Marital Status as a Determinant of Life Expectancy and Wellbeing: The Case of Greece

Abstract

It has been proven that marital status affects health outcomes, with marriage often linked to greater longevity and wellbeing. However, while married individuals generally exhibit higher life expectancy, the ordering among other marital statuses (never married, divorced, widowed) can vary by gender and socio-cultural context. This study examines the evolving relationship between marital status and life expectancy in Greece over a 30-year period (1991–2021). Utilizing Hellenic Statistical Authority (ELSTAT) data specifically commissioned for this research, it constructs life tables by marital status, incorporating, for the first time in Greece, life tables for those in civil partnerships for 2021. While life expectancy improved across all marital statuses, married individuals consistently had the highest longevity, whereas those in civil partnerships are expected to live less than married individuals. Furthermore, widowers experienced a substantial increase in life expectancy, while by 2021, divorced males had the lowest life expectancy among men and divorced females showed the highest mortality rates at older ages among women. The relative position of never-married individuals improved over the period. Never-married women generally outlived never-married men, with this gap widening for the divorced. The most compelling finding is that the difference in mortality among family status categories appears to have diminished over time in Greece.

1. Introduction

Marital status has long been recognized as a significant socio-demographic factor influencing health outcomes, with studies consistently demonstrating associations between marriage on one hand, and improved longevity and wellbeing on the other. Married adults live longer and healthier lives, and they are characterized by improved mental health compared to their unmarried counterparts (Jensen 2022; Shor et al. 2012; Sorlie et al. 1995; Goldman and Hu 1993). At the same time divorced and widowed individuals also experience worse health outcomes than married ones. Yet, the sorting of marital status according to life expectancy is not clear cut. While married ones exhibit higher life expectancy at birth than every other marital status category, the sorting of life expectancy of never married, divorced and widowed is not stable. It depends on the gender and on the particular socioeconomic and cultural circumstances of the examined population and is also contingent on age (Jia and Lubetkin 2020).

Two primary mechanisms are often cited to explain why married individuals tend to live longer and healthier lives: marriage selection and marriage protection.

The marriage selection hypothesis posits that healthier individuals are more likely to marry in the first place. This perspective suggests that the observed health advantages in married people are not necessarily a direct result of marriage itself, but rather that those who are already healthier, both physically and mentally, are more attractive as partners and thus more likely to enter and remain in marital unions (Goldman and Hu 1993; Lillard and Waite 1995).

The marriage protection hypothesis, in contrast, argues that marriage itself provides benefits that lead to improved health and longevity. This perspective emphasizes the supportive and regulatory aspects of marital relationships. Protective factors include social support, social integration and regulation, economic benefits and caregiving. As far as social support is concerned, spouses often provide emotional, instrumental and informational support to each other, which can buffer stress and promote wellbeing. Regarding social integration and regulation, marriage can provide a sense of belonging, purpose and social control, which may reduce loneliness and promote healthier behaviors. Economic benefits include shared resources which can reduce financial stress and improve access to healthcare (Williams and Umberson 2004; Boyce et al. 2016; Gallagher and Waite 2000). The marriage protection hypothesis is reinforced by the fact that marital dissolution results in increased mortality risks, especially among males (Shor et al. 2012)

Both mechanisms claim their part in explaining the differential mortality rates of married, never married and “have been married” (i.e., widowed and divorced people). Yet, demographic studies find it difficult to determine the extent to which differential mortality by marital status is due to “marriage protection” or “marriage selection”.

Another research question that arises when one investigates marital status as a determinant of life expectancy and wellbeing is whether differences in mortality across different categories of marital statuses have remained stable over time or have undergone significant changes or even reordering. One of the few such published studies is that of Murphy et al. (2007), which compares seven european countries, namely Belgium, Finland, the Netherlands, England and Wales, France, Germany and Czech Republic. In most countries and age groups, the trends indicated an increase in the relative mortality difference between married and non-married people. The exceptions were related to the oldest age groups (80–84 and 85–89) in few countries (mainly males in Germany, England, Wales and inconclusively to Belgium). However, the study period of that research was from 1990 to 1999 and one could claim that it was not long enough to produce reliable trends in the long run.

