Repartnering After Widowhood (2024)

,

Article history

Abstract

A growing proportion of individuals are reentering the dating and marriage market in later life because of population aging and the increase in union dissolution (Carr, 2004b; Martin-Matthews, 2011). There is a large literature on union formation, but our knowledge of this topic is based largely on entrance into first unions and remarriage among young and middle-aged adults. Little is known about marital and nonmarital union formation in later life. The findings from these early studies cannot be generalized to older people, however, because the personal motivation and structural opportunities for finding a mate change across the life course (Brown, Bulanda, & Lee, 2012; Bulcroft, Bulcroft, Hatch, & Borgatta, 1989; Carr, 2004b; Vespa, 2013). The routes into repartnering in later life include divorce and widowhood. Our focus is on repartnering after widowhood in Canada. The absolute size of the widowed population in Canada has doubled since 1971 and is projected to further increase as the Baby Boom generation ages (Martin-Matthews, 2011; Statistics Canada, 2012).

Understanding differences in repartnering is important because repartnering can compensate for the losses of instrumental and emotional support that associate with widowhood and protect the widowed from loneliness and social isolation as well (Carr, 2004a; Connidis, 2010; De Jong Gierveld, 2004). Utz, Carr, Nesse, and Wortman (2002) demonstrate that the negative consequences of widowhood can also be offset through interaction with friends, relatives, and other social contacts, but the authors observe that not all widowed people have the resources to increase their social participation enough to compensate for the loss of a spouse. Moreover, Stroebe, Stroebe, Abakoumkin, & Schut (1996) show that, although important for coping with bereavement, support from family and friends cannot alleviate emotional loneliness or replace all the support that a spouse provides. De Jong Gierveld (2002) argues that spouses tend to be the best sources of long-term emotional and instrumental support and that spousal relationships provide the greatest opportunities for social integration. Spousal relationships provide the proximity between individuals (coresidence), interpersonal commitment, and shared interests that define companionship and ensure the exchange of support.

The purpose of this study is to estimate the sex-specific hazard rates of cohabitation and remarriage after widowhood among Canadians aged 45 and older. The literature on repartnering after widowhood is sparse, dated, and focuses on the antecedents of remarriage. The neglect of nonmarital cohabitation is a major gap in the literature. There are good reasons for considering cohabitation as a competing choice to remarriage. First, the prevalence of cohabitation among Canadians aged 50 and older has tripled since 1991, and the rate of cohabitation is growing faster among older adults than younger adults (Brown, Lee, & Bulanda, 2006; Chevan, 1996; Statistics Canada, 2012). Second, cohabitation appears to be an alternative to marriage in later life and is perhaps a more preferable option than remarriage for repartnering among older people (Brown et al., 2012). Our examination of both cohabitation and remarriage provides insight into how widowhood and the characteristics of the widowed associate with different choices and opportunities for repartnering.

Background

De Graaf and Kalmijn (2003) remark that first marriage is primarily a question of “when” since most people get married, but repartnering is a question of “if” because a much smaller proportion of people form a second union, especially if their first union ended in later life. There is no theory that specifically addresses union formation in later life or after widowhood, but marriage market theories offer a useful approach for explaining repartnering differentials among the widowed (Moorman, Booth, & Fingerman, 2006). Marriage market theories focus on the macro-level factors of union formation and refer to the opportunities and constraints that individuals encounter when searching for an eligible spouse (Guzzo, 2006; Moorman et al., 2006). However, life-course stage and union history influence the reasons (needs, goals, desires) for partnering and also the choice between marriage and cohabitation (Carr, 2004b; Davidson, 2001; De Graaf & Kalmijn, 2003). Hence, our conceptual approach adapts marriage market theories to account for the reasons and opportunities for repartnering after widowhood and in later life.

The general premise of marriage market theories is that people search for a spouse within a defined (local) area (Becker, 1981; Guzzo, 2006). Their success in finding a spouse depends on the conditions of the local marriage market and the personal characteristics that determine their attractiveness (bargaining power) in the marriage market. Moorman and colleagues (2006) observe that marriage market theories approach the search for a spouse in similar terms as searching for a job in labor markets. That is, marriage markets are proposed to function under the principles of supply and demand. Finding a spouse is relatively easy in marriage markets that have a plentiful supply of eligible mates and when the person searching for a mate has personal characteristics (e.g., physical beauty, human capital, wealth) that are in demand. In contrast, the chances of finding a partner decline as the number of potential mates in the marriage market decreases and also for individuals who lack in-demand characteristics.

