International Journal of Public Health (2020) 65:627–636 https://doi.org/10.1007/s00038-020-01371-w (0123456789().,-volV)(0123456789().,-volV) ORIGINAL ARTICLE Marital status, living arrangements, and mortality in middle and older age in Europe Pilar Zueras1 • Roberta Rutigliano2 • Sergi Trias-Llimós3 Received: 17 October 2019 / Revised: 2 April 2020 / Accepted: 6 April 2020 / Published online: 29 April 2020 Ó The Author(s) 2020 Abstract Objectives We study the role of marital status and living arrangements in mortality among a 50? population living in Europe by gender and welfare states. Methods Using data from waves 4, 5, and 6 of the Survey of Health Age and Retirement in Europe (n = 54,171), we implemented Cox proportional hazard models by gender and age groups (50–64 and 65–84). We estimated pooled models and separated models for two regions representing different welfare states (South-East and North-West). Results Among people aged 50–64, nonpartnered individuals (except never-married women) showed a higher mortality risk as compared with those partnered. Among the older population (65–84), divorce was associated with higher mortality among men, but not among women, and living with someone other than a partner was associated with higher mortality risk as compared to those partnered. In the South-East region living with a partner at ages 50–64 was associated with lower mortality. Conclusions Partnership and residential status are complementary for understanding the role of family dimensions in mortality. The presence of a partner is mortality protective, especially among 50–64-year-old men in South-East Europe. Keywords Mortality differences  Marital status  Partnership status  Living arrangements  Family systems  Welfare states  Europe Introduction constellations, marital status, and living arrangements in mid- and later life (Carr and Utz 2020; Esteve et al. 2020). Increasing life expectancy together with family changes There is an extensive and consistent body of the literature (e.g., rising divorce rates, cohabitation, nonmarital fertility, that finds a mortality advantage for married people. How- and solo living) is leading to growing diversity in family ever, less is known about the health outcomes associated with both living arrangements and household composition (see Hank and Steinbach 2018 for a review). Although Electronic supplementary material The online version of this living arrangements and marital status in both mid- and article (https://doi.org/10.1007/s00038-020-01371-w) con- tains supplementary material, which is available to autho- later life largely result from individual choices over the life rized users. course, institutional and normative factors also play a role (Pfau-Effinger 2005). However, mainly due to data limi- & Roberta Rutigliano tations, the variation in the association between family r.rutigliano@rug.nl structure and mortality across different welfare states is Pilar Zueras currently under researched (Hank and Steinbach 2018; pzueras@ced.uab.cat Requena and Reher 2020). This study contributes to the 1 Centre d’Estudis Demogràfics (a Member of the CERCA growing body of the literature examining the relationship Programme/Generalitat de Catalunya), Barcelona, Spain between family structure, living arrangements, and mor- 2 Population Research Centre, Faculty of Spatial Sciences, tality. The first contribution of the paper is the use of both University of Groningen, Groningen, The Netherlands marital status and living arrangements in a sample of 3 Faculty of Epidemiology and Population Health, London middle-aged and older Europeans. The second contribution School of Hygiene and Tropical Medicine, Keppel Street, is the measurement of the associations between family London WC1E 7HT, UK 123 628 P. Zueras et al. structure and mortality in two different European welfare advantage of older adults who have been living with a states. This study benefits from the Survey of Health, partner for a long time provide possible explanations for Ageing and Retirement in Europe (SHARE) data (Börsch- different risks of mortality by marital status (Hu and Supan et al. 2013), which allowed us to include family, Goldman 1990). Furthermore, similar lifestyles and edu- health, and socioeconomic variables in the analysis of cational and social homogamy between partners are mortality in and within Europe by distinguishing two big assumed to equally shape the risk of dying of both partners, regions with different welfare states: North-West and consequently explaining the elevated mortality among South-East. widowed people (Drefahl 2012; Manzoli et al. 2007; Marmot 2005). Despite the identification of the mecha- Background nisms linking marital status and mortality, the benefits of being married on survival persist even after adjusting for Survival at adult and later ages is positively influenced by socioeconomic factors, objective health, and smoking the quantity and the quality of both social relationships and behavior (Pijoan-Mas and Rı́os-Rull 2014). support (Holt-Lunstad et al. 2010; Shor et al. 2013). These findings suggest that other mechanisms—related Among the extended social net, family members, particu- to union formation—beyond selection, protection, and larly partners and children, play a crucial role in this