In the case of Greece, there are two published research works concerning life tables by family status. The first is a paper by Kotsyfakis and Tsimpos, which refers to 1991 only (Kotsyfakis and Tsimpos 1997), and the second is an Msc dissertation, which refers to the period 2001–2011 (Pateraki 2018). The paper by Kotsyfakis and Thimpos found that married people of both sexes lived longer in 1991, followed by the divorced, the widowed and the never married in that order. The ordering is different in 2001 and 2011, in Pateraki’s dissertation. She found that never married did not exhibit the lowest life expectancy, except for the never married women in 2001. The widowed and divorced tended to occupy the last positions for both sexes in 2001 and 2011.

The purpose of this study is to investigate how the relationship between marital status and life expectancy in Greece has evolved between two points in time, 1991 and 2021, for both males and females. It also examines the specific trends in age-specific mortality rates across different marital status groups (never-married, married, in civil partnership, divorced, widowed) for males and females in Greece between 1991 and 2021, and what insights the observed differential mortality risks across marital status categories provide regarding the “marriage protection” and “marriage selection” hypotheses. Furthermore, this study investigates for the first time in Greek data the life expectancy and mortality rates of those in civil partnerships and explores why they are different from their married counterparts in 2021.

2. Results

Figure 1 shows life expectancy at birth (e₀) by marital status for males in Greece at two points in time: 1991 and 2021. A general observation is that male life expectancy over this 30-year period presented a noteworthy improvement across all marital-status categories. Married men consistently had the highest life expectancy in both years (e₀ = 74.8 in 1991 and 78.0 in 2021), while widowers had the lowest life expectancy in 1991 (e₀ = 65.2 years), and divorced men had the lowest life expectancy in 2021 (e₀ = 71.6 years). Never-married males had consistently lower life expectancy than both the general population and the married ones in both years, but they were expected to live longer than the widowers in 1991 and longer than both the divorced and the widowers in 2021.

What is noteworthy in Figure 1 is the spectacular improvement in e₀ for the widowers between 1991 and 2021. They exhibited an increase of 9.1 years (from 63.6 to 72.7), while at the same period e₀ for the general population increased “only” by 2.8 years (from 74.6 to 77.4).

Analysis of the age-specific probabilities of death offers further insight into the substantial gains in life expectancy observed among widowers (Figure 2 and Figure 3). Specifically, in 1991, widowed males exhibited markedly elevated probabilities of mortality up to the 40–44 age group when compared to their never-married counterparts. Conversely, at older ages, particularly beyond 60 years, widowers demonstrated significantly lower probabilities of death than never-married males in 1991, although their life expectancy at birth was 3.6 years lower than the never married. This pattern had changed in the 2021 data (Figure 3, whereas widowers had higher probabilities of dying in almost every age group (the age groups 50–54 and 55–59 being the exceptions) but the disparity in life expectancy at birth between widowers and never-married males had narrowed considerably to 0.5 years, down from 3.6 years in 1991 (Figure 1). This convergence in longevity can be attributed to a shift in the age distribution of widowers. In 2021 there were proportionally fewer widowers at young ages compared to 1991. Consequently, the relatively few widowers under the age of 40 in 2021 exerted a diminished influence on the overall life expectancy at birth for this marital status category.