The widowed have fewer opportunities for finding a suitable spouse than the never married. The probability of marriage (or cohabitation) depends on the supply of available members of the opposite sex (Guttentag & Secord, 1983). This factor is a demographic constraint on the chances of repartnering among the widowed, especially women. About 2% of widows and 20% of widowers get remarried (Smith, Zick, & Duncan, 1991). The low rate of remarriage among the widowed reflects age-graded opportunities for finding a spouse. Given that most people prefer to partner with someone similar in age, the marriage market is restricted for the widowed because the supply of single people at older ages is smaller (De Graaf & Kalmijn, 2003). Beyond this structural constraint, older people have fewer social opportunities to meet potential spouses, which is an additional barrier to their repartnering chances. The opportunities to meet spouses are tied to social networks and social participation, which tend to contract with older age, decreasing older people’s exposure to the marriage market (Carstensen, Fung, & Charles, 2003; De Graaf & Kalmijn, 2003; Lee & Markides, 1990).

The experience of widowhood is sex- and age-selective. More than 80% of the widowed in Canada are women and a similar percentage are aged 65 and older (Statistics Canada, 2012). The higher mortality rate of men contributes to an imbalanced sex ratio at older ages (Carr, 2004b). The relative shortage of men on the marriage markets translates into a steep gender difference in the probability of repartnering after widowhood. The likelihood of remarriage is about five times higher for widowers than widows (Connidis, 2010). This imbalanced sex ratio could also have implications for women’s choice of union after widowhood. Guzzo (2006) argues that, in circ*mstances where men have greater opportunities for forming a union than women, women may have to “settle” for cohabitation if their desire for marriage is a deterrent to men. She demonstrates that couples are more likely to cohabit than marry in areas with a shortage of men. How an imbalanced sex ratio influences women’s repartnering choices after widowhood is unknown, however, as Guzzo’s analysis is restricted to first unions among women aged 18–44.

The choice to repartner depends on the perception that the net benefits of forming a union are greater than the benefits of remaining single (Carr, 2004b; Sweeney, 1997; Vespa, 2013). Becker’s (1981) economic approach to marriage markets views potential spouses as “trading partners” who benefit the most from marriage when men specialize in paid labor and women in domestic labor, with the couple exchanging (trading) these services to take advantage of economies of scale and comparative advantage. Carr (2004b) observes that the gains to such economic exchange are weaker in later life, however, because these are a function of needs earlier in the life course. At later ages, gender roles become blurred as most people are retired from the labor force and no longer have regular childcare responsibilities. Carr writes that, although economic gains to trade are not a promising explanation for understanding repartnering among the widowed, the basic assumption of rational choice theories are still applicable because the choice to repartner is still a function of needs and the perception that there is a greater gain to forming a union than remaining single. Besides economic services, couples also exchange emotional and instrumental support and provide one another companionship, and repartnering decisions reflect the desire for these positive benefits.

Our analysis considers several factors that associate with the chances of repartnering among the widowed: gender, age, socioeconomic status, and culture. These factors could also correlate with the choice between marriage and cohabitation. With the rise of cohabitation, the perceived costs and benefits of marriage are no longer weighed simply against the costs and benefits of remaining single. Little is known about cohabitation after widowhood, but recent studies suggest that cohabitation serves different functions across the life course (Brown et al., 2012; King & Scott, 2005). Vespa (2013) observes that the net benefits of conjugal relationships change across the life course, and economic disincentives to marriage could make cohabitation an attractive alternative in later life. Cohabitation could be a preferable choice for repartnering among the widowed because it provides many of the benefits of marriage (e.g., emotional and instrumental support, companionship), but it also allows couples to avoid some of the costs of marriage because it is more flexible with regard to formal roles and the pooling of economic resources (Chevan, 1996; De Jong Gierveld, 2004; Vespa, 2013).

