Vous êtes sur la page 1sur 14

Population Investigation Committee

The influence of offspring's sex and age at parents' divorce on the intergenerational
transmission of divorce, Norwegian first marriages 1980-2003
Author(s): Torkild Hovde Lyngstad and Henriette Engelhardt
Source: Population Studies, Vol. 63, No. 2 (JULY 2009), pp. 173-185
Published by: Taylor & Francis, Ltd. on behalf of the Population Investigation Committee
Stable URL: http://www.jstor.org/stable/40646334
Accessed: 02-11-2017 10:23 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/40646334?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

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

Population Investigation Committee, Taylor & Francis, Ltd. are collaborating with JSTOR
to digitize, preserve and extend access to Population Studies

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
RRoutledqe
RRoutledqe T-ytor^Group

The influence of offspring's sex and age at parents'


divorce on the intergenerational transmission of
divorce, Norwegian first marriages 1980-2003

Torkild Hovde Lyngstad1 and Henriette Engelhardt2


University of Oslo; 2University of Bamberg

Whether a couple remain married or divorce has repeatedly been shown to be of importance for the marital
stability of their children. This paper addresses the related question of whether the intergenerational
transmission of divorce is contingent on the age at which parents divorced and the sex of the spouse who
experienced the parents' divorce. Using a population-wide data-set on Norwegian first marriages followed
from 1980 to 2003, we find that the intergenerational transmission hypothesis holds also for Norway, that
this relationship is stronger for women than for men, and that there is a negative age gradient in the
transmission effect for women. The experience of multiple family transitions, such as a parent's remarriage
or a second divorce, does not affect couples' divorce risk.

Keywords: intergenerational transmission; divorce; parents' divorse; child-parent; Norway; hazard


models; register data

[Submitted March 2008; Final version accepted February 2009]

Introduction and Bianchi 2006), and it is conceivable that parents'


divorce affects the marriages of sons differently from
A consistent finding in divorce research is thatthose of daughters. Finally, we took into account any
whether an individual's parents remain married hassubsequent transitions in their parents' marital lives
an important impact on the stability of that indivi-that the spouses experienced before getting married
dual's marriage. This association has been demon- themselves, such as parents' remarriage or another
strated with different data sources for a large divorce. Some of the intergenerational transmission
number of countries (see, e.g., McLanahan and might operate through this causal pathway. Owing
Bumpass 1988; Kiernan and Cherlin 1999; Kiernan to data restrictions, few other researchers can offer
2000; Wolfinger 2005; Diekmann and Schmidheiny an equally detailed account of age dependencies in
2006; Dronkers and Hrknen 2008). the association between parents' and offspring's
The study presented in this paper adds to the divorce rates. In order to explore these issues, we
existing literature on the intergenerational transmis- used a rich and very large population-wide data-set
sion of divorce by addressing three issues. First, weon married couples taken from Norwegian adminis-
investigated whether transmission was contingent on trative registers.
the offspring's age at the time of their parents'
divorce. Although there are theoretical reasons
to expect an age-contingent transmission effect,A brief review of evidence and explanations of
the evidence for it is somewhat mixed (Kiernanthe intergenerational transmission of divorce
and Cherlin 1999; Furstenberg and Kiernan 2001;
Wolfinger 2005). Second, we took into account A correlation between parents' divorce and the
whether it was the husband or the wife who had divorce risk of their adult children is documented in

a large
experienced the parents' divorce. The importance of number of studies from various industrialized
the child's sex for family processes has been countries, such as Australia (D'Onofrio et al. 2007),
documented for a large number of outcomes (Raley France (Traag et al. 2000), Germany (Diekmann and

ISSN 0032-4728 print/ISSN 1477-4747 online/09/020173-13 2009 Population Investigation Committee


DOI: 10.1080/00324720902896044

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
174 Torkild Hovde Lyngstad and Henriette Engelhar dt

Engelhardt 1999; Engelhardt et al. 2002), the Nether- nerational transmission of divorce is driven by the
lands (Poortman and Kalmijn 2002), Norway (St0rk- lessons children learn about marital commitment
sen et al. 2007), the UK (Kiernan and Cherlin 1999), (Wolfinger 2005). Amato (1996) suggests that the
and the USA (Mueller and Pope 1977; Greenberg and intergenerational transmission of divorce operates
Nay 1982; Keith and Finlay 1988; Wolfinger 1999; Li mostly through spouses' interpersonal behaviour.
and Wu 2008). The strength of the transmission effect Furthermore, it seems that children of divorce
varies, but having experienced the dissolution of the develop behaviour that interferes with the mainte-
marriage of one's parents is in all cases associated nance of intimate relationships, and that this is an
with a higher risk of divorce for the offspring. In part, important factor behind their excess risk of divorce.
the intergenerational transmission of divorce is likely A later study finds that the intergenerational trans-
to be a result of selection. A recent study from mission of divorce is stronger for individuals whose
Australia, using a sophisticated children-of-twins de- parents had relatively low levels of pre-divorce
sign, confirms that there is some biological selection conflict, and concludes that children of divorce
in addition to selectivity on social factors, but that have higher risks of marital dissolution because
there must also be factors exerting a causal influence they have a lower general commitment to the
on the intergenerational transmission of divorce norm of lifelong marriage (Amato and DeBoer
(D'Onofrio et al. 2007). 2001).
A diverse range of causal mechanisms have been
Taken together, the evidence points towards a
suggested as contributors to the transmission pro-
causal effect of the parents' divorce on their off-
cess. One set of these refers to factors related to
spring's divorce risk that operates through one or
growing up with a single mother or in a stepfamily
more of a set of different mechanisms, of which the
with little or no contact with the child's own father.
most important is a lower commitment to marriage
Absence of the father is often offered as an
among children of divorce.
explanation of poor outcomes for children of divorce
(Sigle-Rushton and McLanahan 2002), but it does
not seem to be a very important explanation for the
Reasons for age and sex differences in
intergenerational transmission of divorce. Results
intergenerational divorce transmission
obtained with natural-experiment data from Ger-
many show a strong effect on own divorce risk for
There are several reasons to suppose that the
individuals who experienced their parents' divorce,
intergenerational transmission of divorce might
while no such relationship appears for individuals
depend on the age of the child when the divorce
who lost a parent during the Second World War
occurs. Father absence, for instance, might be more
(Diekmann and Engelhardt 1999).
important if the parents' divorce takes place when
Another prominent explanation is that the experi-
children are very young because they then will face a
ence of parents' divorce impairs the development of
trust in children and adolescents, and that longer
this period
has of little or no contact with their father.
consequences for how much commitment and Theretrust
might also be an age gradient in how parents'
divorce
in each other they can muster in their own romantic potentially affects the offspring's level of
relationships (Hayashi and Strickland 1998; trust in other individuals in general, and a romantic
Amato
partner
and Gilbreth 1999). A study by Franklin et al. (1990)in particular. Older children may, for exam-
ple,
examined the long-term impact of parents' be more conscious of the process leading
divorce
on perceived risk and trust between partners inthe
towards a dissolution of their parents' marriage
and take
romantic relationship. Their results suggest the process as an example of how to deal
that
adult children of divorce are less optimistic withabout
a troubled relationship. Adult children might be
their own future success in a marriage and report better able to deal with the experience of their
less trust in their future partner. However, parents' divorce: they are more mature, they have
it has
been shown that children of divorce do not show had more experience of observing other divorces in
lower levels of trust in intimates and others once the their social network, and they are in contact with a
quality of the parent-child relationship is controlledlarger number of individuals able to offer support
for, and thus that it is the relationships between anincluding their own romantic partners.
adult child and its divorced parents that are of In addition, an early divorce will potentially result
greater importance (King 2002). in more family transitions such as the entry of one or
A fair amount of the empirical evidence supports both parents into a second union, and possibly the
one notable explanation, namely, that the interge- experience of a second dissolution (Wolfinger 2005).

