Family Formation and Offspring Mortality in Sweden

Abstract

Using multigenerational population register data that cover the total Swedish population, we studied relative mortality of offspring whose parents had formed a new family with children. These primarily adult-age children are found to have lower death risks than those with divorced parents who did not form a new family, which highlights that the link between parental family formation and offspring health may be attributed not only to causal factors associated with family disruption but also to social selection in parents. The association differs notably according to whether sibling groups are determined according to the mother or the father. This finding is interpreted as reflecting varying environmental exposure, because most minor children who experience parental divorce remain with the mother. We approximate that parental social selection, which maliciously affects offspring health, raises the offspring mortality risk by 20%.

Keywords

offspring mortality family formation sibling groups

Introduction

Using multigenerational register data that cover the total Swedish population, the aim of this article is to study the relative mortality of offspring whose parents had formed a new family with children. The data contain complete sibling groups on both the mother’s and the father’s side. Since each child’s full-siblings and half-siblings can be identified, the parents can be categorized according to whether they have children with more than one partner. These prerequisites provide an opportunity to illustrate that social selection in parents might affect offspring mortality risks.

It is well established that parental marital status is strongly associated with offspring health and well-being (Amato & Keith, 1991; Maier & Lachman, 2000; Wadsworth & Kuh, 1997). Even when having reached adult ages, people with divorced parents tend to have higher substance use, more behavioral and emotional problems, and higher mortality rates than those with married parents (Buchanan, Brinke, & Flouri, 2000; Emery, 1999; Hetherington, 1999; Pensola, 2003). They also appear to carry traits and behaviors that make them less successful in finding a marriage partner (Erola, Härkönen, & Dronkers, 2012). Many studies attribute the association between parental divorce and offspring health to causal factors, such as inadequate parenting practices, conflict between parents, loss of contact with the noncustodial parent, economic factors, and deleterious effects of stepparents in remarriages (Amato, 2000; Dawson, 1991; Flinn, Leone, & Quinlan, 1999; Lawson & Mace, 2008).

However, selection factors also may account for the correlation, meaning that latent characteristics of the parents can lead to both divorce and difficulties in the offspring. Thus, at least some of the health problems in offspring of divorced families may not be caused only by the divorce but can also result from selection factors that predispose parents both to divorce and to having unhealthy children (Cherlin, Chase-Lansdale, & McRae, 1991; Doherty & Needle, 1991; Emery, Waldron, Kitzmann, & Aaron, 1999). Potential selection factors include parental antisocial behavior, personality disorders, and unobserved sociodemographic characteristics (Emery, 1999; Rowe, 1994; Simons & Chao, 1996), as well as genetic factors transmitted from divorce-prone parents to their offspring (D’Onofrio et al., 2007; Jockin, McGue, & Lykken, 1996; McGue & Lykken, 1992).

Based on previous research, it is not clear whether children with a parent who divorced, remarried, and formed a new family have better health than those with a divorced parent who never remarried. On the one hand, parental divorce and remarriage are known to involve increased stress levels, conflicts within the new household, and a lack of institutionalized social support (Capaldi & Patterson, 1991; Coleman, Ganong, & Fine, 2000; Kurdek & Fine, 1993). On the other hand, a divorced parent who remarries may yet be better equipped to provide material and social support for the children than one who remains divorced (Fritzell & Burström, 2006; Morrison & Ritualo, 2000; Weiss, 1984). This article contributes to the literature by quantifying offspring’s death risks associated with a parent’s formation of a new family with children.

An important issue concerning this is who forms the new family. In the cohorts studied here, the majority, or more than 85%, of all minor children who had experienced parental divorce remained with the mother (Statistics Sweden, 2010). Defining a sibling group by the mother implicates that her potential remarriage and formation of a new family with children reflects not only her parental characteristics but also the living conditions in the new family, which are likely to have an impact on health and long-term mortality of the offspring (Bowlby, 1970; Rossi & Rossi, 1990; Umberson & Chen, 1994). A sibling group as defined by the father, on the contrary, is not to the same extent contaminated by the effects induced by the family environment, because the offspring of a father who divorced did not generally remain with the father. Children from the father’s first marriage are not subject to the environmental exposure of the new family, whereas those from the second marriage have not experienced any parental separation.

Father-defined sibling groups are therefore more capable of reflecting social selection into remarriage and the formation of a new family than mother-defined sibling groups. If offspring health relates to parental social selection and not only to causal effects of remarriage or environmental exposure, we expect to find that children with a father who divorced and formed a new family should have approximately the same low mortality risks as those with a married father who never divorced.