Gender

This study conducts sex-specific analysis for two reasons. First, as noted previously, widowers have higher chances of repartnering than widows. The consequences of demographic constraints on the repartnering chances of widows are compounded because women are also less likely to select spouses from younger age groups (Brown et al., 2012). Second, the gains to marriage could be sex-specific, which could influence both gender differences in the overall repartnering rate and the rate of remarriage versus cohabitation. Inequality in gender roles could create differential incentives for repartnering at older ages. Remarriage could be unattractive for older women who wish to benefit from their new-found independence and avoid the costs of traditional gender roles and caregiving responsibilities (Davidson, 2001; De Jong Gierveld, 2002). In contrast, the need for instrumental support is an incentive for remarriage among widowers, especially among those who were dependent on their late wives for domestic work (Carr, 2004b).

Age

Our analysis considers two age-related variables that could associate with repartnering after widowhood: age at widowhood and age at marriage. The probability of remarriage declines with age because the marriage market (supply of singles) is smaller at older ages (Smith et al., 1991; Wu, 1995). How age influences cohabitation after widowhood is unknown, but it is reasonable to assume that the demographic constraints that decrease the probability of remarriage also decrease the probability of cohabitation. However, previous research demonstrates that cohabitation is a more prevalent choice than remarriage after union dissolution (from divorce and widowhood combined) in later life (Brown et al., 2006; Connidis, 2010). Age at marriage is an important repartnering differential because women who married at a young age could have greater economic and social incentives to get remarried (Wu & Balakrishnan, 1994). Women who get married at a young age tend to specialize in domestic roles and have less human capital than women who spent time in the job market before getting married. This could pressure them into finding a new spouse to replace the human capital that their deceased husbands provided.

Socioeconomic Resources

As noted previously, the economic incentives for getting married are dependent on life-course stage. At earlier stages in the life course, socioeconomic resources have a positive effect on the probability of marriage and a negative effect on the probability of cohabitation (Brown et al., 2012). That is, economic well-being (e.g., high education, stable income) increases the likelihood of marriage for men and cohabitation is selective of men with fewer socioeconomic resources. For older people, socioeconomic status appears to have different effects. Chevan (1996) demonstrates that low income also selects older people into marriage. This effect on repartnering choices could be gender-specific because low-income status could push widows into repartnering, whereas low-income status makes widowers unattractive candidates on the marriage market. What is inconsistent with union formation in the younger population is the effect of economic well-being on the decision to get married or cohabit. Economic resources could be a disincentive to getting married among older people (Chevan, 1996; De Jong Gierveld, 2002; Vespa, 2012). For those wanting to repartner, people could opt for cohabitation to protect their pension benefits, accrued wealth, control over finances, and children’s inheritance, considering that cohabitation does not have the same legal implications as marriage.

Culture

In Canada, there are regional differences in the prevalence of cohabitation that reflect cultural differences in the meaning of marriage and preferences for nonmarital living arrangements (Laplante, 2006; Le Bourdais & Lapierre-Adamcyk, 2004). In Quebec, people have adopted cohabitation in greater numbers than elsewhere in Canada (Laplante, 2006; Le Bourdais & Lapierre-Adamcyk, 2004). Since the 1990s, the rate of cohabitation in Quebec has been more than double the rate in other regions. Le Bourdais and Lapierre-Adamcyk (2004) observe that, outside Quebec, cohabitation is accepted as a prelude to marriage, but it has not been widely accepted as an alternative to marriage or a suitable union for child rearing. In Quebec, cohabitation is overtaking marriage as the modal form of conjugal union, and childbirth is more common in cohabitation than marriage in Quebec. Some older people could be reluctant to cohabit because of previous taboos against it, but regional differences in the social acceptance of cohabitation could influence their repartnering decisions.

This study has following four specific hypotheses:

• Hypothesis 1: The overall prevalence of repartnering after widowhood is lower for widows than widowers.

• Hypothesis 2: Cohabitation is a more prevalent choice of union than remarriage among the widowed.

• Hypothesis 3: High socioeconomic resources increase the hazard rate of cohabitation after widowhood and decrease the hazard rate of remarriage.

• Hypothesis 4: The prevalence of cohabitation after widowhood is higher in Quebec than in the rest of Canada.