A noteworthy finding is that those in civil partnerships live less than the married ones, but more than the never married. This is a novel observation, as the institution of civil partnership is relatively recent in Greece. It was introduced in November 2008 under Law 3719/2008. The difference in life expectancy between married men and those in civil partnerships may reflect a stronger degree of selection effect into marriage versus civil partnership in the Greek context. Marriage is a long-standing institution often associated with greater tradition, perceived stability and higher levels of social and familial expectation. Civil partnerships, being a newer option, might attract a different demographic group that does not fit this traditional mold (Figure 4a–c). Figure 4a,b confirm that civil partnership is preferred by younger people. However, it was the old males (aged 80–84) in civil partnership that exhibited unusually high probabilities of dying in 2021 (Figure 3), while this was not the case with their female counterparts (Figure 5). It may be the case that males who entered a civil partnership at old age are more “fragile”, health wise, than their same-age counterparts in the general population. Another explanation could be the small sample size—only 77 men in this group, with 10 deaths—which makes the data more susceptible to random variations.

Moreover, Figure 4c reveals that in Greece many more men than women enter in civil partnership at old ages (65+). A plausible hypothesis suggests that these partnerships may frequently involve older men and significantly younger women. From a critical perspective, such arrangements could be analyzed through the lens of strategic resource allocation and intergenerational transfer. The younger female partners may be motivated by the prospect of acquiring the male partner’s assets upon his demise, thereby securing their economic future. Further research employing qualitative methodologies, such as in-depth interviews, alongside quantitative analysis of partnership durations and asset transfers, would be crucial to fully understand the motivations and implications of these civil partnerships.

Similarly to the chart for males, female life expectancy (e₀) increased in all categories between 1991 and 2021 (Figure 6). Married women consistently showed the highest life expectancy. Never-married females showed the lowest e₀ both in 1991 and in 2021. However, in 2021 the difference in e₀ between never-married females and the rest of the non-married categories (divorced and widowed) was marginal (0.3–0.9 years).

As far as females in civil partnership are concerned, in 2021 they recorded lower life expectancy than their married counterparts, but higher than the rest of family status categories (Figure 6), as was the case for males. Yet, in the case of females the difference in e₀ between married and those in civil partnership was negligible (0.3), while for males this difference was four full years (78.2 for married versus 74.2 for those in civil partnership).

Figure 7 focuses on 1991 and highlights the difference in life expectancy between sexes for each marital status category. The magnitude of the sex differential in life expectancy varied considerably depending on marital status. The largest gap was recorded in the widowed individuals, with females living 12.3 years longer than males on average. The smallest gap was recorded in divorced (5.1 years) and married (5.3 years) individuals. The difference for never-married individuals was intermediate (6.4 years). Taking into account that the difference between males and females in e₀ for the general population was 5.1 years, these results imply that widows and spinsters can cope better with life difficulties than widowers and bachelors. This is particularly intonated in the case of the windowed with a sex-differential gap in e₀ of over 12 years, suggesting that widowhood had a strong negative association with longevity for men at that time.

Similar are the results for 2021, as far as the sex-differential mortality by family status is concerned (Figure 8). Women without a husband (either never married, or divorced, or widowed) are better off than men without a wife. While the difference between males and females in e₀ for the general population was 5.1 years in 2021, for the never married the difference was 6.5 years, for the divorced it was 7.8 and for the widowed it was 8.9, always favoring the females. Compared to 1991 the sex differential in life expectancy has widened for the divorced, while it has narrowed for the widowed; yet the female advantage in e₀ in these two categories (widowed and divorced) in 2021 was larger than it was for the general population. A greater advantage in life expectancy is also exhibited by females in civil partnerships, who outlive their partners by 8.7 years.

Figure 2, Figure 3, Figure 5 and Figure 9 illustrate the probabilities of dying by age and marital status for men and women. Including the variable age in the analysis reveals mortality patterns by marital status that are not apparent when looking at only e₀. In 1991 the never married males faced the highest probability of dying from the age of 45 and above (Figure 2) and remained significantly above the rest of family status categories until their 80s, although their life expectancy at birth was higher than that of widowers (Figure 1). This is because the widowers had higher probabilities of dying in the young ages (below the age of 35), as can be seen in the relevant life tables. As a result, widower’s life expectancy at birth was a little lower than that of never married, but at ages 30 and above widowers presented higher life expectancy than the never married. In simple words, widowhood exerts more pernicious effect in males if it occurs at a young age.

