Population aging, caused by lower fertility rates and longer life expectancy, is a topic of increasing importance in many of the world’s industrialized nations (Brown, Reference Brown2011; Cooke, Reference Cooke2006; McDaniel & Rozanova, Reference McDaniel and Rozanova2011; McDonald & Donahue, Reference McDonald and Donahue2011; Turcotte & Schellenberg, Reference Turcotte and Schellenberg2006). Between 1981 and 2011, the proportion of Canada’s population made up of those aged 65 years and older increased from 10 per cent to 14 per cent, and it is expected that by 2036, seniors will comprise 25 per cent of Canada’s population (Cooke, Reference Cooke2006). The near future will involve large numbers of people leaving the workforce as members of the baby boom generation, those born between 1946 and 1964, reach their retirement years (Brown, Reference Brown2011; McDonald & Donahue, Reference McDonald and Donahue2011; McMullin, Cooke, & Downie, Reference McMullin, Cooke and Downie2004; Rix, Reference Rix2004). These demographic trends have led to concerns among many policymakers and employers that there will soon be a shortage of skilled workers and that the Canadian health and income safety nets will be compromised (Bergman, Béland, & Perrault, Reference Bergman, Béland and Perrault2002; Denton & Spencer, Reference Denton and Spencer2009). This, in turn, has led to modifications in immigration and pension policy as well as debate and discussion about the benefits of retaining older workers in the labour force (Denton & Spencer, Reference Denton and Spencer2009; Turcotte & Schellenberg, Reference Turcotte and Schellenberg2006).
Some statistics give cause for optimism concerning the availability of jobs for older individuals who desire to work for pay. Statistics Canada’s (2016) Labour Force Survey reports that Canada’s unemployment rate (defined as the proportion of the population not employed for pay and but who have actively searched for paid work in the previous four weeks) among persons aged 55 and older was 5.1 per cent in 2006 and 6.3 per cent in 2011. Within the first 11 months of 2016, Statistics Canada (2016) reported unemployment rates among persons aged 55 and older that ranged from a high of 7 per cent in April to a low of 5.2 per cent in June and September. The Labour Force Survey reported that the employment rate (defined as the proportion of the population employed for pay) among Canadians aged 55 and older was 30.5 per cent in 2006 and 33.9 per cent in 2011. Concerning the first 11 months of 2016, this survey reported employment rates among Canadians aged 55 and older that ranged from 34.9 per cent in January to 36.3 per cent in September (Statistics Canada, 2016). These statistics show that relatively small proportions of Canada’s older population have been unemployed throughout the past decade or so. Furthermore, these statistics show that the proportions of Canada’s older population employed for pay have been much higher than the proportions unemployed throughout the past decade or so. As such, these statistics show that older Canadians desiring paid work are usually able to find employment.
Indeed, increasing numbers of men and women are working later in life (Cahill, Giandrea, & Quinn, Reference Cahill, Giandrea and Quinn2006; Cahill, Giandrea, & Quinn, Reference Cahill, Giandrea and Quinn2007). Currently, about one half of retiring individuals in the United States either undergo a staged exit from paid work or return to the labour force after an episode of retirement (Maestas, Reference Maestas2010). Some scholars claim that many present-day workers will seek to gradually leave the workforce instead of making one transition from full-time employment to full retirement (Davis, Reference Davis2003; Dendinger, Adams, & Jacobson, Reference Dendinger, Adams and Jacobson2005; Morrow-Howell, Reference Morrow-Howell2007; Pengcharoen & Shultz, Reference Pengcharoen and Shultz2010). Workers who are approaching retirement might seek work arrangements involving fewer hours of work and lower levels of responsibility, greater flexibility in how they arrange their work schedule, as well as the opportunity to take up shared positions (Pengcharoen & Shultz, Reference Pengcharoen and Shultz2010). Pengcharoen and Shultz (Reference Pengcharoen and Shultz2010) have defined “bridge employment” as any part-time or full-time transitional paid work undertaken after having exited one’s career employment. This staged withdrawal from the workforce allows for a gradual process of adjustment to life as a retiree (Pengcharoen & Shultz, Reference Pengcharoen and Shultz2010). In reviewing the literature on the non-standard paths to full retirement undertaken by older individuals, we interchangeably use the terms “return to work” and “bridge employment”. Both are conceptually related as both represent paid work undertaken after having left one’s career employment.
