanations;
fitution
of marriageand,
(Berna
nally, we present our empirical
findings.
would not
consider divorc
Rather than reflecting
a direct
causal
relationcept
as a last
resort.
Wome
ship, the positive relationship
between
marriage
may be premarital
more com
sex and the risk of divorce
marriages
may be
work;
due to
they
the fact
ma
that nonvirgins are more
strained
likely by
to the
engage
negative
in other
op
behaviors that themselves are associated with a

and friends toward divorce.

higher risk of divorce (e.g., early marriage, having Conversely, women who engaged in premarital
a premarital birth). One of the strongest predicsex may have done so because they valued personal fulfillment and satisfaction more than traditors of marital disruption is the age at which the
couple marries, which, at least for women, is positional expectations of chastity. If they felt less
constrained by traditional norms prior to marriage, they may also feel less constrained by norms
about the permanence of marriage and therefore
TABLE 2. PROPORTION OF EVER-MARRIED WOMEN AGED
15-44 WHO WERE SEPARATED OR DIVORCED FROM FIRST

HUSBAND, BY VIRGINITY STATUS AT MARRIAGE,


MARRIAGE COHORT, AND RACE, 1988 NSFG
White

Black

Marriage Non- Non-

Cohort Virgin virgin t Test Virgin virgin t Test


1965-69
1970-74
1975-79
1980-83

.30
.30
.21
.14

.50
.46
.34
.24

**
**
**
**

.57
.61
.29
.13

.58
.53
.48
.44

be more willing to dissolve an unhappy marriage.3

Indeed, Lye and Biblarz (1990) recently found

that couples who hold nontraditional attitudes are

less satisfied with their marriages than couples


who hold more traditional attitudes. Similarly,
Thomson and Colella (1991) found that couples
who were less committed to the institution of mar-

riage tended to perceive a higher likelihood that


their marriage would end in divorce, in compariTotal .26 .38 .48 .51
son with other couples. This general argument
n 763 1,983 73 852
(that some individuals may be predisposed to reNote: Sample limited to women who first got marject traditional
norms
throughout their lives) is
ried at least five years prior
to the
interview.
+p < .10. **p < .01.
consistent with what many social commentators

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

848 Journal of Marriage and the Family

FIGURE 1. FULL MODEL PREDICTING THE

have identified as a general ideological shift


toward individual happiness and personal fulfillment and away from self-denial and self-sacrifice
(Marini, 1990; Swidler, 1980; Veroff, Douvan,
and Kulka, 1981; Yankelovich, 1981).
Part of the impact of changing values and ideologies may be conditioned by the type of family
in which a person grew up. Research has shown
that women from intact and religiously conservative households are more likely to postpone having sex until a later age (Hofferth et al., 1987;
Thornton and Camburn, 1986) and also to face
lower risks of divorce (London, Kahn, and Pratt,
1988; McLanahan and Bumpass, 1988; Teachman, 1983; Thornton, 1985). The typical explana-

PROBABILITY OF DIVORCE

Note:

See

text

for

explan

tion is that these kinds of households are more

likely to reinforce traditional family-oriented valthought to intervene between virginity and

ues and exert greater normative constraints,


divorce (e.g., premarital birth, early marriage),
whereas nonintact and less religious households
and Z represents a vector of variables, such as atmay leave crucial decisions up to individual
titudes, values, and family background, thought

choice.

to influence the likelihood of remaining a virgin


In summary, we have discussed three generaluntil marriage as well as the likelihood of divorce.
mechanisms through which premarital sex may beFinding a strong and significant relationship berelated to the risk of divorce. The first can be
tween V and D even after controlling for all other
thought of as a direct effect: this could be duerelevant variables would be support for a direct
either to the impact of premarital intimacy on
effect of virginity status on divorce. Finding that
subsequent marital quality or to the impact of
the inclusion of the X or Z variables explains (i.e.,
having had other premarital sexual experiences on
reduces) the effect of virginity status on divorce
the level of satisfaction with one's spouse. A secwould be support for the indirect or selectivity efond explanation suggests an indirect effect wherefects described above.