homogamy, play a role in understanding the survival association (Giudici et al. 2019; Modig et al. 2017). Dif- advantage of marrieds. Given the diversity of partnership ferential mortality by marital status has largely been situations and family structures, it would be worth con- investigated and has shown the benefits of being married. sidering the role of other family dimensions. Living These benefits are greater among men than among women arrangements, household composition, and the number of and in younger than in older age groups (e.g., they decrease children offer a proxy for the type of social relationship, after the age of 65) (Franke and Kulu 2018; Hank and the family resources, and their association with differential Steinbach 2018; Hu and Goldman 1990; Murphy et al. mortality across population groups. For example, the 2007; Rendall et al. 2011; Shor et al. 2012; Staehelin et al. household size and its structure (e.g., the presence of 2012). Yet, gender differences between married and non- dependent children) reduce mortality variation across married people at older ages are less evident (Manzoli et al. partnership status only for men (Franke and Kulu 2018). 2007; Shor et al. 2012). Although research about difference Yet, living alone results in the highest risk of mortality in mortality between never-married, divorced, and wid- among middle-aged men, upholding the marriage advan- owed individuals is inconclusive, there is a consistent lit- tage hypothesis (Staehelin et al. 2012). A study for England erature that finds elevated mortality risks among divorced and Wales found that the health benefits of living with a people, especially among working age men (Hu and partner were largest for those aged 50–64 and lowest for Goldman 1990; Koskinen et al. 2007; Manzoli et al. 2007; those aged 65–85, and the differences in mortality persisted Murphy et al. 2007). when household size and presence of children were There is evidence that the benefits of marital status on accounted for Franke and Kulu (2018). health operate through different mechanisms such as health It is also known that parents live longer than nonparents selection, health behaviors, social and economic support, and that middle-aged childless women experience higher and social control. These mechanisms are found to be risks of mortality. Previous studies show a U-shaped rela- stronger among men (Bourassa et al. 2019; Drefahl 2012; tionship between the number of children and mortality risk, Manzoli et al. 2007; Umberson 1987). Previous studies and women with either low or moderate number of children have identified selection, protection, and homogamy show also lower risk of death (Doblhammer 2000). The effects of being partnered on mortality. At younger ages, mortality benefits derived from the number of children there is a positive selection of the wealthier and healthier were explained by nonexclusive biomedical and social individuals into marriage (Rendall et al. 2011) and, sub- mechanisms (Barclay and Kolk 2019; Doblhammer 2000; sequently, a negative selection out of union for those who Jaffe et al. 2009), including socioeconomic and health are less healthy or have risky behaviors (Umberson 1987). selection, risk-avoiding behavior associated with parent- Other studies focused on the protective effect of marriage, hood, and the potential supply of economic and social which discourages health-damaging behaviors (Franke and support at older ages provided by children (Barclay and Kulu 2018; Guner et al. 2014; Koskinen et al. 2007; Kolk 2019; Friedman and Mare 2014; Hank and Steinbach Murray 2000). In particular, smoking behavior is associ- 2018). ated with marital dissolution, and it partly explains earlier Overall, the effect of family support as one of the most mortality among separated and divorced mature and older beneficial for individuals’ survival appears to be consistent people (Bourassa et al. 2019). Thus, the combined effects across societies (Holt-Lunstad et al. 2010). However, most of selection into and out of marriage and the cumulative research about differential mortality by marital status or 123 Marital status, living arrangements, and mortality in middle and older age in Europe 629 living arrangements has a national scope and overlooked a longitudinal individual-level data of nationally represen- comparative perspective (see Hu and Goldman 1990; tative samples of population aged 50 or older across Eur- Murphy et al. 2007; Noale et al. 2005; Valkonen et al. 2004 ope, collected from 2004 onward (Börsch-Supan et al. for some exceptions). Therefore, understanding the inter- 2013). It includes information about a wide range of relation between partnership or residential status and socioeconomic, demographic, and health topics, as well as mortality across societal settings remains an open question information of deceased respondents from end-of-life (Hank and Steinbach 2018; Requena and Reher 2020). interviews. Specifically, little is known about the associations We selected individuals who were 50–84 years old between family structures and mortality across European when they were first observed and whose survival status in regions, as they are characterized by dissimilar family the subsequent wave was known (76.4% between wave 4 systems and welfare states. These