In 2021 never married males was not the most disadvantaged group anymore. They exhibited lower mortality rates in all but two age groups (50–54 and 55–59) compared to the widowers and the divorced ones (Figure 3). Their life expectancy at birth is also higher than those who lost their wives either to death or to divorce (Figure 1).

Never-married women also improved their relative position in 2021 compared to 1991 (Figure 5 and Figure 9). In 2021 they exhibited higher mortality risks than every other category at the ages 40 to 64 but lower mortality than the divorced ones at the ages 65 and above and also lower than the widows in age group 80–84. However, their life expectancy at birth remained the lowest of all family status groups, as it was in 1991, though the difference is smaller than in 1991 (Figure 6).

It is noteworthy that men in civil partnerships tend to have higher probabilities of dying than the married ones at every age group but lower than the single ones at most age groups (Figure 2). However, the likelihood of death for those in civil partnership was higher than every other family status category at ages above 80 years. These are people who chose to get into a civil partnership at old age (roughly in their 70s, since the institution of civil partnership had been recently introduced in Greece in 2008) and maybe their health was already fragile when they decided to enter into a civil partnership. This is not the case with females in civil partnership, however, since they do not exhibit unusually high mortality in old ages (Figure 5).

3. Discussion

This study provides a comprehensive analysis of the evolving relationship between marital status and mortality in Greece, examining significant shifts between 1991 and 2021. By presenting both overall life expectancy at birth (e₀) and age-specific probabilities of dying (_nq_x) for males and females across different marital statuses, the paper highlights crucial demographic and societal transformations. A predominant finding across both genders and all marital statuses is that life expectancy at birth improved significantly, increasing by three to six years between 1991 and 2021. Additionally, the likelihood of dying at any specific age decreased over the same period. This substantial gain in longevity reflects broader advancements in healthcare, living conditions and public health infrastructure in Greece over the three decades. The slight dip in male life expectancy between 2011 and 2021 was likely due to the COVID-19 pandemic, which caused a significant death toll among men around 2021. However, this pandemic episode has been overcome and the population as a whole is living longer and healthier lives than ever before, a testament to socioeconomic progress.

Consistent with the established literature and reinforcing the “marriage protection hypothesis” outlined in the introduction, married individuals (both males and females) consistently exhibit the highest life expectancy at birth and, for the most part, the lowest probabilities of dying at various ages across the study period. This pattern underscores the enduring benefits of marital union, likely encompassing enhanced social support, shared resources, healthier lifestyles and mutual caregiving, which collectively contribute to improved longevity (Rendall et al. 2011; Williams and Umberson 2004). Nevertheless, in 2021 the disparity in the probabilities of dying between married and never-married individuals attenuated at very old ages (80–84) for both sexes. This partly aligns with earlier research, which found that the protective effect of marriage decreases with age and vanishes after age 85 (Jia and Lubetkin 2020; Rendall et al. 2011). A crucial novel aspect of this study is the inclusion of individuals in civil partnerships for the first time in the Greek context. The findings reveal that for both sexes in 2021, those in civil partnerships exhibited lower life expectancy than married individuals but higher than never-married individuals. This observation, especially the notable difference for males (a four-year gap compared to married men), suggests a potential selection effect where individuals entering civil partnerships may, on average, possess different socioeconomic profiles or health behaviors than those choosing marriage. Nevertheless, our data revealed that a disproportionately higher number of men aged 65 and above are entering civil partnerships, compared to women in the same age bracket, and this fact warrants further socio-legal investigation. It is plausible that these partnerships often involve older men and much younger women. From a critical viewpoint, such arrangements might be understood as a strategic allocation of resources and an intergenerational transfer. The younger female partners could be motivated by the potential to inherit the man’s assets upon his death, thereby securing their financial future. However, given the relatively recent introduction of civil partnerships in Greece (2008), further longitudinal research is needed to draw more concrete conclusions regarding their long-term health outcomes and to ascertain whether the initial observed differences diminish or persist over time. While the advantage of marriage is firmly established, the relative mortality risks among non-married groups (never-married, divorced and widowed) have undergone significant reordering, differing notably by gender.