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
The intergenerational transmission of divorce 175

Some studies have shown that the experience of Amato and DeBoer (2001) investigated both age
multiple family transitions, such as a divorce, gen- and sex contrasts in the intergenerational transmis-
erally produces worse outcomes for children (Amato sion of divorce, and found no differences. They did,
and Booth 1991; Wu 1996; Wolfinger 2005; but see however, find a small age difference in the like-
Teachman 2002). If the intergenerational transmis- lihood of thinking about divorce. The likelihood of
sion of divorce has a causal component, and if it has having thoughts of divorce was higher if offspring
an additive effect when individuals experience multi- were younger at the time of their parents' break-up.
ple family transitions, an age gradient might emerge Parents' marital discord was associated with a higher
simply because children who experience a divorce likelihood of thinking about divorce for men but no
such association was found for women.
early in life have more exposure to such subsequent
A difference in effect between the sexes could be
transitions than those whose parents divorce when
they are older. a consequence of sex-specific socialization effects.
Several studies have investigated the possible Pope and Mueller (1976) claim that growing up with
dependence of intergenerational transmission on an opposite-sex parent means less exposure to same-
the age and sex of the child at the time of divorce. sex role models, and that the latter give the child
Two important and fairly recent studies in the UK fewer opportunities to learn appropriate roles as
used the British National Child Development Study, spouse and parent. Since most children live mainly
which has followed a cohort of British children with the mother following a divorce, this would
translate into a higher transmission effect for men
continuously since their birth in 1958. Kiernan and
than for women. This idea, often described as the
Cherlin (1999) studied the dissolution rate in part-
'same-sex hypothesis', was not supported in a broad
nerships of any type, and report that a part of the
investigation in which a wide range of outcomes
intergenerational transmission of divorce can be
from childhood, adolescence, and early adulthood
explained by other life-course choices such as the
were studied for individuals who grew up with one
timing of entry into the first co-residential union and
parent (Powell and Downey 1997).
the choice of union type. In addition, they controlled
There may be yet other mechanisms that con-
for the child's behaviour problems, school outcomes
tribute to the production of a sex difference in the
at ages 7 and 16, and class background. Even after
transmission effect. One possibility is that the effect
these controls, the positive effect of parents' divorce
is produced by a combination of two factors: (i) the
on the dissolution rate persisted. In the models in
effect the parents' divorce has on the child's
which control variables were included, separate
commitment to marriage, which currently is the
parameters were estimated for the effects of parents'
leading candidate as an explanation of the causal
divorce at ages 0-19 and 20-33. Although they found
component of intergenerational transmission, and
evidence for the intergenerational transmission ef-
(ii) the fact that when a marriage moves towards a
fect, no age gradient was found for women. For men,
divorce, it is usually the wife who takes the initiative
however, the effect seemed to disappear if the in the process towards a dissolution of the union
divorce had taken place after the child had turned (Kalmijn and Poortman 2006). From this reasoning,
20, indicating a negative age gradient if it had taken one would expect that the transmission effect should
place at earlier ages. be stronger for women, and in fact a fair amount of
Furstenberg and Kiernan (2001) later used the evidence suggests that they do indeed have a higher
same data-set to investigate whether parents should
risk of divorce than men who experienced parents'
delay divorce for the sake of their children. In order
divorce (Glenn and Kramer 1987; Amato 1996; Feng
to make use of a rich set of control variables
et al. 1999). On the other hand, other investigators
measured at age 7, they counted parents' divorces
report no sex difference (Wolfinger 2005; St0rksen
taking place only at that age or later. Controlling
et al.for a
2007).
range of pre-divorce characteristics, such as children's
To summarize, the empirical evidence is quite
malaise and anti-social behaviour, they report strong
mixed on the role the child's sex and age at parents'
evidence for an intergenerational transmission
divorceof
plays in how the intergenerational transmis-
divorce. Consistently with the Kiernan and sion
Cherlin
operates. There are some weak indications of an
(1999) study, the effects were somewhat stronger for
age gradient, and there are similar indications of a
sex difference. It is useful to note the following
men than for women. Despite all parameter estimates
indicating a negative age gradient in this effect, the of the earlier studies: they either used
limitations
broad age categories (at least when other variables
differences in hazard rates by age at parents' divorce
were not statistically significant. were controlled) or data where the individual and