by the association between premarital sex and

The model in Figure 1 can be written as the foldivorce is explained by the intervening impact of
lowing recursive system of equations:

several "high risk" behaviors such as premarital

pregnancy and early marriage. The third explana- D = ao + a1X + a2V + a3Z + el (1)
tion, the selectivity effect, looks to preexisting at- V = bo + bZ + e2 (2)
If all of the relevant Z variables were observed
titudinal differences between virgin and nonvirgin
brides to explain their different risks of divorce.and adequately measured, then the impact of virThe ideal test of these explanations would requireginity status on divorce would simply be a2 .4
However, because many of the hypothesized Z
a costly prospective study that tracked couples
variables are unobserved (e.g., attitudes and prefrom early adolescence throughout their marriages and gathered both demographic and attitudi-dispositions thought to affect both V and D), then
nal information at each stage of the life course.estimates of a2 from Equation 1 will be biased and
inconsistent. In effect, what we have is a problem
Yet, in the absence of such a unique data source,
of omitted-variables bias because we lack meait is still possible to explore the relevance of these
sures of attitudes regarding traditional marriage
explanations with existing life history data.

behaviors. We argue that many of the same un-

THEORETICAL MODEL

measured variables are contained in both e1 and

e2, thus implying that el and e2 (and by extension,

The explanations described above are expressed


el and V) are correlated. Thus, by violating the asgraphically in Figure 1, in which D represents the
sumption that explanatory variables be indepenrisk of divorce, V represents virginity status at
dent and uncorrelated with the disturbances,
marriage, X represents a vector of variables Equation 1 will produce biased and inconsistent

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

Premarital Sex and Risk of Divorce 849

years (DIV1-5);
then,
among marriages
survivin
estimates of a2 (Kmenta,
1971;
Maddala,
1988).
Although we lack direct
of the
these
imat leastmeasures
five years, we model
probability
of di
portant attitudes and ruption
predispositions,
weofare
abl
in the second five years
marriage

to control for their effects


by
modeling
the
(DIV5-10).5
In order
to ensure that
all corre
responlation between the error
dents faced
terms
the same
ofpotential
bothduration
equation
of risk

models for DIV1-5


are restricted
to respondent
(el and e2). We use a bivariate
probit
model
that
by jointly estimating who
Equations
1 least
and
produce
got married at
five2,
years
prior to th
unbiased parameter estimates
areat net
th
interview date.that
Respondents
risk ofof
separatin
correlation between the disturbances and thereor divorcing in the second five years of marriage
fore are corrected for the effects of the omitted
(DIV5-10) are defined as those who got married
variables (Calhoun, 1989; Greene, 1986, 1990;
at least ten years prior to the interview and re
Maddala, 1983, 1988).
mained married for at least five years.
Information on premarital sexual activity is
based
on questions about the age and date at
DATA
which respondents first had sexual intercourse.
The data for this analysis come from the 1988 Na- These two questions, in combination with the date
tional Survey of Family Growth (NSFG), a
of first marriage, allow us to determine whether a
nationally representative sample of women be-respondent was a virgin at marriage. Virgin brides
tween the ages of 15 and 44. We limit our samplewere identified as respondents who reported havto white women whose first marriages took place ing had first intercourse during (or after) their
between 1965 and 1985. Nonwhites are excluded
month of marriage.6
because the small number of nonwhite virgins pre-Several measures of behaviors intervening becludes a detailed analysis. The NSFG is one of tween
the virginity status and marital stability are inbest available data sources for the present analysis
cluded in our model predicting the risk of divorce.
because it includes detailed information on marSince the impact of a premarital pregnancy on
riage and divorce as well as the timing of first sex.
divorce has been shown to depend on whether the
The NSFG also includes a number of other
birth arrived before or after marriage (Morgan
variables that are related to marital instability
and Rindfuss, 1985; Teachman, 1983; Waite and
(e.g., education, religion, and husband's characSpitze, 1981), we include a three-category classifiteristics). Although the survey lacks attitudinal
in-no premarital pregnancy (ending in live
cation: (1)
formation related to premarital sex and birth),
divorce,
(2) premarital birth, and (3) legitimated
our method of analysis will at least allow
us to
birth.
We also include controls for age at marriage
control for its systematic effect. A majorand
limitathe number of years of schooling completed
tion of the NSFG for the present analysis
is the
at the
time of marriage.7
lack of information on men, especially theirAlthough
prewe lack direct measures of the attimarital sexual experiences. However, given
tudesthat
and values hypothesized to predict both virvirginity at marriage has not been a normative
ginity exstatus and marital stability, (i.e., many of
pectation for men in our society, we would
not
exthe
Z variables
in Figure 1), we do have several
pect a very strong effect.
measures of family background that may be assoMeasuring the respondent's experience
with
ciated
with these attitudes and values. First, growmarital disruption is straightforward, as dates
ing upof
in a broken home is coded dichotomously
marriage, separation, and divorce are available.
to indicate whether the respondent did not live
We consider first marriages only and identify
dis- natural parents at age 14. Religious
with both
rupted marriages as those ending in eitherbackground
divorce is coded in three categories identifyor separation prior to the date of interview.
Ideing the
more conservative religious denominaally, we would want to model the risk oftions:
divorce
(1) Catholic, (2) fundamentalist Protestant,
across all durations of marriage (as one might
and (3) do
all others. Finally, on the basis of the rein a continuous-time hazard model). However,
sults of other studies, we also include controls for
because the bivariate probit model requires
marriage a
cohort; urban residence; and the age difdichotomous measure of divorce, we use
a dis-between spouses, coded in three categorference
crete-time hazard approach to model the
ies:prob(1) wife older than husband, (2) husband 0 to
ability of separation or divorce in the first
five
4 years
older than wife, and (3) husband more