regional contexts result and 5 and 80.0% between 5 and 6). We grouped our sample in different levels of intergenerational co-residence, social in two age groups: 50–64 and 65–84 (age at the beginning norms, attitudes toward family support, and public policies of the risk window). We excluded 1 221 individuals (Pfau-Effinger 2005). Regional norms can influence the (2.20%) with missing information on our main explanatory way in which social relationships, support, and care pro- variables (see details below). We ended up with a final vision are understood. Following this reasoning, a previous sample of 29 917 women and 24 254 men. study found positive associations between social vulnera- bility and mortality in Western and Southern European Variables countries but not in the Nordic ones (Wallace et al. 2015). Furthermore, a meta-analysis found different mortality risk The process time in our model was the age of individuals associated with widowhood across European regions (Shor expressed in months, and the outcome variable was time to et al. 2012). Overall, the health benefits associated with death. With this approach, we considered the effect of age family structure are found to be stronger in familialistic on mortality while not specifying any functional form for countries, i.e., South-East Europe, with less generous the hazard function (see Thiébaut and Bénichou 2004 for welfare states and where more people subscribe to norms further detail). Marital status (married or partnered, of strong family obligations, than in more individualistic divorced, never-married, and widowed) and living countries in the North-West, with more socially oriented arrangements (living with a partner in the household (either welfare states (Requena and Reher 2020). Because of these with or without others), living alone, and living with well-established differences in the role of family and someone other than a partner) were defined at the begin- welfare states across European regions, studying the role of ning of each person-month of the exposure to death. We both marital status and living arrangements in health out- included as covariates education and two time-varying comes is assumed to be particularly relevant. characteristics: smoking behavior and self-rated health (SRH). Aims The control variables were coded as follows. Educa- tional attainment: primary (ISCED 0–2), secondary This study examines the relationship between family (ISCED 3–4), or tertiary (ISCED 5–6). Smoking was coded structure, i.e., marital status, living arrangements, and as a binary variable. SRH was recoded into three cate- number of children, and mortality risk for middle-aged and gories: fair and poor, good, and very good and excellent. older European population by age group and gender. The We also included the number of children distinguishing first objective of this research is to account for the identi- between childless, having one child, and having two or fied selective, protective, and homogamy effects described more children. Finally, we created a control variable that above in the associations between family structures and grouped all countries into four regions: North (Sweden, mortality. The second objective of this research is to study, Netherlands, Denmark), West (Austria, Germany, France, for the first time, these associations from a comparative Switzerland, Belgium, and Luxembourg), South (Spain, perspective exploring potential differences within main Italy, and Israel), and East (Czech Republic, Slovenia, and welfare states in Europe. Estonia), which was used to adjust the first models. We, then, gathered those into two big regions: North-West and South-East to run the final models in parallel for different Methods family and welfare contexts. This study analyzed data from waves 4, 5, and 6 of the Survey of Health Age and Retirement in Europe (SHARE), collected in 2011, 2013, and 2015. SHARE is a biennial 123 630 P. Zueras et al. Statistical methods childless was associated with higher mortality risk among younger women as compared to those having two or more We performed separate Cox proportional hazard models children (HR 2.27, 1.41–3.66). for both genders and age groups (50–64 and 65–84) to Among those aged 65–84, divorce was associated with estimate the hazard ratios (HR) of mortality by marital higher mortality among men (M1 HR 1.35, 1.05–1.74), but status and living arrangements. We adjusted for the clus- not among women (M1 HR 0.86, 0.63–1.17), whereas tered structure of the data by computing robust standard being never-married was positively associated with mor- errors (Cleves et al. 2008). In a first set of models, we tality among women (M1 HR 1.22, 0.88–1.69). These focused on marital status and we ran nested specifications results remained steady after adding the controls. The to adjust by both sociodemographic characteristics and number of children was not associated with distinct mor- health variables (1). Formally, for each individual i of age t tality risk. (measured in months), the hazard of dying was modeled as follows: Associations between living arrangements hi ðtÞ ¼ h0 ðtÞ  exp xi;1 bi;1 þ xi;2 bi;2 þ xi;3 bi;3 and mortality  ð1Þ þxi;4 bi;4 þ xi;5 bi;5 Figure 2 shows the associations between living arrange- where hi ðtÞ was the hazard for individual