In this article, we examine return to work after formal retirement, one pathway through which older workers are increasingly engaging in paid work. Although there has been some research done on the factors that contribute to post-retirement return to work in Canada (see, for example, Armstrong-Stassen, Reference Armstrong-Stassen2008; Armstrong-Stassen, Schlosser, & Zinni, Reference Armstrong-Stassen, Schlosser and Zinni2012; Armstrong-Stassen & Staats, Reference Armstrong-Stassen and Staats2012; Lefebvre, Merrigan, & Michaud, Reference Lefebvre, Merrigan and Michaud2011; Schellenberg, Turcotte, & Ram, Reference Schellenberg, Turcotte and Ram2005; Schlosser, Zinni, & Armstrong-Stassen, Reference Schlosser, Zinni and Armstrong-Stassen2012; Turcotte & Schellenberg, Reference Turcotte and Schellenberg2006), this research has tended to be rather simplistic in its consideration of gender and the similarities or differences between men and women in the factors that influence them to engage in post-retirement paid work. Yet gender is a salient feature of the social structure that influences paid employment throughout the life course (Davis, Reference Davis2003; Perkins, Reference Perkins1994; Quinn & Kozy, Reference Quinn and Kozy1996). Hence, it is important to examine the complexities of the relationship between return to work after retirement and gender, and, in particular, whether and how gender intersects with other key variables in its influence on individual engagement in post-retirement paid work.
Factors Associated with Return to Work after Retirement
In Canada, 22 per cent of the people aged 55 and older worked for pay in 1995. This increased to 31.7 per cent in 2007, the year from which we drew the data in the current study (Statistics Canada, 2007b). Past research shows that many factors influence post-retirement paid work. Individuals who have higher levels of socioeconomic status, measured using variables such as occupation, education, and financial security, are more likely to return to work after retirement than are those with lower levels of socioeconomic status (Cahill et al., Reference Cahill, Giandrea and Quinn2006; Griffin & Hesketh, Reference Griffin and Hesketh2008; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011; Maestas, Reference Maestas2010; Raymo, Warren, Sweeney, Hauser, & Ho, Reference Raymo, Warren, Sweeney, Hauser and Ho2010; Turcotte & Schellenberg, Reference Turcotte and Schellenberg2006). Cahill et al. (Reference Cahill, Giandrea and Quinn2006) argued, however, that wage level in the United States holds a U-shaped relationship with bridge employment. Those earning low wages are more likely to return to work for financial reasons, whereas those who earn higher wages engage in post-retirement work for the personal satisfaction that it brings them (Cahill et al., Reference Cahill, Giandrea and Quinn2006). It has been found that, in the United States, good finances imply a lower likelihood of planning to undertake bridge employment (Davis, Reference Davis2003; Kim & Feldman, Reference Kim and Feldman2000; Weckerle & Shultz, Reference Weckerle and Shultz1999).
Some other demographic variables are also linked with propensity to return to work after retirement. Age, for example, influences post-retirement employment decisions. Younger adults and those who retired at younger ages are more likely to return to work after retirement than older individuals and those who retired at later ages respectively (Cahill et al., Reference Cahill, Giandrea and Quinn2006; Davis, Reference Davis2003; Griffin & Hesketh, Reference Griffin and Hesketh2008; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011; Maestas, Reference Maestas2010; Turcotte & Schellenberg, Reference Turcotte and Schellenberg2006). Regarding health, those in better health are more likely to return to work than those who have health difficulties (Cahill et al., Reference Cahill, Giandrea and Quinn2006; Griffin & Hesketh, Reference Griffin and Hesketh2008; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011; Maestas, Reference Maestas2010; Turcotte & Schellenberg, Reference Turcotte and Schellenberg2006). Pleau and Shauman (Reference Pleau and Shauman2013) and Lefebvre et al. (Reference Lefebvre, Merrigan and Michaud2011) found that in the United States and in Canada, respectively, having more children is associated with a higher likelihood of return to work after retirement. Of relevance to these findings, Cahill et al. (Reference Cahill, Giandrea and Quinn2006) showed how, in the United States, both men and women with dependent children are more likely to remain employed in later life; further, they showed how older women with dependent children who transition out of their career employment are more likely to take on a bridge job. These authors linked this later life employment to the need for older parents to provide for their children financially when they suggested that one source of motivation for this later-life paid work might be costly college tuition up ahead.