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

850 Journal of Marriage and the Family

neither type
of premarital conception influences
than five years older than
wife.8
maritalanalysis,
stability, implying
that
the effect isesnot
In the final stage of the
we
jointly
timate the divorce model
proportional
(Equation
throughout the 1)
marital
with
life course.
the

model predicting virginity


As expected, status
both education
at
andmarriage
age at marriage
have model
negative and significant
effects some
on the risk of
of
(Equation 2). The latter
includes
divorce,
though the age-at-marriage
effect
weakthe same variables as the
former:
marriage
cohort
ens somewhat byaffiliation.
the second five yearsIn
of marrifamily stability, and religious
addition, we include mother's education (years of
age.
schooling completed), as well as two measures of
The relationship between virginity status and
the family context when the respondent was grow- the risk of divorce remains strong even after difing up: strictness of rules in the parental home ferences in family background characteristics are
(coded 1 if the rules were strict or very strict, and 0 controlled (Models 3 and 6). Surprisingly, once we
otherwise), and frequency of church attendance control for virginity status at marriage, Catholics
when age 14 (coded 1 if attended church once a and fundamentalist Protestants are no different
week or more, and 0 otherwise).9
from any other respondents in the probability of
divorce. However, our other measure of family
background, coming from a nonintact family,
RESULTS
raises the risk of divorce, at least within the first
We present our results in several stages in order tofive years of marriage. By the second five years,
provide a clear test of each of the explanationsfamily background has no effect.

described above. We start with a series of nested

Thus far, we have found that virgins face a sig-

models of divorce, derived from Equation 1: we nificantly lower risk of divorce than nonvirgins,
first present the simple bivariate model and then even after the effects of various intervening and
add variables corresponding to the observed inter-background variables are controlled. On the survening behaviors and family background charac-face, these findings tend to support the hypothesis
teristics. Finally, we control for the unobserved of a direct relationship between virginity status
variables affecting both virginity status and theand the risk of divorce. However, the results
risk of divorce by jointly estimating Equations 1based on Equation 1 do not take into account the
and 2 using a bivariate probit procedure.'0
potentially confounding effects of differences beTable 3 presents probit results showing the im- tween virgins and nonvirgins in unmeasured varipact of virginity status on the risk of divorce in theables that also affect the risk of divorce (i.e., the
first five years of marriage and the second fiveselectivity argument described earlier). For this
years of marriage net of the effects other explana-reason, we now turn to the bivariate probit results

tory variables. The results of Model 1 are consis-in Table 4.