We determine the extent to which deaths cluster within families, since siblings tend to resemble each other on unobservable family characteristics associated with health and mortality. Death clustering may occur because of shared latent risk factors, such as genetic endowment, social environment, parental competence, and health-related knowledge, or due to a causal process, where one child’s death raises the death risk of the next child in the family (Arulampalam & Bhalotra, 2006; Omariba, Rajulton, & Beaujot, 2008). Numerous studies on infant mortality in developing countries have found siblings’ survival chances to be interrelated (Das Gupta, 1997; Guo, 1993; Zenger, 1993). Socialization with regard to health and risk behaviors within the family, similar environmental exposures at childhood, imitation of the siblings’ norms, and hereditary factors and biological predisposition suggest death clustering in advanced societies also, where sibling groups are small and mortality rates are modest (Rostila & Saarela, 2011). To our knowledge, this is the first study that explicitly accounts for death clustering at noninfant and primarily adult ages.

Data

The data for this study come from the Swedish Work and Mortality Data Base (HSIA), maintained at the Centre for Health Equity Studies (CHESS) in Stockholm. The HSIA is a multiply linked database of national Swedish routine registers. For all persons born in Sweden during the period 1932 to 1980 and alive at the end of 1980, there is individual-level information about basic demographic and socioeconomic variables and the month of death for all who died during the period 1981 to 2002. Each person can be linked to his or her registered mother and father, provided that the parent was alive at the end of 1980. We can hereby determine sibling groups as defined by the mother and the father. To account for both the mother’s and the father’s observable characteristics, we focus on offspring whose both parents were alive at the end of 1980.

Studying sibling groups, and particularly those of noninfant ages, have some notable consequences. First, since the children are nested within parents, the target population should be defined according to the parents’ birth year, in spite of the units under analysis being children. Second, comprehensive analyses that account for potential death clustering require that families are complete. A narrow birth cohort restriction of the parents is therefore necessary. Considering that the fertile period in women roughly stretches from 15 to 45 years, it would be possible to include mothers born between 1916 and 1935. We have chosen to include mothers born between 1921 and 1935, however, as we want to measure their socioeconomic position at active working age. The first information available on this account is from the end of 1980, and many in the cohorts born between 1916 and 1920 were retired already by that time. For similar reasons, and to get comparable results, fathers included also belong to the birth cohorts 1921 to 1935. The sibling groups studied therefore consist of offspring who all are from parents born between 1921 and 1935. Of the 722,522 children analyzed, 62,283 had half-siblings: 39,970 on the mother’s side, 31,606 on the father’s side, and 9,293 on both the mother’s and the father’s side.

The age distribution of the offspring is given in Figure 1. More than 80% of all the children were at least 18 years at the end of 1980, and almost 70% were aged between 18 and 30 years. Since the offspring were followed over a 22-year period, and only 1.3% of all deaths have occurred before age 18 years, the mortality analyses concern primarily adult ages. Complementary analyses that focus on adult-aged offspring only yield results that are close to identical to those reported here.

Figure 1.

Age distribution of the offspring.

Table 1 gives the distribution of families by the number of children and the number of deaths in offspring when siblings groups are defined according to the mother and the father, respectively. More than three quarters of all the families had more than one child, and almost 90% of the children had at least one sibling. There are in total 13,692 offspring deaths, whereof approximately 88% have occurred in families with more than one child. There is evidently some death clustering, as 6% of all deaths have occurred in families that have experienced more than one child’s death.

Table 1.

Distribution of Families by the Number of Children and the Number of Child Deaths When Sibling Groups Are Defined According to the Mother and the Father.

	No. of child deaths in the family
No. of children in the family	0	1	2	3	4	Total families	Percentage of all children	Percentage of all deaths
Mother defined
1	72,443	1,624				74,067	10.3	11.9
2	129,190	4,632	73			133,895	37.1	34.9
3	65,717	3,530	127	1		69,375	28.8	27.7
4	23,138	1,711	94	3	0	24,946	13.8	13.9
5	7,408	782	60	1	0	8,251	5.7	6.6
6	2,573	321	29	3	0	2,926	2.4	2.8
7	933	126	16	1	0	1,076	1.0	1.2
8	358	62	6	0	0	426	0.5	0.5
9	134	25	2	0	0	161	0.1	0.2
10	71	15	1	0	0	87	0.0	0.1
11	15	5	0	0	1	21	0.0	0.1
12	14	7	1	0	0	22	0.0	0.1
13	9	2	0	0	0	11	0.0	0.0
14	0	0	0	0	0	0	0.0	0.0
15	1	0	0	0	0	1	0.0	0.0
16	1	0	0	0	0	1	0.0	0.0
17	0	1	0	0	0	1	0.0	0.0
Total families	302,005	12,843	409	9	1	315,267	100.0	100.0
Percentage of all children	94.8	5.0	0.2	0.0	0.0	100.0
Percentage of all deaths		93.8	6.0	0.2	0.0	100.0
Father defined
1	77,643	1,767				79,410	11.0	12.9
2	128,593	4,627	77			133,297	36.9	34.9
3	64,913	3,440	126	1		68,480	28.4	27.0
4	22,811	1,698	88	3	0	24,600	13.6	13.8
5	7,362	761	54	1	0	8,178	5.7	6.4
6	2,587	329	30	4	0	2,950	2.4	2.9
7	941	120	12	1	0	1,074	1.0	1.1
8	375	61	6	1	0	443	0.5	0.6
9	133	24	2	0	0	159	0.2	0.2
10	64	19	1	0	0	84	0.1	0.2
11	18	6	0	0	1	25	0.0	0.1
12	16	6	1	0	0	23	0.0	0.1
13	10	2	0	0	0	12	0.0	0.0
14	0	0	0	0	0	0	0.0	0.0
15	1	0	0	0	0	1	0.0	0.0
16	1	0	0	0	0	1	0.0	0.0
17	0	1	0	0	0	1	0.0	0.0
Total families	305,468	12,861	397	11	1	318,738	100.0	100.0
Percentage of all children	94.8	5.0	0.2	0.0	0.0	100.0
Percentage of all deaths		93.9	5.8	0.2	0.0	100.0