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
176 Torkild Hovde Lyngstad and Henriette Engelhar dt

not the married couple was the unit of analysis. H4: Those who experienced additional family transi-
When there is homogamy in the family structure of tions after their parents divorced have higher
divorce risks than those who did not experience
the spouses (their family structures are correlated),
such transitions.
the latter problem will bias any estimated effect of
parents' divorce (Wolfinger 2005). An implication of hypothesis 4 is that some of the
intergenerational transmission effects operate
through these additional transitions. Our estimates
Hypotheses of the strength of the effect were expected to
weaken somewhat when experiences of additional
Our first hypothesis concerns the existence of the transitions were accounted for, if these transitions
intergenerational transmission of divorce in Norway. themselves increased divorce risk.

Earlier studies in other European countries and the


USA offer good reasons to believe that the trans-
Data and methods
mission effect operates also in Norway. Hence our
first hypothesis was:
Available data

HI: The divorce rate of couples is higher if one or both


of the spouses experienced parents' divorce. To test our hypotheses, we used a very large data-set
drawn from a range of Norwegian administrative
To turn to one of our main concerns: as noted
registers. The register system in Norway was set up
above, the evidence about the strength of a sex
in the early 1960s and is based on the principle that
difference in the transmission effect is mixed. From
each resident has his or her own unique personal
earlier studies, we expected that the intergenera-identification (ID) number. This ID number identi-
tional transmission of divorce would be stronger for
fies the individual in all administrative records kept
men than for women. Hence our second hypothesis by population registry offices, educational institu-
was:
tions, tax registers, and other government institu-
tions. The registers are population-wide and
H2: Couples where the husband has experienced
longitudinal.his
For more information on this type of
parents' divorce, all else equal, will have a higher
data-set, see R0ed and Raaum 2003.
divorce rate than couples where the wife has
From the population register, we created a data-set
experienced her parents' divorce.
on marriages contracted from 1980 onwards and
their
Another of our main concerns was the outcomes. Data on demographic variables,
possibility
including
of a transmission effect contingent on the the marital
age at history of both spouses' parents
andThe
disruption of the parents' marriage. otherearlier
types of variables, were subsequently
studies suggested we assume that there woulddata-set
linked to the be and provided precise and
accurate
an age gradient in the transmission effect longitudinal
and that it information on both spouses'
educational
would be at its strongest for children attainments and occupation by year,
who experi-
their annual earned incomes, and their fertility
enced their parents' divorce while they were rela-
histories. The final data-set we used contains the
tively young. Therefore, our third hypothesis was:
complete cohorts of first marriages contracted from
1980 to 2002. In total, we included 323,479 marriages
H3: The excess risk of marital dissolution for those
in the analysis. We excluded marriages where at least
who experienced parents' divorce declines with age
at parents' divorce. one spouse was an immigrant because this group
often displays patterns of demographic behaviour
A low age at parents' divorce means that themarkedly
that are child different from those of the host
will have a higher likelihood of experiencing
population.sub-
The marriages were followed until di-
sequent transitions in parents' marital lives, such
vorce, as a
emigration, spouse's death, or censoring
mother's remarriage and a possible second
occurred.divorce
Observations were censored at the end of
(Wolfinger 2005). We expected that the experience
the observation window in 2003. With these data we
of such additional transitions would increase cou- avoided a number of problems that usually plague
ples' divorce risk. Hence our fourth and final
studies using survey data. We had no attrition from
hypothesis was: the sample or any systematic misreporting, as the

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
The intergenerational transmission of divorce 111

measurements were gathered through administrative show that the choice of link function in the

procedures and not directly from the individuals. statistical model (complementary log-log or logit)
did not matter in any important way for our results.
We used the time of judicial divorce as the duration
Statistical approach variable because this represented an irreversible
step in the break-up process and is easier to measure
We employed discrete-time hazard regression mod- than separation.
els in our analysis. This method is frequently used in
studies of marital dissolution, and is both practical
and sufficiently precise for our purposes (Allison Measurement
1995). The time unit is a calendar year because
many of our variables are measured to the accuracy All the variables included in the models were

of 1 year. The observations of marriage periods are categorical variables. This eased interpretation of
split into slightly more than 3.75 million marriage- differences between groups. We could accommodate
years. A logistic regression model was then esti- a large number of categories with this large-N data-
mated on the data-set of marriage-years with a set. All time-varying independent variables were
dummy variable that indicated whether a divorce lagged by one calendar year to avoid problems
took place that year as the dependent variable. Tests related to anticipatory regressors. Table 1 shows

Table 1 Distribution of selected variables used for a study of the intergenerational transmission of divor

Variable Category Percentage

Divorced this year of marriage? Yes 1.5


No 98.5

Duration of marriage, years 0-4 39.1


5-7 18.3
8-10 14.4
11-14 14.4
15+ 13.8
Husband's age at marriage Up to 24 27.0
25-29 43.8
30-34 19.1
35+ 10.2
Wife's age at marriage Up to 24 49.9
25-29 34.3
30-34 10.1
35+ 5.7
Husband's annual income in 1980 NOK Less than 50k 13.6
50-100k 39.3
100-150k 33.3
150-200k 8.8
More than 200k 5.0
Wife's annual income in 1980 NOK Less than 50k 47.9
50-100k 43.1
100-150k 7.6
150-200k 1.0
More than 200k 0.4
Parity and age of youngest child Childless 31.1
One child, 0 years 14.5
One child, 1-6 years 4.9
One child, 7 + years 2.6
Two children, 0 years 13.6
Two children, 1-6 years 10.3
Two children, 7+ years 8.4
Three children, 0 years 6.4
Three children, 1-6 years 5.2
Three children, 7+ years 3.2

Source: Norwegian administrative register data.