tent with Table 2, which shows a strong negative The bivariate probit estimation procedure
relationship between being a virgin and the risk ofjointly estimates two probit equations, taking into
divorce, with year of marriage controlled. Look-account the correlation between their disturing across Table 3, we see that this effect cannotbances. It is a full-information maximum-likelibe explained by the introduction of other vari- hood procedure described in greater detail in
ables into the model. Model 2 shows that alGreene (1986, 1990). Table 4 presents the results
though the intervening variables do indeed affect
of the joint estimation of Equations 1 and 2, for
the risk of divorce, they do not reduce the effect
both the first and second five years of marriage.
of being a virgin. It is interesting that in (Note
this that these jointly estimated results correanalysis, unlike previous studies that found spond
that to the separately estimated results prehaving a premarital birth raised the risk of divorce
sented in Models 3 and 6 in Table 3).
while having a bridal pregnancy did not, we find
The key result in Table 4 is that, for both time
that once we control for virginity status at marriintervals, the impact of virginity status on the risk
age, having a premarital birth no longer signifiof divorce is no longer significant. In other words,
cantly affects marital stability, and having a bridal
once the unobserved factors affecting both the
pregnancy serves to protect against divorce, which
likelihood of being a virgin and the likelihood of

perhaps reflects the stabilizing effect of young


marital disruption are taken into account (i.e., the
children. After the first five years of marriage,
common systematic component of the distur-

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

Premarital Sex and Risk of Divorce 851

TABLE 3. PROBIT RESULTS PREDICTING THE PROBABILITY OF SEPARATION OR DIVORCE WITHIN THE FIRST AND

SECOND FIVE YEARS OF MARRIAGE (STANDARD ERRORS IN PARENTHESES): WHITE WOMEN, 1988 NSFG

First Five Yearsa Second Five Yearsb


Variable

Virgin
No

at

marriage?
-

Yes -.567*** -.614*** -.601*** -.419*** -.410*** -.402***

(.072) (.077) (.078) (.090) (.097) (.098)

Intervening behaviors
Premaritally pregnant?

Noc

Yes, premarital birth .166 .139 -.168 -.194


(.121) (.122) (.228) (.228)
Yes, marital birth -.182* -.171* .057 .066
(.086) (.086) (.115) (.115)
Age at marriage -.080*** -.073*** -.052** -.045*
(.012) (.013) (.020) (.020)
Education at marriage -.043** -.040* -.044* -.049*

(.016) (.016) (.023) (.023)

Family background characteristics


Intact family?
No

.234***

.077

(.069)

(.106)

Yes

Religion

Catholic

-.065

-.160

(.065)
Fundamentalist

(.093)

.098

(.125)

.039

(.180)

All others

Note: All models control for the effects of marriage cohort, urban residence, and age difference between spouses.
aThe sample is limited to white women who first got married between 1965 and 1983 (n = 2,746).
bThe sample is limited to white women who first got married between 1965 and 1978 and remained married for at

least five years (n = 1,562).

CNo premarital pregnancy ending in a live birth.

*p < .05. **p < .01. ***p < .001.

bances in both equations), we find that being aat marriage are consistent with previous findings
virgin no longer has a significant direct effect on(see Whyte, 1990, and Hofferth et al., 1987), rethe risk of divorce. We should point out, how-gardless of whether they are estimated separately
ever, that rho, the correlation between the distur-or jointly (separate estimates not shown). Most

bance terms, while negative in sign, is small innotably, we find that women were more likely to
value. The negative sign obviously implies a negahave been virgins at marriage if they came from
tive relationship between the disturbances, whichintact families, went to church regularly as teen-

we contend is attributable to common unobserved

agers (especially, fundamentalist Protestants),


variables. When we control for them in the joint
and were subject to strict rules while growing up.
estimates, we come to quite different conclusionsInterestingly, we find that women with betterabout the effect of virginity status on the risk ofeducated mothers were more likely to be nonvir-

divorce, despite the fact that rho fails to reach stagins at marriage, perhaps because they got martistical significance. According to these findings, ried at a later age."
virgin brides are not significantly less likely than

nonvirgins to separate or divorce, at least not

DIscussION
within the first ten years of marriage. The effects
of all of the other predictors of divorce, however,This analysis started with the observation that
remain unchanged from the single-equation estiwomen who were virgin brides faced a lower risk
mates presented earlier.
of divorce than women who were not virgins at
Moreover, the determinants of virginity statusmarriage, at least among white marriage cohorts