Of the 315,267 mothers, 12,698 had children with more than one partner, meaning that these children have half-siblings. The corresponding number for the fathers is 9,311 out of 318,738 (Table 2). Only 594 mothers and 501 fathers had children with more than two partners.

Table 2.

Distribution of Mothers and Fathers by the Number of Partners With Whom They Have Children.

No. of partners	Mothers	Fathers
1	302,569	309,427
2	12,104	8,810
3	549	459
4	43	35
5	2	3
6	0	3
7	0	0
8	0	0
9	0	1
Total	315,267	318,738

Analytical Framework

The offspring death risk, which is the outcome of interest, is analyzed by age-specific mortality hazard models. As discussed below, there is no within-family variation on the family variables used. We therefore use a random effects specification to incorporate the possibility of death clustering (see Allison, 2005).

Variables

Since the children are grouped according to whether they are siblings, men and women must be included in the same models. The child’s sex is therefore used merely as a covariate, in spite of the generally large sex differences in mortality rates. A strategy to circumvent the problem would be to focus on same-sex sibling groups, but even with total data for a population of Sweden’s size, the number of deaths is too small to obtain enough statistical power. Stratifying the hazard models by child’s sex does not affect estimates for the other variables of interest, so we report only the results of the more parsimonious models.

We use a piecewise linear spline function for age. Age at entry is the child’s age at the end of 1980, that is, the clock for the hazard models start ticking from the child’s age when he or she enters the observation window.

Offspring’s socioeconomic background is measured by each parent’s socioeconomic position at the end of 1980, that is, when the parent was aged between 45 and 59 years. It distinguishes among blue-collar workers, lower-level white-collar workers, upper-level white-collar workers, self-employed, early retired, house-workers, and others. Offspring’s county of residence at the end of 1980 (25 categories) is included to capture any regional variation in mortality rates.

As for socioeconomic background and residence, information about a parent’s marital status is available only from the end of 1980. However, by using it as a complement to the data on each child’s full-siblings and half-siblings, we know if a separated parent had formed a new family with children. This means that in the data, we can separate and compare offspring of parents who differ on the outcomes of the family formation process. The process itself, in terms of creation and disruption of families, cannot be observed explicitly. Hence, we categorize the offspring according to a summary of the reference parent’s (not each child’s) family history.

In our preliminary analyses, we attempted to utilize information about the offspring’s own socioeconomic position, which is available for the years 1980, 1985, and 1990, but soon we realized that parental socioeconomic position is to be preferred. Using the offspring’s own socioeconomic position will lead to a considerable part of the observations being inadequately classified, particularly those who die young or early in the process as they have not established a socioeconomic position. Studies have found that socioeconomic background as measured by parental socioeconomic position and own socioeconomic position have similar effects on offspring mortality risks and lead to similar conclusions with regard to the effects of other common mortality predictors (Saarela & Finnäs, 2009, 2011). Information about offspring marital status cannot be used, as it is available only from the end of 1980.

Table 3 shows how the children are distributed when the information on siblings and parental marital status is combined. The reference parent is the mother if sibling groups are defined on basis of the mother, and the father in case sibling groups are defined on basis of the father. This description certainly contains some ambiguities derived from the information about parents’ marital status, but since the sibling information is complete and fully reliable, it serves well to reflect the focal point of interest, which is the parents’ formation of families with children.

Table 3.

Number of Children by the Parent’s Marital Status and the Presence of Siblings When Sibling Groups Are Defined According to the Mother and the Father.