Data and Methods

Data

The empirical analysis is based on data from the 2007 Canadian General Social Survey, Cycle 21 (GSS-21), conducted by Statistics Canada. Statistics Canada’s GSS program is an annual national (cross-sectional) survey that gathers individual- and household-level data on Canadian adults to monitor changes in social conditions and the well-being of Canadians (Statistics Canada, 2009). Each cycle of the GSS has a thematic focus, such as family, time-use, and victimization. The GSS-21 focuses on family, social support, and aging. In addition to standard demographic, social, and economic information, it collected detailed information on family history, childbearing history, social support, health conditions, and retirement planning.

Due to the aging focus, the target population of the GSS-21 included Canadians aged 45 and older living in all 10 provinces, excluding individuals living in the northern territories (remote areas) and full-time residents of institutions. The data were collected through telephone interviews, using the random digit dialing (RDD) method. Although households without telephones were excluded, they represented only 0.9% of the target population (Pelot & Kemeny, 2009). Households with cellular phone service only (6.4%) were also excluded, which is a limitation of the study. Although we are unaware of any Canadian studies on coverage bias in the traditional RDD method, research in the United States has shown that the rate of households with cellular phone service only is much higher in the United States (13.5% in 2007) and is especially common among young-adult and/or low-income households (Blumberg & Luke, 2007, 2008). Given our study population, it is unlikely that the exclusion of households without landlines would significantly bias our regression estimates on the hazard rate of cohabitation and remarriage.

The GSS-21 includes a nationally representative sample of 23,404 Canadians aged 45 and older, with an overall response rate of 57.7%. To study repartnering after widowhood, we restricted our study sample to respondents who were widowed at or after age 45. With this restriction, the original sample was reduced to 3,261. To construct reliable event-history data for survival analysis, we further removed the cases where either age at widowhood, age at cohabitation (after widowhood), and/or age at remarriage (after widowhood) are missing (n = 142 or 4.6%). Missing values for the covariates were minimum, ranging from 0% to 2.6% (educational attainment), which were imputed using the multiple imputation method (Little & Rubin, 1987). The final study sample includes 2,479 widows and 640 widowers (N = 3,119).

Our study of repartnering after widowhood was retrospective in design. As noted, the GSS-21 focuses on family. It collected detailed retrospective data on the histories of union formation and dissolution. Using these retrospective questions, we were able to establish the timing of repartnering (remarriage or cohabitation) after widowhood, if it occurred. Although retrospective (survival) data allow us to study the timing of repartnering and estimate the (hazard) rates of competing union choices (i.e., remarriage and cohabitation), they have some inherent limitations. First, retrospective studies may not have data on all potential confounding factors in part because the events of interest had already occurred prior to the time of the survey and much of information that may be pertinent to exposure to the event was collected only at the time of the survey. For example, income/wealth and health are well-known correlates of union formation in later life (Brown et al., 2006; Smith et al., 1991; Vespa, 2013). But the data on socioeconomic resources and health were collected at the time of the survey, reflecting the respondents’ current status, and do not necessarily reflect their socioeconomic resources and health at the time when they were searching for a partner. In our study, efforts were made to include proxy measures of economic well-being (e.g., saving and pension) and health (chronic illness), reducing potential bias that may have been introduced due to the absence of relevant exposure data. Second, unlike prospective (longitudinal) studies, selective mortality cannot be taken into account in the study design because the study includes only “survivors.” This potential selection bias could overestimate the rate of union formation because frail people would probably be less represented in the study and they have lower propensity to repartner (Fu & Goldman, 1996). Although we do not anticipate that the rate of repartnering is grossly overestimated because repartnering generally occurs at relatively young ages and mortality is low among this segment of the population, caution should be exercised particularly when comparing our findings with those using prospective data.

Measures

Our analysis considers cohabitation and remarriage as two competing union choices after widowhood. As such, the dependent variable is measured as a three-level categorical variable indicating if the respondent entered a cohabiting union, a remarriage, or remained single after widowhood. Exposure time to the “risk” of repartnering is measured from the date of widowhood to the date of repartnership (either a cohabitation or remarriage). If a repartnership did not occur (the censored cases), exposure time is measured from the date of widowhood to the time of the survey. Overall, 93% of the female respondents and 73% of the male respondents are censored in the study.