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
178 Torkild Hovde Lyngstad and Henriette Engelhardt

the distribution of marriage-years over selected as not having experienced any disruption of his or
variables included in the analysis. her parents' union.
Both spouses' education levels were included as For a proportion of cases, we could not establish
two time-varying variables, each with three levels: with certainty whether the parents did divorce, or if
primary education, secondary education, and uni- they did, the age of the spouse at the time it
versity-level education. Another two dynamic cate- occurred. The reason for this is that the time series

gorical variables measured the annual income from for the marital status of the spouses' parents started
labour-market activity for each spouse. Both vari- in 1964. Some parents' marriages may have been
ables had five levels each. The incomes were dissolved by either death or divorce earlier than
1964 or the parents may not have been married at
inflation-adjusted with 1980 as the baseline year.
Age at marriage was included separately for all. Before 1964, these parents may also have
each
spouse and had four levels. Parity and remarried
age of and divorced again without this being
the couple's youngest child was grouped into recorded
a ten- in our data. Such indeterminate cases were
grouped
level variable for which childless couples made up in a separate category, and experiments
were conducted to check the robustness of our
the reference group, and there were nine combina-
conclusions
tions of parity and age categories. We supposed that with the inclusion of the indeterminate
all of these variables were likely to be correla-
cases. The results from these experiments suggested
that
ted with parents' divorce and the couple's the indeterminate cases did not introduce any
own
divorce risk. Contributions by trends in divorceimportant
rates bias.
(period effects) and the couple's year of marriage
(cohort effects) were captured by categorical vari-
ables. Limitations related to cohabitation
Our main variables were those that track the
family structure of the two spouses' familiesAof
complicating factor is that over the past few
decades the popularity of cohabitation in Norway,
origin to show whether and at what age the spouses
both as a precursor and as an alternative to
experienced their parents' break-up. For each of the
marriage, has increased dramatically. This data-set
two partners, we constructed a separate categorical
comprises married couples only, and has no informa-
variable. These two variables had the following main
tion on cohabiting couples. Parents' divorce is
categories: parents still married, parents' marriage
associated with the choice of (premarital) cohabita-
dissolved by death, and parents' marriage dissolved
tion, and this seems to explain some of the inter-
by divorce. The last category was then broken down
generational transmission of divorce (Kiernan and
into eleven 3-year age groups. The first category was
for those whose parents' divorce took place in the Cherlin 1999; Wolfinger 2005). Thus, in the data we
first 3 years of the individual's life, the next for those couples who had experienced parents' divorce
had
for whom it occurred in the next 3 years, and so on. but nevertheless chose to marry. These couples may
have valued marriage more highly or had stronger
The final category was open-ended and was for
barriers to union dissolution than those children of
those whose parents divorced 30 years or more after
the child's birth. divorcees who had so far chosen not to marry. This
To take account of the possibility that parents limitation clearly weakens the generalizability of our
might have remarried and then dissolved a second findings, but also implies that we can view the results
marriage, a set of dummy variables was included. as conservative tests of our hypotheses.
For both spouses, one dummy variable identified a Moreover, transitions occurring after parents'
mother's remarriage and another identified a divorce were recorded only if they involved a change
mother's second divorce. Two other variables iden- in marital status. Thus, any cohabitation in which the
tified the same transitions for the spouse's father. In spouses' parents were involved was not counted. For
total there were eight dummy variables picking up example, if a father of a spouse, after having
the effect of subsequent transitions, four for each divorced the mother, had entered an informal union
spouse. with a new partner, this would not have been
All variables measuring aspects of the maritalcaptured by our data. However, the parents included
histories of the spouses' parents were 'frozen' at thein this study were from cohorts in which cohabita-
time of the spouses' marriage. This means that for an tion had not been a common experience and
individual whose parents divorced after the year ofparents' cohabitation should therefore not have
his or her own marriage, that individual was treatedjeopardized the validity of our results.

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
The intergenerational transmission of divorce 179