	The child has
	No siblings	Full-siblings only	Half-siblings
Mother’s marital status
Married	62,649	550,369	28,904
Divorced	8,609	56,343	10,307
Never married	2,473	1,302	625
Widow	336	471	134
Father’s marital status
Married	68,224	559,296	24,371
Divorced	8,965	50,482	6,899
Never married	1,911	1,398	280
Widower	310	330	56

The two largest groups in each panel, offspring with a married reference parent and no siblings or full-siblings only (Table 3, row 1, columns 1 and 2), represent the offspring of parents who did not form a new family with children. The lion’s share of all offspring in this category comes from stable marriages. There is some misclassification, however, since we cannot separate parents who remarried without forming a new family with children. Columns 1 and 2 of row 2 represent offspring with a reference parent who did not find a new marriage partner after divorce. These children have consequently no siblings or full-siblings only. If the reference parent is married and the offspring have half-siblings (row 1, column 3), the parent had formed a new family with children. Offspring with half-siblings whose reference parent is divorced (row 2, column 3) generally come from parents who had split their family more than once. In the parent cohorts studied here, consensual unions among childless persons were rare and parents who never married few (Andersson, 1998). Children with never-married parents (row 3) therefore constitute a marginal group, but they are included in the analyses for the sake of completeness. Parents to these children represent consensual unions, mothers who constantly had been sole supporters and fathers who had acknowledged their paternity but never went into marriage (with any woman). A widowed reference parent (row 4) means that he or she has remarried and afterward became a widow or widower, as we condition on that a child’s both parents were alive at the end of 1980.

Our primary aim is to study the relative mortality of offspring who have half-siblings, and particularly those with (re)married parents. Inference about the potential role of social selection is drawn by comparing the relative mortality risk from models based on father-defined families with that from corresponding models based on mother-defined families.

Statistical Models

To determine the importance of death clustering, we estimate one set of ordinary hazard models with independent observations and another set of models where the offspring of the same sibling group are treated as dependent observations.

The hazard model for offspring mortality with independent observations, that is, which does not account for potential death clustering, takes the following form if the sibling groups are defined by the mother:

ln μ (t) = y (t) + β_{1} s e x + β_{2} r e g i o n + β_{3}^{m d} {[m a r \times s i b]}^{m d} + β_{4} m s o c e c + β_{5} f s o c e c + z,

where y(t) represents the baseline duration pattern by child’s age, captured by piecewise linear splines together with a constant. Explanatory variables included are the child’s sex, the child’s region of residence, the variable combining the reference parent’s marital status and the child’s siblings, the mother’s socioeconomic position, and the father’s socioeconomic position. The vector of parameters associated with the explanatory variables is given by the βs. The term z represents a normally distributed error term. The corresponding specification if sibling groups are defined by the father is the following:

ln μ (t) = y (t) + β_{1} s e x + β_{2} r e g i o n + β_{3}^{f d} {[m a r \times s i b]}^{f d} + β_{4} m s o c e c + β_{5} f s o c e c + z .

Differences between Equations 1 and 2 are in the term β₃[mar × sib], since mother-defined sibling groups (md) specify a variable that differs from father-defined sibling groups (fd).

If observations are dependent, then the mother-defined model is

{ln μ}_{j} (t) = y (t) + β_{1} s e x + β_{2} r e g i o n + β_{3}^{m d} {[m a r \times s i b]}^{m d} + β_{4}^{m d} m s o c e c + β_{5} f s o c e c + u_{j} + v_{j i},

and the father-defined model is

{ln μ}_{j} (t) = y (t) + β_{1} s e x + β_{2} r e g i o n + β_{3}^{f d} {[m a r \times s i b]}^{f d} + β_{4} m s o c e c + β_{5}^{f d} f s o c e c + u_{j} + v_{j i} .

The lower indexes j and i stand for the family (sibling group) and the child, respectively. Since the children are nested within mothers in the mother-defined specification, and within fathers in the father-defined specification, the parameters β₄ and β₅ (and of course the parameter β₃ and its associated variable) differ between the two equations. Unobserved family-specific random effects are represented by u_j, which is assumed uncorrelated with the explanatory variables. The child-specific error term is given by v_ji. Both these components are assumed normally distributed with zero means and independent of each other. The variance of the composite error term, ϵ_ji = u_j + v_ji, is thereby

V a r (ε_{j i}) = σ_{u}^{2} + σ_{v}^{2} = σ_{u}^{2} + 1 .

Offspring death clustering within families means that

C o r r (ε_{j k}, ε_{j q}) \equiv ρ = \frac{σ_{u}^{2}}{σ_{u}^{2} + 1}, k, q = 1, 2, K, n_{i}, k \neq q .

It is the standard deviation of the parameter representing unobserved family-specific effects, σ_u, that will be integrated out. Following Equation 6, it is defined as

σ_{u} = \sqrt{ρ / (1 - ρ)} .

Results

The estimation results are summarized in Table 4. The first two columns refer to models where observations (siblings) are considered independent, and models are estimated separately according to the mother and the father.

Table 4.