We considered several independent variables known to influence union formation in mid and later life (e.g., Brown et al., 2006; Vespa, 2012; Wu & Schimmele, 2005). As noted, using retrospective data, much of the cross-sectional information collected in the survey that is seemingly pertinent to the study cannot be used because these measures are not tied to the time of the event (cohabitation/remarriage). Given this limitation, in addition to gender, we chose a set of 10 independent variables. Table 1 presents the definitions and descriptive statistics for the independent variables.

We considered three demographic variables. Age at widowhood is measured in years. Table 1 shows that the mean age of widowhood is 64 for women and 65 for men. The mean age of widowhood reported here should not be taken as an estimate of mean age of widowhood for the Canadian population because widowhoods that occurred prior to age 45 were not included in the estimation. Age at first marriage is measured in years. We observe that the mean age at first marriage is 23 for women and 26 for men. The number of children is also considered. As a source of social support, the number of children could reduce the need to repartner among the widowed (Carr, 2004b; Silverstein & Bengtson, 1994). Table 1 shows that the mean number of children is 2.9 for widows and 2.6 for widowers.

We included four socioeconomic variables. Education is an ordinal variable in 10 levels, ranging from elementary-level education or less to some postgraduate education or higher. It is treated as a continuous variable in the analysis. The mean level of education for women is 3.7 (between high school diploma and trade/technical school) and 4.2 for men (between trade/technical school and community college). We considered three indicators for economic well-being. Pension is a dummy variable indicating whether the respondent has a pension plan through employment beyond government-sponsored pensions (i.e., the Canada or Quebec Pension Plans). Table 1 shows that 25% of widows and 56% of widowers have an employment-sponsored pension. Savings is also a dummy variable identifying those who have built up their savings or made investments (e.g., stocks, bonds, mutual funds, rental income, and equity in business) to prepare for retirement. Though this indicator reflects the circ*mstances at the time of the survey, the assumption is that saving for retirement is a habitual behavior and is a long-term goal for many Canadians in mid and later life (Moussaly, 2010). We observe that 45% of widows and 57% of widowers have built up their savings or made investments toward retirement. The final indicator for economic well-being is bankruptcy, which is coded as a time-varying variable indicating whether the respondent ever experienced bankruptcy that had a major impact on their life. Table 1 shows that 3.2% of widows and 5.2% of widowers ever experienced bankruptcy.

The analysis includes two cultural variables. Region is measured as a dummy variable indicating whether the respondent resides in the province of Quebec. This variable reflects inter-regional cultural differences in the prevalence and social acceptance of cohabitation and declining rates of marriage (Wu, 1995). About a quarter of the widowed population live in Quebec, which is comparable to the overall distribution of the national population. Religious denomination is measured as a four-level categorical variable. We observe that 42% of women and 41% of men are Catholics, and comparable figures for Protestants are 44% and 36% for women and men, respectively. Table 1 shows that 6% of women and 16% of men reported no religious orientation.

We used a single indicator for health status, which is an important repartnering differential (Brown et al., 2006). Chronic illness is a dummy variable indicating the presence of one or more chronic illnesses reported by the respondent. We understand that chronic conditions can and will change with age and that the presence of chronic illness at the time of the survey is only a proxy for health status. Our data show that 68% of women and 62% of men reported one or more chronic conditions.

Statistical Methods

Statistical methods that are appropriate for retrospective (survival) data are known as survival models (Lee, 1992). These techniques were developed to model the time to the occurrence of an event of interest (remarriage or cohabitation). In essence, our analysis involved following a retrospective cohort of widowed individuals at or after age 45 and examining how selected individual-level characteristics may influence the timing (rate) of repartnering over time. We began our analysis with a double-decrement life table procedure to estimate the monthly rates of cohabitation and remarriage following widowhood and provide the description of these probabilities in terms of the cumulative experience of repartnering by successive years after widowhood. Separate life tables were estimated for women and men, for Quebec and elsewhere in Canada, and for each exit status.