Results smaller in magnitude. For husbands, the highest risk


is still found for those who experienced parents'
Two discrete-time hazard regression models were divorce at a relatively young age. However, the risk
estimated. In both models, we included a set of for other age groups is fairly similar, except that
independent variables to account for other back- some of the estimates are not significantly different
ground factors: spouses' education levels, their from each other. The effect of parents' divorce is
annual incomes from labour-market activity, parity stronger for wives than for husbands, as it was in the
and age of youngest child, their ages at the time of previous model. The effects for wives, measured in
their marriage, and period effects. The complete odds ratios, range from 2.0 for couples where the
regression results from both regression models are wife's parents divorced when she was 3-5 years old
shown in Table 2. to 1.43 for couples where the wife's parents divorced
In our first model, we included indicators of when she was aged 30 or older. Evidently, for wives
parents' divorce for both the husband and the wife there is a decline in the intergenerational transmis-
and let the effect of parents' divorce depend on the sion effect with an increasing age at parents' divorce.
age at which the divorce took place. From this Formal significance tests confirm this, but only for
model, we obtained parameter estimates for all the some of the differences between parameters.
age categories in which divorce could have occurred. In both models, couples in which either the
husband or the wife experienced the dissolution of
A likelihood ratio test from a preliminary model
the parents' union owing to the death of a parent
tells us that distinguishing between ages at parents'
have a very small, but significant, excess risk of
divorce significantly improves the model fit
divorce compared with those who grew up in intact
(p <0.0001). For both spouses, there seems to be a
families. This finding accords with findings from the
decline in the transmission effect as age at the time
USA (Teachman 2002). The other variables included
of the parents' divorce increases. The decline is also
in the model behave as we would expect from the
stronger for wives than for husbands. All but one of
general divorce literature. The husband's income is
the parameter estimates for the various age cate-
negatively related to divorce risk, while the wife's
gories are measurably larger than zero. The range of
high income is associated with a higher divorce risk.
effects, measured in odds ratios of divorce compared
A high age at marriage decreases the likelihood of
with the reference group of those with parents
marital dissolution. Couples with several children
whose marriages are still intact, is now 1.55-1.22
experience very low divorce rates, but this difference
for husbands and 2.0-1.53 for wives. Thus, there
diminishes as the youngest child grows older. Higher
seem to be important differences in the strength of
divorce risks are also found among couples with
the intergenerational transmission of divorce by the lower educational attainments.
age at which the wife experienced her parents'
divorce. Formal significance tests of differences
between parameter estimates (made by changing Discussion and conclusions
reference categories for the variables) confirm that
there is a significant decline in the excess risk with In these results we find support for the first of our
increasing age at parents' divorce. three hypotheses, that the divorce rate of couples is
Finally, in a second regression model we included higher if one or both of the spouses experienced the
variables that took account of any remarriage and divorce of their parents. This confirms that the
second divorce on the part of the spouses' parents. intergenerational transmission of divorce operates
Separate effects for both spouses' parents were in Norway (St0rksen et al. 2007), as it does in all
estimated for a father's remarriage and second other countries for which we have results. According
divorce and a mother's remarriage and second to the second hypothesis, because the transmission
divorce. In total, eight dummy variables were added operates in slightly different ways for men and
to the model to capture any effects of having women, we should expect to find that couples where
experienced these subsequent transitions for the the husband has experienced his parents' divorce
two spouses' parents. Somewhat surprisingly, the have a higher divorce rate than couples where the
inclusion of these variables does not affect their wife has had this experience. A theoretical reason
offspring's divorce risk. None of the eight parameter
for expecting a higher risk for men is found in the so-
estimates is measurably different from zero. called 'same-sex hypothesis', which predicts that
Compared with the first model, the parameterchildren who live with a single parent of the same
estimates for parents' first divorce are somewhat
sex will do better than children living with a single

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
180 Torkild Hovde Lyngstad and Henriette Engelhar dt

Table 2 Parameter estimates from discrete-time hazard regression models of divorce estimated on Norwegian first
marriages 1980-2003

Model 1 Model 2

Variable Category Beta SE p Beta SE p

Intercept -3.30 0.03 0.00 -3.29 0.03 < 0.0001


Duration of marriage 0 -15.77 23.85 0.51 -15.75 23.64 0.51
1 -5.25 0.13 <0.0001 -5.25 0.13 <0.0001
2 -1.64 0.03 <0.0001 -1.64 0.03 <0.0001
3 -0.53 0.02 < 0.0001 -0.53 0.02 < 0.0001
4 -0.12 0.02 <0.0001 -0.12 0.02 <0.0001
5-7 0.00 0.00
8-10 -0.04 0.01 0.01 -0.04 0.02 0.01
11-14 -0.19 0.02 <0.0001 -0.19 0.02 <0.0001
15+ -0.39 0.03 <0.0001 -0.39 0.03 <0.0001

Husband's educational High -0.14 0.01 <0.0001 -0.14 0.01 <0.0001


attainment
Medium 0.00 0.00
Low 0.25 0.01 < 0.0001 0.25 0.01 < 0.0001

Wife's educational attainment High -0.21 0.01 <0.0001 -0.21 0.01 <0.0001
Medium 0.00 0.00
Low 0.21 0.01 <0.0001 0.20 0.01 <0.0001

Husband's age at marriage -24 0.07 0.01 < 0.0001 0.07 0.01 < 0.0001
25-29 0.00 0.00
30-34 0.00 0.01 0.89 0.00 0.01 0.94
35+ -0.01 0.03 0.68 -0.01 0.03 0.72

Wife's age at marriage -24 0.27 0.01 <0.0001 0.27 0.01 <0.0001
25-29 0.00 0.00
30-34 -0.24 0.02 <0.0001 -0.24 0.02 <0.0001
35+ -0.56 0.04 <0.0001 -0.56 0.04 <0.0001
Husband's annual income 0-50k 0.41 0.01 < 0.0001 0.40 0.01 < 0.0001
in 1980 NOK
51-100k 0.07 0.01 <0.0001 0.07 0.01 <0.0001
101-150k 0.00 0.00
151-200k -0.01 0.02 0.69 -0.01 0.02 0.59
201k + 0.06 0.02 0.01 0.05 0.02 0.02
Wife's annual income 0-50k -0.45 0.02 <0.0001 -0.45 0.02 <0.0001
in 1980 NOK
51-100k -0.41 0.02 <0.0001 -0.41 0.02 <0.0001
101-150k 0.00 0.00
151-200k 0.09 0.04 0.03 0.09 0.04 0.03
201k + 0.14 0.06 0.02 0.14 0.06 0.03
Period 1980^84 -0.18 0.04 <0.0001 -0.18 0.04 <0.0001
1985^89 -0.15 0.02 <0.0001 -0.15 0.02 <0.0001
1990-94 0.00
1995-99 -0.12 0.02 <0.0001 -0.12 0.02 <0.0001
2000+ -0.05 0.02 0.02 -0.05 0.02 0.01

Parity and age of Childless 0.00 0.00


youngest child
One child/0-1 years old -0.95 0.02 < 0.0001 -0.95 0.02 < 0.0001
One child/2-6 years old -0.02 0.02 0.29 -0.01 0.02 0.37
One child/7 + years old 0.07 0.02 0.00 0.07 0.02 0.00
Two children/0-1 years old -1.09 0.02 <0.0001 -1.09 0.02 <0.0001
Two children/2-6 years old -0.42 0.02 <0.0001 -0.42 0.02 <0.0001
Two children/7 + years old -0.24 0.02 <0.0001 -0.24 0.02 <0.0001
Three + children/ -1.45 0.03 < 0.0001 -1.44 0.03 < 0.0001
0-1 years old
Three + children/ -0.59 0.02 <0.0001 -0.59 0.02 <0.0001
2-6 years old