Estimates of Hazard Models for Offspring Mortality.

	Mother-defined family, independent deaths	Father-defined family, independent deaths	Mother-defined family, dependent deaths	Father-defined family, dependent deaths
Parent’s marital status × siblings
Married × No siblings	−0.1134 (0.0478)	−0.0135 (0.0503)	−0.1167 (0.0493)	−0.0184 (0.0522)
Married × Full-siblings only	−0.1269 (0.0402)	−0.0356 (0.0441)	−0.1258 (0.0418)	−0.0358 (0.0460)
Married × Half-siblings	—	—	—	—
Divorced × No siblings	0.0344 (0.0760)	0.2172 (0.0751)	0.0348 (0.0778)	0.2161 (0.0773)
Divorced × Full-siblings only	0.0908 (0.0472)	0.1856 (0.0514)	0.0971 (0.0491)	0.1905 (0.0536)
Divorced × Half-siblings	0.2524 (0.0690)	0.1666 (0.0841)	0.2623 (0.0722)	0.1700 (0.0874)
Never married × No siblings	0.0979 (0.1360)	0.3263 (0.1431)	0.0986 (0.1388)	0.3294 (0.1463)
Never married × Full-siblings only	0.3452 (0.1870)	0.5783 (0.1719)	0.3292 (0.1955)	0.5840 (0.1788)
Never married × Half-siblings	0.5175 (0.2169)	−0.0386 (0.4087)	0.5268 (0.2298)	−0.0388 (0.4382)
Widow(er) × No siblings	0.3014 (0.2365)	0.0227 (0.2882)	0.3008 (0.2480)	0.0325 (0.2937)
Widow(er) × Full-siblings only	0.3231 (0.2118)	−0.2699 (0.3598)	0.3316 (0.2253)	−0.2684 (0.3850)
Widow(er) × Half-siblings	0.4289 (0.4100)	0.5174 (0.5708)	0.4695 (0.4401)	0.5519 (0.6477)
Mother’s socioeconomic position
Blue-collar worker	—	—	—	—
Lower-level white-collar worker	−0.0452 (0.0271)	−0.0465 (0.0271)	−0.0457 (0.0278)	−0.0474 (0.0278)
Upper-level white-collar worker	0.0989 (0.0334)	0.0960 (0.0334)	0.1030 (0.0342)	0.0996 (0.0341)
Self-employed	0.0824 (0.0386)	0.0729 (0.0386)	0.0860 (0.0398)	0.0765 (0.0398)
Pensioner	0.0771 (0.0372)	0.0941 (0.0371)	0.0762 (0.0390)	0.0943 (0.0389)
House-worker	0.0552 (0.0260)	0.0441 (0.0259)	0.0566 (0.0269)	0.0456 (0.0268)
Other	0.0532 (0.0346)	0.0536 (0.0346)	0.0576 (0.0357)	0.0587 (0.0356)
Father’s socioeconomic position
Blue-collar worker	—	—	—	—
Lower-level white-collar worker	−0.1432 (0.0297)	−0.1409 (0.0298)	−0.1467 (0.0306)	−0.1453 (0.0307)
Upper-level white-collar worker	−0.1051 (0.0236)	−0.0992 (0.0237)	−0.1054 (0.0242)	−0.0997 (0.0243)
Self-employed	−0.1032 (0.0302)	−0.0985 (0.0302)	−0.1047 (0.0311)	−0.1006 (0.0313)
Pensioner	0.0435 (0.0333)	0.0378 (0.0334)	0.0418 (0.0346)	0.0350 (0.0348)
House-worker	−0.1405 (0.2128)	−0.1145 (0.2133)	−0.1331 (0.2210)	−0.1047 (0.2222)
Other	0.1058 (0.0405)	0.1037 (0.0405)	0.1085 (0.0423)	0.1069 (0.0425)
Child’s sex
Man	—	—	—	—
Woman	−0.5690 (0.0180)	−0.5690 (0.0180)	−0.5774 (0.0183)	−0.5779 (0.0183)
Constant	−8.4586 (1.1997)	−8.5295 (1.2003)	−8.8201 (1.2002)	−8.9023 (1.2008)
Age interval of duration spline
<20 years	−0.1393 (0.0649)	−0.1397 (0.0650)	−0.1358 (0.0650)	−0.1359 (0.0650)
20-30 years	0.2093 (0.0210)	0.2092 (0.0210)	0.2066 (0.0210)	0.2063 (0.0210)
>30 years	0.0690 (0.0006)	0.0691 (0.0006)	0.0702 (0.0007)	0.0703 (0.0007)
σ_u			0.7729 (0.0367)	0.7825 (0.0362)
Log likelihood	−132807.8725	−132822.7855	−132748.9038	−132762.0398

Note. Standard errors are given in parentheses. Each model accounts also for the child’s county of residence, but the estimates are not shown here. The total number of children, mothers, and fathers is 722,522, 315,267, and 318,738, respectively. The total number of child deaths is 13,692, and the total risk time is 15,781,657 person years.