i at age t; Xi is a ments and number of children, and mortality. For both men vector of covariates with coefficients b. Specifically, and women aged 50–64, those living alone showed excess starting from the null model (only marital status,xi;1 , we mortality (M1 HR 1.73, 1.30–2.30, and 1.41, 1.01–1.97, added educational level (xi;2  M2), geographical region respectively) that reduced after adjusting for the control (xi;3  M3), SRH and smoking behavior (xi;4  M4), and variables, while remaining significant for men (HR 1.53, the number of children (xi;5  M5). Finally, h0 ðtÞ was the 1.11–2.09), but not for women (HR 1.14, 0.80–1.63). Men living with others without a partner have higher mortality baseline hazard, i.e., the hazard when the vector Xi ¼ 0. risk as compared to those living with a partner (M1 HR In the second set of models, we focused on living 2.13, 1.36–3.34) and differences diminished after adjusting arrangements and followed the same nested specifications for SRH (M4 HR 1.65, 1.05–2.58). [see formula (1)]. We examined the relationship between Among the older population (65–84) living with others living arrangements and mortality both in the pooled without a partner was associated with higher mortality risk model, including all individuals from all countries, and in compared to living with a partner. HR decreased after separate models for different welfare states. adjusting for education, SRH, and smoking among men and for region and SRH among women, but slightly changed Results when controlled by the number of children (M5 HR 1.41, 1.04–1.89 among men and 1.47, 1.18–1.82 among women). Descriptive statistics Associations between living arrangements Table 1 shows descriptive statistics by age, sex, and region, and mortality by region stratified for our main explanatory variables (marital status and living arrangements). Figure 3 shows the region-specific associations between For each region-, age-, sex-specific combination, we living arrangements and number of children, and mortality. counted at least 86 deaths (North-West, 50–64, female) and Among population aged 50–64 living in North-West 151,773 person-months at risk (over 12,640 person-years). countries, there were no benefits of living with a partner Figure 1 shows the associations between marital status (Fig. 3). However, in the South-East region living with a and number of children, and mortality. Among individuals partner was found to be protective, especially among men aged 50–64, all nonmarried categories (except never-mar- in all model specifications (M5, HR for living alone 2.04, ried women) were associated with higher mortality risk as 1.40–2.97 and with others 2.10, 1.30–3.40). Childless compared with marrieds. These estimates were statistically women had higher mortality risk in both regions (North- significant among never-married and widowed men (HR West region HR 2.17, 1.19–3.98; South-East region HR 1.85, 95% CI 1.32–2.60 and 2.73, 1.71–4.37, respectively). 1.44, 0.74–2.78). Adding SRH reduced the HR, especially for never-married Among the older population, we found higher mortality men (which remained significant) and for divorced and risk among those who were living with others without a widowed women. In the fully adjusted model, which con- partner as compared to those living with a partner. After trolled for the number of children, only excess mortality of adjusting for the covariates these HRs remained similar and widowers remained high (HR 2.52, 1.59–3.99). Being significant only among women (M5, HR 1.64, 1.02–2.63 in 123 Marital status, living arrangements, and mortality in middle and older age in Europe 631 Table 1 Descriptive statistics for the dependent variables. Survey of Health, Ageing and Retirement in Europe, waves 4 (2011), 5 (2013), and 6 (2015) for Sweden, Netherlands, Denmark, Austria, Germany, France, Switzerland, Belgium, Luxembourg, Spain, Italy, Czech Republic, Slovenia, and Estonia Women Men 50–64 65–84 50–64 65–84 Person- n. N. Person- n. N. Person- n. N. Person- n. N. months deaths months deaths months deaths months deaths South-East Living arrangements Partner 158,105 5259 74 112,313 3929 218 12,937 4263 147 137,300 4774 605 Alone 28,802 868 25 69,402 2230 205 14,814 479 41 18,815 617 92 Other than 18,676 562 12 23,468 734 100 7641 251 19 5214 171 37 partner Marital status Partnership 154,400 5140 74 113,752 3962 231 124,054 4110 138 135,600 4728 598 Divorced 23,316 711 15 13,635 439 26 12,937 409 22 6050 202 31 Single 9979 321 5 7763 262 24 11,786 390 33 5833 206 27 Widow 17,888 517 17 70,033 2230 242 2996 84 14 13,846 426 78 Total sample 205,583 6689 111 205,183 6893 523 151,773 4993 207 161,329 5562 734 North-West Living arrangements Partner 188,459 6845 61 116,763 4330 131 160,673 5813 75 145,864 5406 317 Alone 45,858 1585 21 80,628 2733 141 30,957 1071 18 31,566 1099 95 Other than 16,895 592 4 8100 250 21 6684 228 2 2541 82 11 partner Marital status Partnership 180,683 6576 57 119,249 4381 145 152,166 5507 73 144,582 5346 320 Divorced 36,002 1262 16 19,579 689 20 22,692 786 11 11,711 419 32 Single 17,619 622 6 8736 306 17 18,808 666 6 7775 279 16 Widow 16,908 562 7 57,927 1937 111 4648 153 5 15,903 543 55 Total sample 251,212 9022 86 205,491 7313 293 198,314 7112 95 179,971 6587 423 the North-West and HR 1.34, 1.06–1.71 