One of the most striking findings for males is the “spectacular improvement” in life expectancy for widowers, increasing by 6.5 years between 1991 and 2021 (from 65.2 to 71.7) compared to the general population’s 2.8-year increase. This is primarily attributed to a shift in the age distribution of widowers in 2021, with proportionally fewer widowers at young ages whose high early mortality previously depressed overall life expectancy at birth. This suggests that young widowhood is becoming continuously less common. However, widowers continue to exhibit the lowest life expectancy among all male marital statuses.

Another noteworthy finding is the improved relative position of the never married males. In 1991, never-married males faced the highest mortality probabilities among marital statuses at age 45 onwards. However, by 2021, they demonstrated a significantly improved relative position, exhibiting lower mortality rates than widowers and divorced men across most adult age groups. Their life expectancy at birth consequently surpassed that of divorced (it was already higher than that of widowers since 1991) from 2001 onwards. This suggests potential changes in the socioeconomic profile or lifestyle choices of never-married men in Greece, or improvements in social support networks enabling them to mitigate historical disadvantages.

As far as divorced males are concerned, in 2021 they retained the second-lowest life expectancy after widowers all over the period 1991–2021. This indicates that the consequences of divorce are detrimental for male longevity.

Concerning the females, the paper reaffirms that the effect of widowhood on females appears less detrimental than on males. In 1991, widowed women already had lower mortality and higher life expectancy than never-married women across all age groups, a pattern that largely persisted in 2021. This may be due to stronger social networks, greater resilience or a selection effect among women who survive their spouses.

The analysis of sex differentials reveals the extent to which family status disproportionately affects the longevity of each sex. While women generally live longer than men in the general population, this advantage is significantly amplified for unmarried women compared to unmarried men. In 1991, the largest sex gap in e₀ was observed for widowed individuals. However, by 2021, the largest sex gap in e₀ for divorced individuals had dramatically increased (from 5.1 years to 7.8 years), highlighting a substantially greater survival advantage for divorced women over divorced men. This suggests that women without a husband (either never married, or divorced, or widowed) are better off than men without a wife in terms of life expectancy, and this disparity has widened for divorced individuals.

The observed trends likely stem from a combination of factors, and this study cannot definitively disentangle marriage selection from marriage protection. However, it is insightful that the difference in mortality among family status categories appears to have diminished over time in Greece. This is especially felt if one compares the “never married” category with the “married” one. Single men are no longer (in 2021) the category with the highest mortality and the lowest life expectancy, as was the case in 1991; also, the difference in age-specific mortality rates has diminished over the 30-year study period. This fact comes in contraditions with the findings of Murphy et al. (2007), which established that in six out of seven northwest European countries the mortality advantage of married individuals increased during the 1990s. In Greece this advantage was decreasing continously from 1991 to 2021 for both sexes.

In conclusion, it is important to acknowledge the limitations of this study. The main limitation is that the aggregate nature of the life table data prevents deeper causal analysis. The study lacks information on crucial confounding variables such as socioeconomic status (e.g., income, education level), health behaviors (e.g., smoking, diet, physical activity), pre-existing health conditions, or specific causes of death. These factors are vital for fully explaining the observed disparities. Furthermore, the duration of marital status is not accounted for, meaning the mortality risks for a recently divorced (or widowed, or married, or in civil partnership) individual are averaged with those divorced (or widowed, or married or in civil partnership) for many years.

This study clearly illustrates that while Greece has achieved substantial progress in overall female and male longevity, the gains have not been distributed uniformly across all marital statuses. Marriage continues to exert a protective effect in 21st century Greece, despite the destabilization of its institution. This finding suggests the need for targeted social and health policies designed to support vulnerable non-married populations, especially older divorced women and men. This includes strengthening social safety nets, promoting community integration and ensuring equitable access to healthcare and mental health services, particularly for those facing life transitions such as divorce or widowhood. By addressing these disparities, Greece can ensure that the benefits of improved longevity are shared more equitably across its population.