The effects of the child’s sex and socioeconomic background are generally as expected. The mortality hazard of women is naturally much lower than that of men. The influence of father’s socioeconomic position is stronger than that of mother’s socioeconomic position, which is the reason why some of the estimated effects for the latter appear awkward. For instance, offspring with fathers who were upper-level white-collar workers have a log-hazard of mortality that is 0.1 lower than that of children with fathers who were blue-collar workers, whereas the difference is the opposite, or 0.1 higher, as viewed on basis of mother’s socioeconomic position.

The two latter columns refer to mother-specific and father-specific models where offspring of the same sibling group are treated as dependent observations. The standard deviation of the parameter representing unobserved heterogeneity is about 0.77 (which corresponds to a correlation coefficient of 0.37), with a small standard error of 0.037, saying that the families vary systematically in terms of their offspring’s death risks, even at these primarily adult ages. Some latent family-specific factors consequently underlie the variation in offspring mortality risks. Incorporating the random effect does not change the estimated effects of the other variables or their standard errors to any noteworthy degree or in any systematic way, however.

The variable of central interest is the one that combines the reference parent’s marital status and the child’s siblings. Its estimates have been reorganized and transformed into relative death risks in order to improve readability. Table 5 gives these exponentiated coefficients together with 95% confidence intervals. Children with half-siblings and whose parents were (re)married serve as the reference group, which facilitates the comparisons of greatest interest.

Table 5.

Offspring Mortality Risk Ratios by the Reference Parent’s Marital Status and the Child’s Siblings When Sibling Groups Are Defined According to the Mother and the Father.

	The child has
	No siblings	Full-siblings only	Half-siblings
Mother’s marital status
Married	0.89 (0.81-0.98)	0.88 (0.81-0.96)	1
Divorced	1.04 (0.89-1.21)	1.10 (1.00-1.21)	1.30 (1.13-1.50)
Never married	1.10 (0.84-1.45)	1.39 (0.95-2.04)	1.69 (1.08-2.66)
Widow	1.35 (0.83-2.20)	1.39 (0.90-2.17)	1.60 (0.67-3.79)
Father’s marital status
Married	0.98 (0.89-1.09)	0.96 (0.88-1.06)	1
Divorced	1.24 (1.07-1.44)	1.21 (1.09-1.34)	1.19 (1.00-1.41)
Never married	1.39 (1.04-1.85)	1.79 (1.26-2.55)	0.96 (0.41-2.27)
Widower	1.03 (0.58-1.84)	0.76 (0.36-1.63)	1.74 (0.49-6.18)

Note. 95% confidence intervals are provided within parentheses. The total number of child deaths is 13,692, and the total risk time is 15,781,657 person years. The numbers are based on estimates from the two hazard models with dependent observations (see Table 4).

The results correspond with our expectations. Offspring without siblings and with married parents have practically the same mortality risks as those with full-siblings and married parents. Having a divorced parent and full-siblings, on the other hand, is associated with elevated mortality, approximately 25% higher than that of offspring who come from stable families. The offspring of mothers who divorced and remain unmarried have a 10% higher mortality risk as compared with children whose mothers remarried and formed a new family, that is, children who have half-siblings. In the absence of any environmental influence, which is unlikely in the case of mother-defined sibling groups, this can be interpreted as an effect of parental social selection. If the mother experienced more than one divorce, in the sense that the offspring with half-siblings have a divorced mother, the mortality risk is 30% higher as compared with the reference group. Since these families are defined by the mother, this elevated mortality risk is likely a result of detrimental effects of both environmental exposure and parental social selection.

The pattern is partly different if the family is defined by the father. This has to do with the fact that offspring with half-siblings on the father’s side do not generally live together and therefore cannot be exposed to the same environmental effects as if they had been half-siblings on the mother’s side. We see that the mortality risk of offspring with half-siblings and a (re)married father is the same as the mortality of offspring from stable families. Those with full-siblings and a divorced father, on the other hand, have a 21% higher mortality risk. If the father was divorced, remarried and divorced again, so that the offspring have half-siblings and a divorced father, the offspring mortality risk is practically the same as that of offspring of fathers who remained divorced, that is, of children with full-siblings and a divorced father. Since half-siblings on the father’s side do not generally live together, we interpret these results as induced by parental social selection. Hence, unobserved parental characteristics that maliciously affect offspring health raise the offspring mortality risk by approximately 20%.

Mortality risks in the marginal groups consisting of offspring with never-married parents and divorced parents who had become widow(er)s are generally higher than among offspring who come from stable families. Considering their “nonstandard” family background, this is expected and presumably reflects both parental characteristics and specific rearing conditions at childhood that influence offspring’s long-term health and well-being.