in the South-East). results show the importance of being partnered, especially Finally, the results suggest a protective effect of having an among middle-aged men. At older ages, living with only child in South-East region (HR 0.87, 0.72–1.06 among someone other than the partner is associated with higher men and HR 0.81, 0.65–1.02 among women). mortality as compared to living with the partner for both Overall, the direction of the association between both genders. Results for the two European welfare regions the socioeconomic and the health control variables, and examined suggest living with a partner to be mortality mortality was in line with our expectations: a mild negative protective in the South-East region and especially among association with education, a strong negative association middle-aged men, but not in the North-West. Furthermore, with SRH, and higher mortality of smokers as compared to older people living with others than the partner show nonsmokers. excess mortality in both regions. Our results on the excess mortality among middle-aged nonmarried individuals but not among older individuals are Discussion consistent with previous findings suggesting wider differ- ences among men than among women and among the This study examined how marital status and living working aged population than among the over 65 (Franke arrangements are related to (1) differences in mortality in a and Kulu 2018; Guner et al. 2014; Hu and Goldman 1990; sample of European adults aged 50 and over and (2) Koskinen et al. 2007; Manzoli et al. 2007; Murphy et al. potential variation across European welfare states. The 2007; Murray 2000; Rendall et al. 2011). In line with 123 632 P. Zueras et al. Fig. 1 Hazards ratio (HR) of Men 50-64 Women 50-64 mortality by marital status and Divorced Divorced by number of children. Survey Single Single of Health, Ageing and Retirement in Europe, waves 4 Widowed Widowed (2011), 5 (2013), and 6 (2015) Childless Childless for Sweden, Netherlands, 1 child 1 child Denmark, Austria, Germany, France, Switzerland, Belgium, 0 1 2 3 4 0 1 2 3 4 Luxembourg, Spain, Italy, Czech Republic, Slovenia, and Men 65-84 Women 65-84 Estonia Divorced Divorced Single Single Widowed Widowed Childless Childless 1 child 1 child 0 1 2 3 4 0 1 2 3 4 M1 M4 M5 M1: unadjusted; M4: educational level, region, self-reported health and smoking; M5: M4 plus number of children ref. marital status:in a partnership; number of children: two Fig. 2 Hazards ratio (HR) of Men 50-64 Women 50-64 mortality by living Alone Alone arrangements and by number of children. Survey of Health, Others Others Ageing and Retirement in Europe, waves 4 (2011), 5 Childless Childless (2013), and 6 (2015) for 1 child 1 child Sweden, Netherlands, Denmark, Austria, Germany, France, 0 1 2 3 0 1 2 3 Switzerland, Belgium, Luxembourg, Spain, Italy, Men 65-84 Women 65-84 Czech Republic, Slovenia, and Estonia Alone Alone Others Others Childless Childless 1 child 1 child 0 1 2 3 0 1 2 3 M1 M4 M5 M1: unadjusted; M4: educational level, region, self-reported health and smoking; M5: M4 plus number of children ref. living arrangements: partner in the house; number of children: two earlier research (Pijoan-Mas and Rı́os-Rull 2014), differ- confounders, especially SRH, except for widowed men ential mortality by marital status persisted even after (HR 2.73). Probably young widowers are a highly selected adjusting for socioeconomic, health status, and smoking group of more disadvantaged men in terms of health and behavior variables. This finding upholds the protection and SES who have not re-partnered despite their young age. In selection mechanisms related to marriage mortality addition to the protective effect of both marriage and advantage. Indeed, at ages 50–64, both widowed men and socioeconomic homogamy, the sharing living conditions of women, and never-married men experienced higher mor- couples (e.g., healthcare access, health literacy) could tality risk, which generally diminished after adjusting for explain the higher mortality risk among widowed men. 123 Marital status, living arrangements, and mortality in middle and older age in Europe 633 Fig. 3 Hazards ratio (HR) of Men 50-64 Women 50-64 North-West mortality by living Alone Alone arrangements and number of children by regions. Survey of Others Others Health, Ageing and Retirement in Europe, waves 4 (2011), 5 Childless Childless (2013), and 6 (2015) for 1 child 1 child Sweden, Netherlands, Denmark, Austria, Germany, France, 0 1 2 3 4 0 1 2 3 4 Switzerland, Belgium, Luxembourg, Spain, Italy, Men 50-64 Women 50-64 Czech Republic, Slovenia, and Estonia Alone Alone South-East Others Others Childless Childless 1 child 1 child 0 1 2 3 4 0 1 2 3 4 M4: educational level, self-reported health and smoking; M5: M4 plus number of children Men 65-84 Women 65-84 North-West ref. living arrengements: partner in the house; number of children: Alone Alone two Others Others Childless Childless 1 child 1 child 0 1 2 3 4 0 1 2 3 4 Men 65-84 Women 65-84 Alone Alone South-East Others Others Childless Childless 1 child 1 child 0 1 2 3 4 0 1 2 3 4 M1 M4 M5 M1: unadjusted; M4: educational level, self-reported health and smoking; M5: M4 plus number of children ref. living arrengements: partner in the house; number of children: two Among the oldest population, we