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

852 Journal of Marriage and the Family

TABLE 4. BIVARIATE PROBIT RESULTS OF JOINTLY ESTIMATED MODELS OF VIRGINITY STATUS AT MARRIAGE AND THE

RISK OF SEPARATION OR DIVORCE: WHITE WOMEN, 1988 NSFG


First Five Yearsa Second Five Yearsb

Variable
Virgin

at

No

Virgin

Divorce

Virgin

Divorce

marriage?

Yes

-.511

-.274

(.514)
Mother's

Strict
No

Yes

education

Rules

-.031***

(.842)

-.026*

(.009)
(.011)
at
age
14?

.249**

.307**

(.082) (.105)
religious at age 14?

Very
No

Yes

.246***

.127

(.062)

(.079)

Intervening behaviors
Premaritally pregnant?
Noc

Yes,

premarital

birth

.140

-.192

(.120)
(.241)
Yes, marital birth -.171* .065
(.087) (.114)
Age at marriage -.073*** -.045*
(.012)
(.019)
Education at marriage -.040* -.048
(.017) (.025)

Family background
Intact family?
No

-.207**

.241**

(.070)

(.078)

Yes

Religion

Catholic

characteristics

-.187*

.034

-.068

(.059)
Fundamentalist

.336**

(.114)

-.168

(.074)

.087

(.145)

(.122)

.131

(.068)

.086

(.094)

(.106)

.360*

(.150)

.211

(.225)

All others

Rho

2,746

-.054

(p

<

.86)

1,562

-.079

(p

<

.88)

Note: All models control for the effects o


aThe sample is limited to white women w
bThe sample is limited to white women wh

least

CNo

five

years

premarital

*p < .05. **p < .01. ***p < .001.

(n

1,562).

pregnancy

ending

in

live

since the mid-1960s. In the attempt to understand On the basis of simple probit estimates of the

this relationship, we used data from the 1988probability of separation or divorce in the first
NSFG to test several competing explanations: (a)
five and second five years of marriage, we found
a direct causal effect reflecting the impact of prethat, when we controlled for a variety of intervenmarital intimacy on marital quality or satisfacing and background variables, virgins still faced a
significantly lower risk of divorce than nonvirtion, (b) an indirect effect through intervening behaviors such as early marriage or premarital preggins. This provided tentative support for a direct

link. However, bivariate probit estimates showed


nancy, and (c) a selectivity effect representing
that, once the unmeasured differences between
prior (observed and unobserved) differences between virgins and nonvirgins.
virgins and nonvirgins that also influence the risk

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

Premarital Sex and Risk of Divorce 853

of divorce were takenvorce


into
account,
direct
e
as acceptable
options for the
themselves.
It is
fect of virginity status
thewho
risk
of divorce
w
likelyon
that people
feel constrained
by tradino longer significant. tional
In expectations
otherearly
words,
the this
result
in life will maintain
suggest that the observed
relationship
vi
orientation
throughout their lives,between
predisposing

ginity status and thethem


risk
of marriage
divorce
reflects
to wait until
to begin sexual
activneither a direct nor ity
indirect
causal
relationsh
and to reject divorce
as an option
if they

but rather is a function


of
the
selectivity
of
should
become
unhappy
with their marriage.
Un- virgi

with respect to unobserved


variables
that
also pr
fortunately, we
are unable to test this
explanation
without
direct measures of attitudes about what
tect against the risk of
divorce.
What are these unobserved variables that we