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
The intergenerational transmission of divorce 181

Table 2 {Continued)

Model 1 Model 2

Variable Category Beta SE p Beta SE p

Three + children/ -0.36 0.03 < 0.0001 -0.35 0.03 < 0.0001
7+ years old
Marriage cohort 1980-86 0.00 0.00
1987-94 0.02 0.02 0.32 0.02 0.02 0.15
1995 and up 0.06 0.03 0.03 0.06 0.03 0.04
Husband's mother's Up to 1900 -1.24 0.25 <0.0001 -1.22 0.25 <0.0001
birth cohort
1901-20 -0.25 0.02 <0.0001 -0.25 0.02 <0.0001
1921-30 -0.11 0.01 <0.0001 -0.11 0.01 <0.0001
1931^40 0.00 0.00
1941-50 0.09 0.01 <0.0001 0.10 0.01 <0.0001
1951-60 0.22 0.04 <0.0001 0.21 0.04 <0.0001
1961 and up 1.90 0.40 <0.0001 1.49 0.41 <0.0001
Wife's mother's Up to 1900 -0.64 0.34 0.06 -0.67 0.34 0.05
birth cohort
1901-20 -0.34 0.02 <0.0001 -0.34 0.02 <0.0001
1921-30 -0.13 0.01 <0.0001 -0.13 0.01 <0.0001
1931^40 0.00 0.00
1941-50 0.17 0.01 <0.0001 0.19 0.01 <0.0001
1951^60 0.35 0.03 <0.0001 0.36 0.03 <0.0001
1961 and up 0.40 0.30 0.18 0.34 0.30 0.25
Husband's parents' Intact family 0.00 0.00
divorce experience
Parents' death 0.06 0.01 < 0.0001 0.06 0.01 < 0.0001
Indeterminate family 0.19 0.01 <0.0001 0.26 0.03 <0.0001
structure

Parents' divorce any age


Parents' divorce- ages 0-2 0.44 0.10 <0.0001 0.46 0.07 <0.0001
Parents' divorce- ages 3-5 0.44 0.05 < 0.0001 0.41 0.04 < 0.0001
Parents' divorce- ages 6-3 0.48 0.04 < 0.0001 0.42 0.04 < 0.0001
Parents' divorce- ages 9-11 0.44 0.04 < 0.0001 0.45 0.03 < 0.0001
Parents' divorce- age 12-14 0.42 0.03 < 0.0001 0.37 0.03 < 0.0001
Parents' divorce- age 15-17 0.36 0.03 <0.0001 0.34 0.03 <0.0001
Parents' divorce- age 18-20 0.41 0.03 <0.0001 0.37 0.03 <0.0001
Parents' divorce- age 21-23 0.33 0.04 <0.0001 0.34 0.03 <0.0001
Parents' divorce- age 24-26 0.27 0.05 < 0.0001 0.22 0.05 < 0.0001
Parents' divorce- age 27-29 0.20 0.08 0.01 0.28 0.08 0.00
Parents' divorce- age 30 + 0.31 0.12 0.01 0.27 0.12 0.02
Husband's mother's Not remarried
subsequent transitions
Remarried -0.02 0.03 0.49
Second divorce 0.09 0.04 0.05
Husband's father's Not remarried
subsequent transitions
Remarried 0.02 0.03 0.41
Second divorce 0.04 0.04 0.41

Wife's parents' Intact family 0.00 0.00


divorce experience
Parents' death 0.06 0.02 0.00 0.07 0.02 < 0.0001
Indeterminate family 0.26 0.01 < 0.0001 0.39 0.03 < 0.0001
structure

Parents' divorce any age


Parents' divorce- age 0-2 0.70 0.07 < 0.0001 0.70 0.05 < 0.0001
Parents' divorce- age 3-5 0.77 0.04 < 0.0001 0.64 0.04 < 0.0001
Parents' divorce- age 6-8 0.67 0.03 < 0.0001 0.59 0.03 < 0.0001
Parents' divorce- age 9-11 0.56 0.03 <0.0001 0.51 0.03 <0.0001

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
182 Torkild Hovde Lyngstad and Henriette Engelhardt

Table 2 (Continued)

Model 1 Model 2

Variable Category Beta SE p Beta SE p

Parents' divorce- age 12-14 0.50 0.03 <0.0001 0.51 0.03 <0.0001
Parents' divorce- age 15-17 0.48 0.03 < 0.0001 0.50 0.03 < 0.0001
Parents' divorce- age 18-20 0.45 0.03 < 0.0001 0.39 0.03 < 0.0001
Parents' divorce- age 21-23 0.38 0.04 <0.0001 0.36 0.04 <0.0001
Parents' divorce- age 24-26 0.44 0.06 < 0.0001 0.41 0.06 < 0.0001
Parents' divorce- age 27-29 0.46 0.11 <0.0001 0.41 0.10 <0.0001
Parents' divorce- age 30+ 0.24 0.20 0.22 0.43 0.18 0.02
Wife's father's subsequent Not remarried
transitions
Remarried -0.01 0.02 0.72
Second divorce 0.01 0.04 0.75

Wife's mother's subsequent Not remarrie


transitions
Remarried 0.01 0.02 0.54
Second divorce 0.05 0.04 0.14
-2 log likelihood 525,033.6 524,655.3
Source: As for Table 1.

parent of the opposite sex. We cannot directly testeffect with the roles husband and wife play in their
this hypothesis because we do not have information marriages in general and the divorce decision-mak-
about the spouses' actual living arrangements aftering process in particular. When a couple divorce, it is
the wife who most often initiates the dissolution
the dissolution of their parents' marriage. However,
if the hypothesis were true, we would expect the(Kalmijn and Poortman 2006). If her parents'
divorce results in a weaker commitment on her
intergenerational transmission of divorce to be
stronger for men than for women, because most part to the maintenance of a lifelong marriage, as is
children live with their mothers after the dissolution. suggested by other contributions to the literature
such as that of Amato and DeBoer (2001), her
This means that the 'same-sex hypothesis' is not
supported by our evidence because in fact the effects tolerance for periods with low relationship quality
could be lower than that of the man. This, in
for wives are markedly stronger. One possible
interpretation of this sex difference in the transmis- combination with a greater propensity to take the
sion is that it is a combination of the transmission initiative in dissolving the relationship, might ex-

2.0 i - N^-

^v

1 -8 - ^V

.2 16~ ** ^ ^'
O 1.4- ^-- "~ ^~^x

1.2-

l.o H

0-2 3-5 6-8 9-11 12-14 15-17 18-20 21-23 24-26 27-29 30+
Age at parents' divorce

Figure 1 Odds ratios of divorce by age at parents' divorce. Norwegian first marriages 1980-2003.
Source: Regression coefficients in Table 2.