found higher mortality divorced men (HR 1.35). A possible explanation for this risk for divorced men, but not for women, which contrasts gender difference among the oldest cohorts could be a with earlier findings (Koskinen et al. 2007; Manzoli et al. diverging selection by income. That is, men who divorced 2007; Murphy et al. 2007). Our results are consistent with and did not partner again might be less wealthy (negative the literature finding that there are more pronounced dif- selection), whereas women who divorced and stayed non- ferences in lifestyles between married and divorced men, partnered might be wealthier and could afford to live by who are more likely to engage in unhealthy and risky themselves (positive selection) (Shafer and James 2013). behaviors, as compared to these differences between In line with previous results, men aged 50–64 showed divorced and married women (Bourassa et al. 2019; the strongest benefits of living with a partner as compared Umberson 1987). However, adjusting for SRH, smoking to living alone or with someone other than a partner behavior, and education did not reduce the HR for older (Franke and Kulu 2018; Staehelin et al. 2012), while for 123 634 P. Zueras et al. women only living alone was associated with higher healthier lifestyles of parents compared to nonparents), mortality risk in the crude models but not in the adjusted whereas cross-national higher mortality of childless mid- models. Results show that the presence of a partner is more dle-aged women can be explained by biomedical factors critical for middle-aged men than for women, consistently (Barclay and Kolk 2019; Doblhammer 2000; Jaffe et al. with previous research (Koskinen et al. 2007; Staehelin 2009). et al. 2012) that suggest a central role of wives in men’s This study took advantage of the longitudinal aspect of social support and network connection. SHARE data to study the associations between marital At older ages, those living alone did not have different status, living arrangements, and differential mortality mortality risk as compared to those living with a partner, across different European contexts. As compared to vital especially in South-East Europe. In the North-West region statistics registers or linked censuses data, SHARE data the elevated mortality risk of those living alone reduces provide information on marital status and other family after adjusting for health status. This suggests that there is a resources, together with social, socioeconomic, and health positive selection of those living alone at older age, and measures. As for every longitudinal analysis attrition might that the level of independence and good health required to represent an important limitation. In our case, because of live alone may be higher in South-East than in North-West the relatively short follow-up period (around 2 years) we welfare states (Requena and Reher 2020). These results could follow [ 75% of the cases. Furthermore, those should be taken cautiously because recording mortality in individuals who drop-off of the sample do not report worse survey data seems harder among those people living alone health status (see Supplementary Table 1). As many other (Chatfield et al. 2005). However, this seems not to be the health surveys, SHARE sample is selected. However, the case in our sample (see Supplementary Table 1). share of institutionalized individuals is rather low at the age Furthermore, individuals living with someone other than groups of our interest (50–84); i.e., below 8% at age 80–84 a partner presented higher mortality risk (except for mid- in the countries under study (EUROSTAT). Nonetheless, dle-aged women) compared to those living with a partner. following a previous study (Solé-Auró et al. 2015) we Adjusting for controls, especially for SRH, reduced those compared SHARE mortality data with mortality register differences, which suggests that old people in those living data. This comparison resulted in SHARE mortality to be arrangements might have poorer health and functional slightly lower than population-level mortality (Supple- conditions (Requena and Reher 2020). In other words, mentary Fig. 1). Finally, to make sure that our sample individuals living with someone other than a partner might selection does not bias our results, we run some sensitivity be less capable of living alone. This latter finding holds analysis. Specifically, we checked the distribution of the also in the South-East, but not in the North-West region missing values over the sample and we re-run the models (except for older women), and could be related to social including a specific category for missing values for each norms and distinct regional patterns of co-residence with variable. Our results prove robust and the distribution of adult children and institutionalization (as discussed in the missing values does not highlight any selection bias. Requena and Reher 2020). Dependent individuals tend to live more at home (with relatives or caregivers) in Southern Conclusion and Eastern Europe, whereas these individuals are more likely to enter residential care in Western and Nordic The relationship between family structure and mortality in systems (EUROSTAT; Laferrère et al. 2013). Future different welfare states has not been studied before. research