constitutes appropriate roles and behavior, both


inside and outside of marriage. We hope that fuonly speculate as to their nature; however, we ture surveys will include more questions that can
would argue that they probably reflect a set of
tap long-standing attitudes reflecting a person's
orientation toward the traditional institution of
attitudes related to marriage. We believe that we
have reasonably good measures of the demo- marriage.
graphic variables and many of the background Our findings demonstrate the selective nature
characteristics related to marital stability. But
of marital behaviors and have important implicabecause our data come from a large demographic
tions for interpreting past research on marriage
survey, we lack adequate measures of attitudesand divorce as well as for conducting future
toward the institution of marriage. As we dis-analyses. For instance, one might interpret the
cussed earlier, these attitudes may affect marital
previously discussed finding in the 1950s that virsatisfaction, or they may affect the likelihood of
ginity at marriage was associated with greater
considering divorce as an option at any given level
marital happiness and satisfaction to mean that
of marital satisfaction (Lye and Biblarz, 1990). couples could assure themselves of happier and
successful marriages by refraining from preTraditional views have regarded marriage as more
a
multipurpose institution that provides an eco-marital sex. Our findings suggest, however, that
nomic base, an environment for raising children,women who did not already hold traditional attiand the only socially approved context for sexual
tudes about premarital sex and marriage would
relations (D'Emilio and Freedman, 1988; Furstenprobably not achieve them simply by abstaining
berg, 1990; Goode, 1963). Happiness and per- from sex until marriage. Rather, it is prior attitudes, and not sexual activity per se, that influsonal fulfillment were also benefits of many marence the risk of divorce.
riages, but they were not necessarily the primary
benefit. Over time, many of the traditional functions of marriage have become less important.
NOTES
Women have become less dependent on marriage
for economic support. Sexual activity and child- This research was supported in part by a Summer
bearing outside of marriage have become more ac- Research Award from the Office of Graduate Studcepted. In general, traditional social arrangements ies and Research at the University of Maryland at

have been unable to include in our model? We can

based on self-sacrifice have weakened as the satis-

College Park. Data analysis was supported by the

faction of individual needs has become the criter-

Computer Science Center, also at the University of

ion by which most things are measured (Veroff et

Growth was jointly planned and funded by the National Center for Health Statistics, the National In-

al., 1981). Society has placed increasing emphasis


on individual choice and less emphasis on selfdenial with respect to work and family obligations

as well as sexual relations (Yankelovich, 1981).


For many people, individual happiness and satisfaction have become the most important measures
of marital success.

We interpret the results of our analysis to mean


that women who continue to hold traditional atti-

tudes about marriage are less likely than other


women to consider both premarital sex and di-

Maryland. The 1988 National Survey of Family

stitute of Child Health and Human Development,


and the Office of Population Affairs, all of the
U.S. Department of Health and Human Services.
The authors are grateful for the valuable comments
and suggestions from Charlie Calhoun, David Guil-

key, Colin Loftin, Bill Mason, Phil Morgan, Don


Negri, Bill Pratt, Harriet Presser, Jay Teachman,
and several anonymous reviewers. However, the

opinions expressed are solely those of the authors.


An earlier version of this article was presented at
the 1989 annual meeting of the American Sociological Association, San Francisco, August 9-13.

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

854 Journal of Marriage and the Family

REFERENCES
1. It should be emphasized, however,
that the black
estimates are based on a very small number of virBalakrishnan, T. R., K. V. Rao, Evelyne Lapierregins (n = 78) and are therefore probably not very
Adamcyk, and Karol Krotki. 1987. "A hazard
reliable. Furthermore, given the extremely small
model analysis of marriage dissolution in Canada."
proportion of black virgins, we would not anticiDemography 24: 395-406.
pate a strong relationship among black women.
Bennett, Neil, Ann Blanc, and David Bloom. 1988.
2. It is interesting, however,
that
premarital
preg"Commitment
and the
modern union: Assessing
the
nancies that were legitimated
by
marriage
are
not
link between
premarital
cohabitation
and subseassociated with a higher risk
of marital
disruption,
quent marital
stability." American
Sociological Review 53: 127-138.
possibly because of the protective
impact of children in the household or because the mother is marBennett, Neil, Heidi Goldstein, and Rikki Abzug. 1989.
rying the child's biological father (Furstenberg,
"Marital choice and instability." Paper presented at
1976).
the 1989 annual meeting of the American Sociological Association, San Francisco.
3. A similar selectivity argument has been suggested,
Booth, Alan, and David Johnson. 1988. "Premari-

though not tested, by several scholars regarding the

positive relationship between premarital cohabitation and divorce (see Bennett et al., 1988; Booth
and Johnson, 1988; Thomson and Colella, 1991).