Conclusion

This article illustrates that, like marriage, remarriage and the formation of a new family reflect parental social traits that promote offspring health. We have used Swedish data on complete siblings groups from parents born between 1921 and 1935 and studied offspring mortality over a 22-year period, primarily in adult ages. We find that offspring with divorced parents who formed a new family have lower death risks than those with divorced parents who did not form a new family. The association differs notably according to whether families are defined according to the mother or the father. Offspring with a father who formed a new family have mortality risks that are in parity with those of children from stable marriages, whereas an equally beneficial outcome cannot be observed for offspring with a mother who formed a new family. Few children live with the father after a divorce, meaning that half-siblings on the father’s side are much less likely than half-siblings on the mother’s side to share the same family environment. This finding consequently suggests that the link between parental family formation and offspring health may be attributed not only to causal factors associated with family disruption but also to social selection in parents. Based on the estimates from father-defined families, we approximate that parental social selection, which maliciously affects offspring health, raises the offspring mortality risk by approximately 20%. In contrast to previous research on parental marital status and offspring health, we thus provide an explicit estimate for the potential influence of parental social selection on offspring mortality.

We have accounted for death clustering, since siblings tend to resemble each other on family characteristics associated with health. The data indicate that the families differ on latent risk factors that influence offspring mortality. However, as compared with results from standard models with independent observations, the estimates and their standard errors change modestly.

During the past decades, parents’ separation and formation of new families have become increasingly more common. In Sweden, only 60% of all children who still live at home at age 17 years are found in a household with both their parents (Statistics Sweden, 2007). Many have also experienced a parent’s formation of a new family, meaning that the child has half-siblings. The data used here say that, among the offspring of mothers born before 1915, one in 10 had half-siblings. In the younger mother cohorts born after the mid-1920s, with children born primarily in the late 1950s and early 1960s, the corresponding proportion was about 15%. Considering the high separation rates of parents in today’s Sweden, this development will not taper off in the near future.

The increasing number of persons who come from nonnuclear families implies that future studies concerned with the interrelation between parent’s marital status and offspring health ought to incorporate some dimension of parents’ family history. Yet a growing number of separated parents have shared custody of their minor offspring. The consequences for children who come from later-born parent cohorts will not therefore necessarily be the same as documented here.

Footnotes

Acknowledgements

The authors thank the seminar participants at the Aboa Centre for Economics in Turku, Stockholm University Demography Unit, and the anonymous referees for their comments on previous versions of this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Allison

P. D.

(2005). Fixed effects regression methods for longitudinal data using SAS. Cary, NC: SAS Institute.

Amato

P. R.

(2000). The consequences of divorce for adults and children. Journal of Marriage and the Family, 62, 1269-1287.

Amato

P. R.

Keith

(1991). Parental divorce and adult well-being: A meta-analysis. Journal of Marriage and the Family, 53, 43-58.

Andersson

(1998). Trends in marriage formation in Sweden 1971-1993. European Journal of Population, 14, 157-178.

Arulampalam

Bhalotra

(2006). Sibling death clustering in India: State dependence versus unobserved heterogeneity. Journal of the Royal Statistical Society A, 169, 829-848.

Bowlby

(1970). Attachment and loss (Vol. 1). London, England: Hogarth.

Buchanan

Brinke

J. T.

Flouri

(2000). Parental background, social disadvantage, public “care”, and psychological problems in adolescence and adulthood. Journal of the American Academy of Child and Adolescent Psychiatry, 39, 1415-1423.

Capaldi

D. M.

Patterson

G. R.

(1991). Relation of parental transitions to boys: I. A linear hypothesis. II. Mothers at risk for transitions and unskilled parenting. Developmental Psychology, 27, 489-504.

Cherlin

A. J.

Chase-Lansdale

P. L.

McRae

(1998). Effects of parental divorce on mental health throughout the life course. American Sociological Review, 63, 239-249.

10.

Coleman

Ganong

Fine

(2000). Reinvestigating remarriage: Another decade of progress. Journal of Marriage and the Family, 62, 1288-1307.

11.

Das Gupta

(1997). Socio-economic status and clustering of child deaths in rural Punjab. Population Studies, 51, 191-202.

12.

Dawson

D. A.

(1991). Family structure and children’s health and well-being: Data from the 1998 national health interview survey on child health. Journal of Marriage and the Family, 53, 573-584.

13.

Doherty

W. J.

Needle

R. H.

(1991). Psychological adjustment and substance use among adolescents before and after a parental divorce. Child Development, 62, 328-337.

14.

D’Onofrio

B. M.

Turkheimer

Emery

R. E.

Maes

H. H.

Silberg

Eaves

L. J.

(2007). A children of twins study of parental divorce and offspring psychopathology. Journal of Child Psychology and Psychiatry, 48, 667-675.