should further investigate whether differential Overall, our results confirm that, regardless of welfare associations between the family structure and mortality states, being partnered is associated with lower mortality, exist across European welfare states. especially among middle-aged men in South-East Europe. Finally, we expected to find evidence of social support Moreover, this advantage persisted after accounting for related to the number of children among the older popu- smoking behavior and socioeconomic and health status, lation (65–84). Having a child suggested a protective which are related to well-known selection and protection effect, consistent with higher involvement of only children mechanisms of being partnered. At older ages, living in parental care (Campbell and Martin-Matthews 2003), arrangements are more strongly associated with mortality but only in the familialistic welfare state. Previous studies than marital status, especially among men. This suggests did not find strong evidence on the importance of social that individuals who are not living alone nor with the support in explaining higher longevity of parents in Swe- partner show higher mortality risk as compared to those den (Barclay and Kolk 2019; Modig et al. 2017). However, who live with their partner. Furthermore, our results gender similarities of elevated mortality risk among the pointed at living arrangements and the number of children childless middle-aged population in the North-West sug- as important factors for mortality risk in South-East Eur- gest that protective social mechanisms are in action (e.g., ope, where the welfare system is less generous than in 123 Marital status, living arrangements, and mortality in middle and older age in Europe 635 North-West Europe and in which the well-being of older Campbell LD, Martin-Matthews A (2003) The gendered nature of individuals depends more on their family members. In sum, men’s filial care. J Gerontol Ser B: Psychol Sci Soc Sci 58:S350– S358 both marital status and living arrangements are comple- Carr D, Utz RL (2020) Families in later life: a decade in review. mentary variables in the complex associations between J Marriage Fam 82:346–363 family and mortality at middle and old age in Europe. Chatfield MD, Brayne CE, Matthews FE (2005) A systematic literature review of attrition between waves in longitudinal Acknowledgements This paper uses data from SHARE Waves 4, 5, studies in the elderly shows a consistent pattern of dropout and 6 (https://doi.org/10.6103/share.w4.700, https://doi.org/10.6103/ between differing studies. J Clin Epidemiol 58:13–19 share.w5.700, https://doi.org/10.6103/share.w6.700). The SHARE Cleves M, Gould W, Gutierrez R, Marchenko Y (2008) An data collection has been funded by the European Commission through introduction to survival analysis using Stata. Stata press, College FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006- Station 062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4- Doblhammer G (2000) Reproductive history and mortality later in CT-2006-028812), FP7 (SHARE-PREP: GA No. 211909, SHARE- life: a comparative study of England and Wales and Austria. LEAP: GA No. 227822, SHARE M4: GA No. 261982), and Horizon Popul Stud 54:169–176 2020 (SHARE-DEV3: GA No. 676536, SERISS: GA No. 654221) Drefahl S (2012) Do the married really live longer? The role of and by DG Employment, Social Affairs, and Inclusion. Additional cohabitation and socioeconomic status. J Marriage Fam funding from the German Ministry of Education and Research, the 74:462–475 Max Planck Society for the Advancement of Science, the U.S. Esteve A, Reher DS, Treviño R, Zueras P, Turu A (2020) Living National Institute on Aging (U01_AG09740-13S2, P01_AG005842, alone over the life course: cross-national variations on an P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, emerging issue. Popul Dev Rev 46:169–189. https://doi.org/10. IAG_BSR06-11, OGHA_04-064, HHSN271201300071C) and from 1111/padr.12311 various national funding sources is gratefully acknowledged (see EUROSTAT The Census Hub: 2011 Census Database. Eurostat. www.share-project.org). Available at https://ec.europa.eu/CensusHub2/ Accessed 9 Oct 2018 Franke S, Kulu H (2018) Mortality differences by partnership status Funding PZ was supported by the Spanish Ministry of Education in England and Wales: the effect of living arrangements or health Culture and Sports [CAS18/00152] and the Spanish Ministry of selection? Eur J Popul 34:87–118 Economy, Industry and Competitiveness, National R&D&I Plan Friedman EM, Mare RD (2014) The schooling of offspring and the [CSO2017-89721-R]. survival of parents. Demography 51:1271–1293 Giudici C, Polettini S, de Rose A, Brouard N (2019) Which aspects of Compliance with ethical standards elderly living conditions are important to predict mortality? The complex role of family ties at home and in institutions. Soc Indic Conflict of interest The authors declare that they have no conflict of Res 142:1255–1283 interest. Guner N, Kulikova Y, Llull J (2014) Does marriage make you healthier? IZA Discussion Paper No. 8633 Open Access This article is licensed under a Creative Commons Hank K, Steinbach A (2018) Families and health: a review. In: Attribution 4.0 International License, which permits use, sharing, Doblhammer