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
The intergenerational transmission of divorce 183

plain a stronger intergenerational transmission of Any interpretation of the results must be made
divorce for wives. with caution, since the results come with several
The degree of support for our third hypothesis, caveats. Because we cannot control for all other
which predicted an age gradient in the transmission relevant factors, some of the transmission effects we
effect, is somewhat mixed. Figure 1 shows odds observe will be spurious owing to the effects of both
ratios of divorce by age at parents' divorce. The social and genetic variables. The magnitude of the
excess risk of marital dissolution for those who
effects is therefore difficult to assess precisely. It is
experienced parents' divorce does decline with age
possible that the magnitude would be lower if we
at parents' divorce, and it does so regardless
wereof
able to control for more of the unmeasured
which spouse experienced it. However, the gradient
characteristics of parents and children. In this
is not very steep for husbands. For wives,context,
the it is relevant to note that Furstenberg and
changes in the regression coefficients by age at
Kiernan (2001) report that selection on variables
parents' divorce correspond in strength to those for
such as behaviour problems and low school achieve-
changes in other predictors of divorce that are in childhood is unimportant for the interge-
ment
recognized as important in the divorce literature nerational transmission of divorce.
such as the wife's age at marriage. Future research could improve on this study in a
A simple prediction for four different hypothe- number of ways: by attempting to include cohabi-
tical marriage cohorts experiencing the divorce tants because an increasing proportion of children
probabilities estimated from our third modelexperienceover the break-up of the relationship of their
their first 10 years of marriage can serveunmarried as an parents; by seeking a better understand-
ample illustration of the strength of the ing age of the role played by choice of union type and
gradient. Indeed, the prediction indicates that thethe
selectivity from cohabitation into marriage; by
expected percentages of divorced couples where the
removing the influence of other unmeasured factors
husband experienced parents' divorce at ages related 4 and both to the spouses and their parents or by
22 are 44 and 42 per cent, respectively. The taking such factors into account; and by incorporat-
hypothetical marriage cohorts belong to the refer- ing data on the different pathways taken after the
ence group on all variables except those measuring divorce including the various forms of step-family
parents' divorce. The corresponding proportions for children of divorce live in.
couples where the wife's parents divorced at the
same ages are 40.4 and 32.5 per cent, suggesting
that, all other things equal, almost 10 per cent fewer Notes
of such couples in a hypothetical marriage cohort
would not have divorced if the wife's parents had 1 Torkild Hovde Lyngstad is a postdoctoral research
divorced after she reached adulthood and an age at fellow in the Department of Sociology and Human
which she had most probably already left the Geography, University of Oslo, PO Box 1095 Blindem,
parental home. 0317 Oslo, Norway. E-mail: t.h.lyngstad@sosgeo.uio.no.
Our fourth hypothesis stated that those who Henriette Engelhardt is Professor of Demography in the
experienced subsequent transitions after their par- Department of Population Studies, University of Bam-
ents' first divorce would be even more affected by berg, Kapuzinerstrae 16, D-96047 Bamberg, Germany.
the intergenerational transmission than those who E-mail: henriette.engelhardt-woelfler@uni-bamberg.de
did not experience them. In the event, unlike some 2 For comments on an earlier version, the authors are
earlier studies, we do not find that multiple transi- grateful to two anonymous reviewers, Maria Stanfors,
tions affect divorce risk. When their risks of experi- and audiences at seminars in Budapest, Klaekken, and
encing multiple transitions (parents' remarriage and London. They also wish to thank the Vienna Institute
a second divorce) are taken into account, the for Demography and the Centre for Advanced Study at
estimates for the effect of parents' divorce by age the Norwegian Academy of Letters and Science for
at disruption are somewhat reduced, but the main office space, and the Research Council of Norway for
conclusions hold. Why these later life-course financial support.
changes do not affect divorce risk and the strength
of the intergenerational transmission in Norway is
unclear. Our results are consistent, however, with References
those of Teachman (2002), who reports that parents'
remarriage is not an additional factor increasing the Allison, Paul D. 1995. Survival Analysis using the SAS
offspring's divorce risk. System. Cary, NC: SAS Institute.

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
184 Torkild Hovde Lyngstad and Henriette Engelhar dt

Amato, Paul R. and Alan Booth. 1991. Consequences of Greenberg, Ellen F. and W. Robert Nay. 1982. The
parental divorce and marital unhappiness for adult intergenerational transmission of divorce reconsidered,
well-being, Social Forces 69(3): 895-914. Journal of Marriage and Family 44(2): 335-347.
Amato, Paul R 1996. Explaining the intergenerational Hayashi, Gina M. and Bonnie R. Strickland. 1998. Long-
transmission of divorce, Journal of Marriage and Family term effects of parental divorce on love relationships:
58(3): 628-640. divorce as attachment disruption, Journal of Personality
Amato, Paul R. and J. G. Gilbreth. 1999. Non-resident and Social Psychology 15(1): 23-38.
fathers and children's well-being: a meta-analysis, Kalmijn, Matthijs and Anne-Rigt Poortman. 2006. His or
Journal of Marriage and Family 61(3): 557-573. her divorce? The gendered nature of divorce and its
Amato, Paul R and Danelle D. DeBoer. 2001. The determinants, European Sociological Review 22(2):
201-214.
transmission of marital instability across generations:
relationships skills or commitment to marriage?, Keith, Verna M. and Barbara Finlay. 1988. The impact of