15.

Emery

R. E.

(1999). Marriage, divorce, and children’s adjustment. Thousand Oaks, CA: Sage.

16.

Emery

R. E.

Waldron

Kitzmann

K. M.

Aaron

(1999). Delinquent behavior, future divorce or nonmarital childbearing, and externalizing behavior among offspring: A 14-year prospective study. Journal of Family Psychology, 13, 568-579.

17.

Erola

Härkönen

Dronkers

(2012). More careful or less marriageable? Parental divorce, spouse selection, and entry into marriage. Social Forces, 90, 1323-1345.

18.

Flinn

M. V.

Leone

D. V.

Quinlan

R. J.

(1999). Growth and fluctuating asymmetry of stepchildren. Evolution and Human Behaviour, 20, 465-479.

19.

Fritzell

Burström

(2006). Economic strain and self-rated health among lone and couple mothers in Sweden during the 1990s compared to the 1980s. Health Policy, 79, 253-264.

20.

Guo

(1993). Use of sibling data to estimate family mortality effects in Guatemala. Demography, 30, 15-32.

21.

Hetherington

E. M.

(Ed.). (1999). Coping with divorce, single parenting, and remarriage: A risk and resiliency perspective. Mahwah, NJ: Lawrence Erlbaum.

22.

Jockin

McGue

Lykken

D. T.

(1996). Personality and divorce: A genetic analysis. Journal of Personality and Social Psychology, 71, 288-299.

23.

Kurdek

Fine

(1993). The relation between family structure and young adolescents’ appraisals of family climate and parenting behavior. Journal of Family Issues, 14, 279-290.

24.

Lawson

D. W.

Mace

(2008). Sibling configuration and childhood growth in contemporary British families. International Journal of Epidemiology, 37, 1408-1421.

25.

Maier

E. H.

Lachman

M. E.

(2000). Consequences of early parental loss and separation for health and well-being in midlife. International Journal of Behavioral Development, 24, 183-189.

26.

McGue

Lykken

D. T.

(1992). Genetic influence on risk of divorce. Psychological Science, 3, 368-373.

27.

Morrison

D. R.

Ritualo

(2000). Routes to children’s economic recovery after divorce: Are cohabitation and remarriage equivalent? American Sociological Review, 65, 560-580.

28.

Omariba

D. W. R.

Rajulton

Beaujot

(2008). Correlated mortality of siblings in Kenya: The role of state dependence. Demographic Research, 18, 311-336.

29.

Pensola

T. H.

(2003). From past to present: Effect of lifecourse on mortality, and social Class differences in mortality in middle adulthood (Unpublished doctoral dissertation). University of Helsinki, Finland. Yearbook of Population Research in Finland, 39, Supplement.

30.

Rossi

A. S.

Rossi

P. H.

(1990). Of human bonding: Parent-child relations across the life course. New York, NY: Aldine de Gruyter.

31.

Rostila

Saarela

(2011). Time does not heal all wounds: Mortality following the death of a parent. Journal of Marriage and Family, 73, 236-249.

32.

Rowe

D. C.

(1994). The limits of family influence: Genes, experience, and behavior. New York, NY: Guilford Press.

33.

Saarela

Finnäs

(2009). Geographic ancestry and cause-specific mortality in a national population. Population Research and Policy Review, 28, 169-194.

34.

Saarela

Finnäs

(2011). Family origin and mortality: Prospective Finnish cohort study. BMC Public Health, 11, 385.

35.

Simons

R. L.

Chao

(1996). Conduct problems. In Simons

R. L.

(Ed.), Understanding differences between divorced and intact families: Stress, interaction, and child outcome (pp. 125-143). Thousand Oaks, CA: Sage.

36.

Statistics Sweden. (2007). Barn och deras familjer 2006 [Children and their families 2006]. (Demographic Report No. 4). Stockholm, Sweden: Author.

37.

Statistics Sweden. (2010, January). Antal hushåll efter hushållstyp 1, FoB75-FoB90 [Number of household according to household type 1, FoB75-FoB90]. Retrieved from http://www.scb.se/Pages/TableAndChart____26362.aspx

38.

Umberson

Chen

(1994). Effects of a parent’s death on adult children: Relationship salience and reaction to loss. American Sociological Review, 59, 152-168.

39.

Wadsworth

M. E. J.

Kuh

D. J. L.

(1997). Childhood influences on adult health: A review of recent work from the British 1946 National Birth Cohort Study, the MRC National Survey of Health and Development. Paediatric and Perinatal Epidemiology, 11, 2-20.

40.

Weiss

R. S.

(1984). The impact of marital dissolution on income and consumption in single-parent households. Journal of Marriage and the Family, 46, 115-127.

41.

Zenger

(1993). Siblings’ neonatal mortality risks and birth spacing in Bangladesh. Demography, 30, 477-488.