4. Data and Methods

Biostatistics by family status (that is age and sex specific mortality rates, survival probabilities and life expectancies) have been derived from life tables. Data for the construction of life tables by family status have been taken from ELSTAT (the former National Statistical Service of Greece, now renamed into Hellenic Statistical Authority) upon request. More specifically, deaths by age, sex and marital status (which are not published by ELSTAT) had to be ordered for the years 1991, 2001, 2011 and 2021 (ELSTAT 2025). These deaths were subsequently associated with the population by age, sex and marital status for the same years (ELSTAT 2024, n.d.) so that life tables by marital status could be derived. For the construction of the life tables, the central death rates (_nm_x) were transformed to probabilities of death (_nq_x) based on the Keyfitz method, as is described in Papadakis and Tsimpos (2004, pp. 456–58). For individuals under the age of 20, mortality rates from the general population were applied uniformly across all family statuses. This approach was adopted due to the negligible or non-existent number of individuals and, consequently, deaths within specific family status categories (excluding the never-married group) for this age range. An exception to this approach is the life tables that have undergone smoothening procedure. In these life tables, which are indicated by a footnote, mortality rates may differ from the general population even for the young ages.

However, in some of these initial life tables, age-specific mortality rates (_nm_x) were not increasing monotonically from the age of 10 onwards, as should be the case in modern populations in a peaceful period (Coale et al. 1983). This inconsistency affected the life tables of widows and widowers across most years, and the life table of females in civil partnerships in 2021. The root cause of this discrepancy was the limited number of deaths within certain age groups for these demographic segments. Therefore, the initial inconsistent life tables were treated to conform to the UN General model life table pattern (UN 1982). More particularly, the non-inconsistent _nq_x values from the initial life tables were input into a best-fit application of the software MORTPACK 4.3 and predicted _nq_x values, absolved from inconsistencies, were derived. A summary of all life tables for 1991 and 2021 can be seen in Appendix A. Life tables that have undergone a smoothening procedure, based on one component fit as described above, are noted in Appendix A, while for 2001 and 2011 only summary values are presented because our focus is on the two years that define the study period, mainly 1991 and 2021.

Another methodological challenge emerged from the mismatch between the family status categories of the census and those in civil registers. Greek censuses from 2001 onwards include “Separated” as a distinct category. However, the “Separated” category is absent in the civil registration of deaths as it does not constitute a legally recognized family status. “Separated” are those who are legally married, but they do not live together, and they are not divorced. Usually, separation is a step before divorce. It is plausible that a portion of those who declared “separated” in the census had applied for divorce and were waiting for the divorce to get issued. On the other hand, in the death registers only legal categories are mentioned. Consequently, individuals who are separated and die prior to the issuance of a divorce decree will be recorded as “married” in mortality statistics. The time required for a divorce in Greece in 2021 largely depended on whether it was a consensual divorce or a contested divorce. The first could be completed in approximately 2–3 weeks (as per Law 4509/2017), while contested divorce can take from 8 to 12 months, but in some cases, it can extend to two years or even more. The 2021 Population-Housing Census in Greece was conducted from the end of October until February 2022 (although the reference day was 22 October 2021). Therefore, those who declared “separated” in the census were most likely still alive by the end of 2021. Hence, in this paper we allocated the “separated” as married. The same treatment was followed for the “separated” in 2011 and 2001 (in the 1991 census the category “Separated” did not exist), although these censuses took place in springtime (18 March 2001 and 10–24 May 2011), and some of those enumerated as “Separated” in the census could have been divorced by the end of the year. This treatment was based on the fact that before the implementation of Law 4509/2017 (which came into force in December 2017 and drastically streamlined consensual divorces), the divorce process in Greece, both consensual and contested, generally took significantly longer (8 months to 1 year for consensual divorces and much more for the contested ones).