Journal of Marriage and Family 63(4): 1038-1051. parental divorce on children's education attainment,
Diekmann, Andreas and Henriette Engelhardt. 1999. Themarital timing, and likelihood of divorce, Journal of
Marriage and Family 50(3): 797-809.
social inheritance of divorce: effects of parents' family
Kiernan, Kathleen and Andrew Cherlin. 1999. Parental
type in postwar Germany, American Sociological Re-
divorce and partnership dissolution: evidence from a
view 64(6): 783-793.
British cohort study, Population Studies 53(1): 39^8.
Diekmann, Andreas and Kurt Schmidheiny. 2006. The
Kiernan, Kathleen. 2000. The state of European unions: an
intergenerational transmission of divorce: a fifteen-
analysis of FFS data on partnership formation and
country study with family and fertility surveys. Paper
dissolution. Paper presented at the FFS Flagship
presented at the Population Association of America
Conference, UNECE Population Activities Unit,
Annual Meeting 2006 in Los Angeles, CA.
Brussels.
D'Onofrio, B. M., E. N. Turkheimer, R. E. Emery, K. P.
King, Valarie. 2002. Parental divorce and interpersonal
Harden, W. Slutske, A. Heath, P. A. F. Madden, and N.
trust in adult offspring, Journal of Marriage and Family
G. Martin.. 2007. A genetically informed study of the
64(3): 642-656.
intergenerational transmission of marital instability,
Li, Jui-Chung Allen and Lawrence L. Wu. 2008. No trend
Journal of Marriage and Family 69(3): 793-803.
in the intergenerational transmission of divorce, Demo-
Dronkers, Jaap and Juho Hrknen. 2008. The interge-
graphy 45(4): 875-883.
nerational transmission of divorce in cross-national
McLanahan, Sara and Larry L. Bumpass. 1988. Interge-
perspective: results from the Fertility and Family
nerational consequences of family disruption, American
Surveys, Population Studies 62(3): 273-288.
Journal of Sociology 94(1): 130-152.
Engelhardt, Henriette, Heike Trappe, and Jaap Dronkers.
Mueller, Charles W and Hallowell Pope. 1977. Marital
2002. Differences in family policy and the intergenera-
instability: a study of its transmission between genera-
tional transmission of divorce: a comparison between
tions, Journal of Marriage and Family 39(1): 83-93.
the former East and West Germany, Demographic
Poortman, Anne-Rigt and Matthijs Kalmijn. 2002. Wo-
Research 6(11): 295-324. men's labor market position and divorce in the Nether-
Feng, Du, Roseann Giarrusso, Vern L. Bengtson, and
lands: evaluating economic interpretations of the work
Nancy Frye. 1999. Intergenerational transmission of
effect, European Journal of Population 18(2): 175-202.
marital quality and marital instability, Journal of Hallowell and Charles W Mueller. 1976. The
Pope,
Marriage and Family 61(2): 451-463. intergenerational transmission of marital instability:
Franklin, K. M., R. Janoff-Bulman, and J. E. Roberts.
comparisons by race and sex, Journal of Social Issues
1990. Long-term impact of parental divorce on opti-
32(1): 49-66.
mism and trust: changes in general assumptions Powell,
or Brian and Douglas B. Downey. 1997. Living in
narrow beliefs, Journal of Personality and Social Psy-
single-parent households: an investigation of the same-
chology 59(4): 743-755. sex hypothesis, American Sociological Review 62(4):
Furstenberg, Frank F and Kiernan Kathleen. 2001.521-539.
Delayed parental divorce: how much do children
Raley, Sara and Suzanne Bianchi. 2006. Sons, daughters,
benefit?, Journal of Marriage and Family 63(2): 446-
and family processes: does gender of children matter?,
457. Annual Review of Sociology 32: 401^421.
Glenn, Norval D. and B. Kramer Kathryn. 1987. R0ed,
The Knut and Oddbj0rn Raaum. 2003. Administrative
marriages and divorces of the children of divorce,
registers - unexplored reservoirs of scientific knowl-
Journal of Marriage and Family 49(4): 811-825. edge?, The Economic Journal 113(488): F258-F281.

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms
The intergenerational transmission of divorce 185

Sigle-Rushton, Wendy and Sara McLanahan. 2002. Father Traag, Tanja, Jaap Dronkers, and Louis- Andr Vallet.
absence and child well-being: a critical review. Center 2000. The intergenerational transmission of divorce in
for Research on Child Wellbeing Working Paper #02- France. Paper presented at the ISA RC28 Spring
20, Center for Research on Child Wellbeing, Princeton Meeting 2000, Libourne, France.
University, Princeton, NJ. Wolfinger, Nicholas H 1999. Trends in the intergenera-
St0rksen, Ingunn, Espen R0ysamb, Hkon K. Gjessing, tional transmission of divorce, Demography 36(3):
Torbj0rn Mourn, and Kristian Tambs. 2007. Marriages 415^20.

and psychological distress among adult offspring of Wolfinger, Nicholas H 2005. Understanding the Divorce
divorce: a Norwegian study, Scandinavian Journal of Cycle. Cambridge, MA: Cambridge University Press.
Psychology 48(6): 467-476. Wu, Lawrence L 1996. Effects of family instability, income,
Teachman, Jay 2002. Childhood living arrangements and and income instability on the risk of a premarital birth,
the intergenerational transmission of divorce, Journal American Sociological Review 61(3): 386-406.
of Marriage and Family 64(3): 717-729.

This content downloaded from 121.58.232.35 on Thu, 02 Nov 2017 10:23:46 UTC
All use subject to http://about.jstor.org/terms

Vous aimerez peut-être aussi