Next, we estimated a series of Cox’s proportional hazard models of duration dependence to estimate the correlation between the selected independent variables and union formation after widowhood (Cox, 1972). We used the Cox model as it allows the form of the baseline hazard function to be unspecified. As noted, we treated cohabitation and remarriage as competing risks. Separate models were estimated for each event type, censoring the other event type at the beginning of the interval when it occurred. We estimated separate models for women and men. We also modeled cohabitation and remarriage jointly and estimated models of overall union formation following widowhood, which are reported in the Supplementary Appendix. In unreported analysis, we tested the proportionality (proportional hazards) assumption of our main regression models, a key assumption underlying the Cox model, by creating and examining the time-dependent covariates (see Hosmer, Lemeshow, & May, 2008). We found that the proportionality assumption generally holds well with the exception of three covariates in the model of men’s postwidowhood cohabitation (see Table 2). The hazard rate appears to decline with an increase in age at widowhood, age at first marriage, and the number of children (the results are available from the authors). These results will be further discussed in the following section.

Results

The Life Table Analysis

Figure 1 provides a graphic representation of the life table estimates of the cumulative proportions of repartnering (remarriage and cohabitation combined) for widows and widowers aged 45 and older (the full life tables are available from the authors). There is a large gender difference in the prevalence of repartnering. For both widows and widowers, the rate of repartnering is low during the first year of widowhood. Under 0.5% of widows and 3% of widowers formed a new union one year of widowhood. The gender gap in repartnering emerges soon after and widens over time. Three years after widowhood, under 3% of widows have formed a new union, compared with 17% of widowers. Five years after widowhood, about 4% of widows and 21% of widowers are remarried or cohabiting. Ten years after spousal loss, about 7% of widows and 29% of widowers have formed a new union. There is little change in the cumulative proportion of repartnering after this, which suggests that most repartnering occurs within 10 years of widowhood. Most widows and widowers who have not repartnered during this timeframe are single 15 and 20 years after widowhood as well. Given the average age at widowhood (64 years for women and 65 years for men), it is reasonable to assume that most of these individuals never repartner.

Figure 2a and b present the sex-specific cumulative proportions of ever cohabited or ever remarried after widowhood. These figures illustrate that marriage is a more prevalent choice of repartnering than cohabitation after widowhood. The aforementioned gender difference in repartnering is persistent in both remarriage and cohabitation after widowhood, but most of the gender difference in repartnering is attributable to men’s higher prevalence of remarriage after widowhood. The gender difference in cohabitation after widowhood is less pronounced, but it is still large.

Figure 2a presents the results for cohabitation after widowhood. The rate of cohabitation is low during the first year of widowhood. The cumulative proportion of widows and widowers who have cohabited within a year of widowhood is 0.1% and 1%, respectively. In comparison, Figure 2b shows that 0.4% of widows and 2% of widowers have remarried within one year of widowhood. Our results suggest that cohabitation does not provide a quicker route than remarriage into a new union after widowhood. Three years after widowhood, 0.8% of widows and 5% of widowers have cohabitated. In comparison, 2% of widows and 11% of widowers have remarried. Within five years of widowhood, 1% of widows and 6% of widowers have cohabitated. The proportion of widows and widowers who cohabit after widowhood plateaus at around the 15 year mark. At this time after widowhood, over 2.3% of widows and 8.5% of widowers have cohabited.

In comparison to remarriage, the cumulative proportion of cohabitation after widowhood is small. Overtime, the increase in the cumulative proportion of cohabitation is also comparatively flat. Figure 2b shows that 2.2% of women and 11.4% of men have remarried three years after widowhood. After five years, 3.1% of widows and 15.3% of widowers have remarried. Ten years after widowhood, about 4% of widows and 20% of widowers have remarried. The gender differences in repartnering after widowhood are consistent with our expectations. Overall, the rate of remarriage for men is about twice that of women, while men’s rate of cohabitation is about four times that of women. Regardless, marriage continues to be the more prevalent union choice after widowhood: 20 years after widowhood, the rate of remarriage is about twice as high as the rate of cohabitation for both widows and widowers.