G, Gumà J (eds) A demographic perspective on adaptation, distribution and reproduction in any medium or format, as gender, family and health in Europe. Springer, Cham, pp 23–39 long as you give appropriate credit to the original author(s) and the Holt-Lunstad J, Smith TB, Layton JB (2010) Social relationships and source, provide a link to the Creative Commons licence, and indicate mortality risk: a meta-analytic review. PLoS Med 7:e1000316 if changes were made. The images or other third party material in this Hu Y, Goldman N (1990) Mortality differentials by marital status: an article are included in the article’s Creative Commons licence, unless international comparison. Demography 27:233–250 indicated otherwise in a credit line to the material. If material is not Jaffe DH, Neumark YD, Eisenbach Z, Manor O (2009) Parity-related included in the article’s Creative Commons licence and your intended mortality: shape of association among middle-aged and elderly use is not permitted by statutory regulation or exceeds the permitted men and women. Eur J Epidemiol 24:9–16 use, you will need to obtain permission directly from the copyright Koskinen S, Joutsenniemi K, Martelin T, Martikainen P (2007) holder. To view a copy of this licence, visit http://creativecommons. Mortality differences according to living arrangements. Int J org/licenses/by/4.0/. Epidemiol 36:1255–1264 Laferrère A, Van den Hende A, Van den Bosch K, Geerts J (2013) Entry to institutional care: predictors and alternatives. In: Börsch-Supan A, Brandt M, Litwin H, Weber G (eds) Active ageing and solidarity between generations in Europe. First References results from SHARE after the economic crisis. De Gruyter, Berlin, pp 253–264 Barclay K, Kolk M (2019) Parity and mortality: an examination of Manzoli L, Villari P, Pirone GM, Boccia A (2007) Marital status and different explanatory mechanisms using data on biological and mortality in the elderly: a systematic review and meta-analysis. adoptive parents. Eur J Popul 35:1–23 Soc Sci Med 64:77–94 Börsch-Supan A, Brandt M, Hunkler C et al (2013) Data resource Marmot M (2005) Social determinants of health inequalities. Lancet profile: the Survey of Health, Ageing and Retirement in Europe 365:1099–1104 (SHARE). Int J Epidemiol 42:992–1001 Modig K, Talbäck M, Torssander J, Ahlbom A (2017) Payback time? Bourassa KJ, Ruiz JM, Sbarra DA (2019) Smoking and physical Influence of having children on mortality in old age. J Epidemiol activity explain the increased mortality risk following marital Community Health 71:424–430 separation and divorce: evidence from the English Longitudinal Murphy M, Grundy E, Kalogirou S (2007) The increase in marital Study of Ageing. Ann Behav Med 53:255–266 status differences in mortality up to the oldest age in seven European countries, 1990–99. Popul Stud 61:287–298 123 636 P. Zueras et al. Murray JE (2000) Marital protection and marital selection: evidence Solé-Auró A, Michaud P-C, Hurd M, Crimmins E (2015) Disease from a historical-prospective sample of American men. Demog- incidence and mortality among Older Americans and Europeans. raphy 37:511–521 Demography 52:593–611. https://doi.org/10.1007/s13524-015- Noale M et al (2005) Predictors of mortality: an international 0372-7 comparison of socio-demographic and health characteristics Staehelin K, Schindler C, Spoerri A et al (2012) Marital status, living from six longitudinal studies on aging: the CLESA project. Exp arrangement and mortality: does the association vary by gender? Gerontol 40:89–99 J Epidemiol Community Health 66:e22–e22 Pfau-Effinger B (2005) Culture and welfare state policies: reflections Thiébaut AC, Bénichou J (2004) Choice of time-scale in Cox’s model on a complex interrelation. J Soc Policy 34:3–20 analysis of epidemiologic cohort data: a simulation study. Stat Pijoan-Mas J, Rı́os-Rull J-V (2014) Heterogeneity in expected Med 23:3803–3820 longevities. Demography 51:2075–2102 Umberson D (1987) Family status and health behaviors: social control Rendall MS, Weden MM, Favreault MM, Waldron H (2011) The as a dimension of social integration. J Health Soc Behav protective effect of marriage for survival: a review and update. 28:306–319 Demography 48:481 Valkonen T, Martikainen P, Blomgren J (2004) Increasing excess Requena M, Reher D (2020) Residential status and health in middle mortality among non-married elderly people in developed and late life: a population-based study with new data from Spain. countries. Demogr Res 2:305–330 BMJ Open 10(1):e033330 Wallace LM, Theou O, Pena F et al (2015) Social vulnerability as a Shafer K, James SL (2013) Gender and socioeconomic status predictor of mortality and disability: cross-country differences in differences in first and second marriage formation. J Marriage the survey of health, aging, and retirement in Europe (SHARE). Fam 75:544–564 Aging Clin Exp Res 27:365–372 Shor E, Roelfs DJ, Curreli M et al (2012) Widowhood and mortality: a meta-analysis and meta-regression. Demography 49:575–606 Publisher’s Note Springer Nature remains neutral with regard to Shor E, Roelfs DJ, Yogev T (2013) The strength of family ties: a jurisdictional claims in published maps and institutional affiliations. meta-analysis and meta-regression of self-reported social support and mortality. Soc Netw 35:626–638 123