4. This, of course, assumes that the equations are


properly identified.

5. An estimated 64% of all divorcing couples in 1987


had been married for less than ten years at the time

of the divorce decree (National Center for Health


Statistics, 1990).

6. We realize that some women may have had premarital intercourse during their month of marriage,

but without an explicit question on whether they


were virgins at marriage, we do not know for sure.
Yet, having waited until the month of marriage to
first have sex, these women are probably very similar to virgin brides in their attitudes and values.

7. Although information about premarital cohabitation is available in Cycle IV of the NSFG, we cannot include it in our model because it is so highly
collinear with virginity status at marriage (i.e., all
cohabitors were nonvirgins at marriage).

8. Although the 1988 NSFG contains some questions


about the current husband's characteristics (e.g.,
his race, religion, and prior marital status), this in-

formation is not available for the first husbands of

respondents who have been married more than


once. Thus we cannot include this information in

the present analysis of first marriages.


9. The inclusion of these last three variables in the vir-

ginity status model but not the divorce model clearly ensures the unique identification of the virgin effect in the divorce equation.

tal cohabitation and marital success." Journal of

Family Issues 9: 255-272.


Bernard, Jessie. 1982. The Future of Marriage. New
Haven, CT: Yale University Press.
Burgess, Ernest W., and Paul Wallin. 1953. Engagement and Marriage. Chicago: Lippincott.

Calhoun, Charles A. 1986. BIVOPROB: MaximumLikelihood Program for Bivariate Ordered-Probit


Regression. Washington, DC: Urban Institute.

Calhoun, Charles A. 1989. "Estimating the distribution of desired family size and excess fertility."
Journal of Human Resources 24: 709-724.

Davis, Katherine B. 1929. Factors in the Sex Life of


Twenty-two Hundred Women. New York: Harper.
D'Emilio, John, and Estelle Freedman. 1988. Intimate Matters: A History of Sexuality in America.
New York: Harper and Row.
Duncan, Greg J., with Richard D. Coe, Mary E. Corcoran, Martha S. Hill, Saul D. Hoffman, and James
N. Morgan. 1984. Years of Poverty, Years of Plenty: The Changing Economic Fortunes of American
Workers and Families. Ann Arbor: Institute for Social Research, University of Michigan.
Furstenberg, Frank F., Jr. 1990. "Divorce and the
American Family." Annual Review of Sociology 16:
379 -403.

Garfinkel, Irwin, and Sara McLanahan. 1986. Single


Mothers and Their Children: A New American Di-

lemma. Washington, DC: Urban Institute.


Goode, William J. 1963. World Revolution and Family Patterns. New York: Free Press.
Greene, William H. 1986. LIMDEP User's Manual.
New York: Econometric Software.

Greene, William H. 1990. Econometric Analysis.


New York: MacMillan.

Hofferth, Sandra L., Joan R. Kahn, and Wendy Bald10. Our results are estimated by using the Bivariate
win. 1987. "Premarital sexual activity among U.S.
Probit routine in the LIMDEP statistical package
(see Greene, 1986). Very similar results are also obtained with the program BIVOPROB (see Calhoun,

1986).

Planning Perspectives 19(2): 46-53.

Kmenta, Jan. 1971. Elements of Econometrics. New


York: MacMillan.

11. This is difficult to assess because virginity status


and age at marriage are probably both jointly deter-

mined.

teenage women over the past three decades." Family

Lewis, Robert A., and Graham B. Spanier. 1979.


"Theorizing about the quality and stability of mar-

riage." In Wesley R. Burr, Reuben Hill, F. Ivan

Nye, and Ira L. Reiss (eds.), Contemporary Theories


about the Family (Vol. 1). New York: Free Press.