Figure 3a and b illustrate the life table estimates of the cumulative proportions of cohabitation and remarriage after widowhood in Quebec and elsewhere in Canada (the full life tables are available from the authors). The figures confirm that cohabitation has largely replaced remarriage in the widowed population (as in the younger population) in the province of Quebec, but not elsewhere in Canada, where remarriage remains the most popular repartnering choice. Three years after widowhood, 3.1% of Quebeckers versus 1.4% non-Quebeckers have formed a cohabiting relationship (see Figure 3a). After five years, the figures are 5% and 2% for Quebeckers and non-Quebeckers, respectively. After ten years, the comparable figures are 6.2% and 2.4%. A reversed pattern is observed for remarriage. Figure 3b shows that the rate of remarriage is considerably lower in Quebec than in the rest of Canada. Three years after widowhood, only 2.1% of Quebeckers have entered into remarriage, compared with 4.4% of Canadians living in other provinces. Ten years after widowhood, only 4% of Quebeckers have remarried, compared with nearly 8% of non-Quebeckers.

Because the life table estimates were estimated from a double-decrement life table, we can combine the rate of cohabitation and remarriage to obtain the overall rate of union formation following widowhood (the estimates are available from the authors). The combined rates of union formation suggest that for the first three years after widowhood, the rate of repartnering is lower in Quebec than elsewhere in Canada, primarily because of the low rate of remarriage. In fact, the rate of remarriage in Quebec within the first year of widowhood is virtually zero, but it is about 1% elsewhere in Canada. The overall rate of repartnering becomes similar between Quebec and the rest of Canada after four years since widowhood. These findings suggest that, though cohabitation is the modal choice of repartnering in Quebec, the region appears to have more stringent cultural expectations surrounding widowhood and the length of the mourning period.

The Cox Models

Table 2 presents the parameter estimates from Cox’s proportional hazard models of postwidowhood cohabitation and remarriage. For both women and men, age at widowhood has a negative and significant association with the hazard rate of cohabitation. For women, a one-year increase in age at widowhood reduces the hazard rate of cohabitation by 11% [(e^−.113 – 1) × 100]. The comparable figure for men is 6%.

Age at first marriage has a similar negative association with cohabitation. For both women and men, older age at first marriage, or a delayed entry into first marriage, reduces the hazard rate of cohabitation after widowhood. For example, a three-year delay in marriage reduces the hazard rate of cohabitation by 43% for women and 16% for men.

Contrary to our expectations, the associations of socioeconomic variables are generally nonsignificant with only one exception. We find that savings appear to have a positive association with women’s postwidowhood cohabitation, although the estimate is only marginally significant (p < .1). Moreover, chronic illness has a nonsignificant association with cohabitation for both widows and widowers.

Table 2 shows that number of children has no significant association with cohabitation after widowhood for women and has a marginally significant association for men (p < .1). An increase in the number of children reduces men’s hazard rate of cohabitation. Religious denomination does not have a significant association with cohabitation after widowhood for either widows or widowers. Moreover, consistent with the life table analysis, for both widows or widowers, the hazard rate of cohabitation is significantly higher in Quebec than elsewhere in Canada.

Finally, as noted in the Methods section, we uncovered that the associations of three covariates in the model of men’s postwidowhood cohabitation change with time (since widowhood), violating the proportionality assumption that underlies the Cox model. Specifically, we found that the negative association of age at widowhood, age at marriage, and number of children become more pronounced over time. These findings suggest that as men age, the decline in the hazard rate of cohabitation accelerates with increases in their age at widowhood, age at first marriage, or the number of children they have.

Discussion and Conclusion

The main purpose of this study was to describe the prevalence of cohabitation after widowhood and compare it to the prevalence of remarriage. We presented sex-specific estimates of repartnering because the opportunities for entering a union in later life are dependent on gender (Guttentag & Secord, 1983). Considering that the supply of older single males is comparatively lower than the supply of older single females, we hypothesized that the rate of repartnering would be lower for widows than widowers. Our findings supported this hypothesis. In comparison to women, the cumulative prevalence of cohabitation after widowhood is about 4 times higher for men. The cumulative prevalence of remarriage is 5 times higher for men. Although this gender gap in repartnering appears to be largely a function of demographic constraints (low supply of eligible men) in marriage markets, it is possible that it also reflects the perceived costs of repartnering for women. Prior research demonstrates that some women prefer to remain uncoupled after widowhood to preserve the benefits of independence and to avoid the personal costs that associate with the gender division of labor and caregiving responsibilities (Davidson, 2001; De Jong Gierveld, 2002).