London, Kathryn A., Joan R. Kahn, and William F.

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

Premarital Sex and Risk of Divorce 855

Pratt. 1988. "Are daughters


divorced
parents
(eds.), Themesof
of Work
and Love in Adulthood.
more likely to divorce Cambridge,
as adults?"
Paper
presented
MA: Harvard University
Press.
at the 1988 annual meeting
of
the
Population
Teachman, Jay
D. 1983.
"Early
marriage, premaritalAssociation of America, New
Orleans.
fertility,
and marital dissolution." Journal of FamLye, Diane N., and Timothy
Biblarz. 1990. "The
ily Issues 4:J.
105-126.
effects of attitudes towards
family
life Factors
and
Terman, Lewis
M. 1938. Psychological
in gender
Mariroles on marital satisfaction."
Paper
presented at
tal Happiness. New
York: McGraw-Hill.
the annual meeting of the
Population
Association
of
Thomson,
Elizabeth, and Ugo Colella.
1991. "CohabiAmerica, Toronto.
tation and marital stability: Quality or commitMaddala, G. S. 1983. Limited-Dependent
and
ment?" National Survey of Families and
House- Qualit
tive Variables in Econometrics.
Cambridge:
Camholds Working Paper
No. 23, Center for Demogbridge University Press. raphy and Ecology, University of WisconsinMadison.
Maddala, G. S. 1988. Introduction to Econometrics.
New York: MacMillan.
Thornton, Arland. 1985. "Changing attitudes toward
Marini, Margaret M. 1990. "The rise of individualism
separation and divorce: Causes and consequences."
in advanced industrial societies." Paper presented at
American Journal of Sociology 90: 856-872.
the annual meeting of the Population Association
of
Thornton,
Arland, and Donald Camburn. 1986. "ReAmerica, Toronto.
ligious commitment and adolescent sexual behavior
McLanahan, Sara, and Larry Bumpass. 1988. "Inter- and attitudes." Unpublished paper of the Survey
generational consequences of family disruption."
Research Center, University of Michigan.
American Journal of Sociology 94: 130-152. Udry, J. Richard, and R. L. Cliquet. 1982. "A crossMoore, Kristin A., and Linda J. Waite. 1981. Marital
cultural examination of the relationship between
dissolution, early motherhood, and early marriage."ages at menarche, marriage, and first birth." DeSocial Forces 60: 20-40.
mography 19: 53-63.
Morgan, S. Philip, and Ronald R. Rindfuss. 1985.
Veroff, Joseph, Elizabeth Douvan, and Richard A.
Kulka. 1981. The Inner American: A Self-Portrait
"Marital disruption: Structural and temporal difrom 1957 to 1976. New York: Basic Books.
mensions." American Journal of Sociology 90:
1055-1077.

National Center for Health Statistics. 1990. Advance

Waite, Linda J., and Glenna Spitze. 1981. "Young

women's transition to marriage." Demography 18:


681-694.
Report of Final Divorce Statistics, 1987. Monthly
Vital Statistics Report, Vol. 38, No. 12, Suppl. 2. Weed, James A. 1980. National Estimates of Marriage
Dissolution and Survivorship. National Center for
Hyattsville, MN: Public Health Service.
Pratt, William F., and Shelly D. Eglash. 1990. "PreHealth Statistics, DHHS Pub. No. (PHS) 581-1403,
Series 3.
marital sexual behavior, multiple partners, and
marital experience." Paper presented at the annual Whyte, Martin K. 1990. Dating, Mating, and Marriage.
New York: Aldine.
meeting of the Population Association of America,
Toronto.
Yankelovich, Daniel. 1981. New Rules: Searching for
Swidler, Ann. 1980. "Love and adulthood in Ameri- Self-fulfillment in a World Turned Upside Down.
can culture." In N. J. Smelser, and E. H. EriksonNew York: Random House.

This content downloaded from 213.114.142.238 on Thu, 13 Oct 2016 15:38:41 UTC
All use subject to http://about.jstor.org/terms

Das könnte Ihnen auch gefallen