Remarriage is losing ground to cohabitation following union dissolution (divorce) and the prevalence of late-life cohabitation is increasing (Brown et al., 2006; Connidis, 2010). Given the legal distinction between marriage and cohabitation, the latter could be the preferable option as well as a quicker route into repartnering. Brown and colleagues (2012) observe that cohabitation appears to be an alternative to marriage in later life, whereas it is primarily a prelude to marriage at earlier stages in the life course. The authors also suggest that older adults could be less interested in marriage than younger adults because the costs and benefits of marriage change across the life course. For these reasons, we hypothesized that cohabitation would be more prevalent than remarriage among those than repartnered after widowhood. Our findings failed to support this expectation. Within 20 years of widowhood, the cumulative prevalence of remarriage is about twice as high as the prevalence of cohabitation. The prevalence of remarriage is also comparatively higher during the first few years after widowhood, which suggests that informality of cohabitation does not offer a quicker route into repartnering.

Some studies posit that there are economic disincentives to remarriage in later life, but still a desire for companionship, and these factors could “push” older people into cohabitation. Cohabitation provides much of the benefits of marriage, but it is more flexible with regard to sharing economic resources (Chevan, 1996; Vespa, 2012, 2013). Hence, we hypothesized that increases in economic resources would increase the hazard rate of cohabitation. Our findings did not support this hypothesis. For the present cohort of widowed people, the putative economic disincentives of remarriage are not pushing them into cohabitation. Socioeconomic resources have a similar association with cohabitation after widowhood as for remarriage. The relationship between repartnering and socioeconomic resources is nonsignificant or marginally significant. Although our measures of wealth/savings are limited, previous studies that use detailed measures also demonstrate that socioeconomic resources have weak or nonsignificant effects on union formation in later life (Brown et al., 2012; Vespa, 2012, 2013).

There is little difference in the overall repartnering rate between the widowed in Quebec and other regions of Canada, but the rate of cohabitation after widowhood is far higher in Quebec and the rate of remarriage is much lower. These findings are consistent with our expectations and with the idea that cohabitation has evolved into an alternative to marriage in Quebec. Among all age groups, about one third of unions in Quebec are cohabitations (Kerr, Moyser, & Beaujot, 2006). In contrast, about 12% of unions in the rest of Canada are nonmarital cohabitations. Le Bourdais and Lapierre-Adamcyk (2004) offer two explanations for this divergent pattern of union formation. First, during the 1960s, Quebec experienced a widespread secularization movement. This led to fundamental changes in union formation behavior as a growing number of people rejected the Catholic Church as well as institutional control over people’s sexuality and conjugal relationships. Second, the feminist movement is stronger in Quebec than in other provinces, and the rise of cohabitation reflects the struggle for greater gender equality. Le Bourdais and Lapierre-Adamcyk (2004) observe that, in comparison to marriage, cohabitation is based on less gender role specialization and greater independence of women.

Our conclusions are tempered by data limitations. First, the retrospective design of this study limits our understanding of the effects of the selected explanatory variables. For example, the measures of socioeconomic resources occurred at the time of the survey. Longitudinal data or detailed retrospective questions are needed to properly understand how socioeconomic status factors into repartnering decisions. However, most of our selected socioeconomic variables are fairly robust since factors such as education and having a retirement pension plan almost surely predate repartnering after widowhood. The only factor that is in doubt is retirement savings, which could change (increase) upon repartnering. Our measure of the effects of health status on repartnering is also limited because it is not measured prior to repartnering. Second, our data are limited by a lack of a measure of the desire for repartnering. Previous research demonstrates that the desire for repartnering is higher than repartnering itself, largely because of constraints in marriage markets (Carr, 2004b). We cannot discern if there are social differences in the desire to cohabit or remarry after widowhood. Despite these limitations, this study contributes to our understanding of repartnering after widowhood and shows that remarriage remains the predominant choice of repartnering after widowhood.

Acknowledgments

The authors gratefully acknowledge support from the Social Sciences and Humanities Research Council of Canada, and Merril D. Silverstein and two JGSS reviewers for their helpful comments. An earlier version of this paper was presented at the International Symposium on Aging Families, Victoria, BC, Canada, June 3–4, 2013.

References

Author notes

Decision Editor: Merril Silverstein, PhD

© The Author 2014. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Issue Section:

Special article

Email alerts

Citing articles via

More from Oxford Academic