Davis (Reference Davis2003) argued that individuals make life decisions, including whether to engage in bridge employment, within an “opportunity structure” that is based upon biographical characteristics and the features of their environment. All of the factors discussed in the preceding two paragraphs have both biographical and structural components. Gender, too, is a factor with significant biographical and structural components, yet studies in Canada and the United States that have examined the relationship between gender and return to work post-retirement have produced mixed results (Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011; Maestas, Reference Maestas2010).
Some research shows that men are more likely than women to resume labour force activity after a first retirement (Davis, Reference Davis2003; Griffin & Hesketh, Reference Griffin and Hesketh2008; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011; Maestas, Reference Maestas2010). Turcotte and Schellenberg (Reference Turcotte and Schellenberg2006) found that in Canada, men have a 25 per cent probability of returning to work post-retirement, whereas women have an 18 per cent probability of doing so. However, Griffin and Hesketh (Reference Griffin and Hesketh2008) explained that research on this topic has produced some inconsistent results. Some studies have shown women to be equally or more likely than men to engage in bridge employment (Griffin & Hesketh, Reference Griffin and Hesketh2008). Griffin and Hesketh (Reference Griffin and Hesketh2008) emphasized the need to control for finances in studies on this topic since the effect of gender might be based on the fact that women tend to hold lower-paying jobs and, therefore, hold expectations of the future that differ from those of men.
Gendered patterns of paid work after retirement are based in the differences between the work trajectories of men and women in early and middle adulthood. In his study, Davis (Reference Davis2003) hypothesized that women are more likely to take a bridge job because of their likelihood of having had more fragmented work histories. Fragmented work histories, according to this argument, likely prevent women from achieving financial security in their later years. Bridge employment therefore becomes necessary in order for women to maintain their standard of living (Davis, Reference Davis2003), yet Davis (Reference Davis2003) found that men are more likely to engage in bridge employment. On the other hand, Perkins (Reference Perkins1994) found that, in the United States, women are likely to return to work after retirement because of financial difficulties resulting from them having spent their lives employed in sex-segregated work environments marked by sex discrimination. Quinn and Kozy (Reference Quinn and Kozy1996) found that American women are more likely to be involved in bridge employment; however, this is because the women in their sample were more likely to have had work records characterized by part-time or temporary jobs that resemble typical bridge jobs. When the authors restricted their sample to men and women who had had full-time career jobs, rates of bridge employment became nearly equal for men and women.
Because gender influences most aspects of social life, we agree with Beehr and Bennett (Reference Beehr and Bennett2015) who, in their study regarding gender influences in the United States, argued that more research needs to be conducted that considers how gender intersects with other variables in its influence on bridge employment. In particular, there is a need for greater understanding of how gender interacts with relationship status in its association with return to work after retirement. Most studies in Canada and the United States indicate that marital status on its own does not influence the likelihood of engaging in bridge employment (Armstrong-Stassen, Reference Armstrong-Stassen2008; Davis, Reference Davis2003; Kim & Feldman, Reference Kim and Feldman2000; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011 Maestas, Reference Maestas2010; Pleau, Reference Pleau2010; Schlosser et al., Reference Schlosser, Zinni and Armstrong-Stassen2012). Some studies in the United States have investigated interactions between gender and marital status as they relate to patterns of retirement and work in later life (McNamara & Williamson, Reference McNamara and Williamson2004; Szinovacz, DeViney, & Davey, Reference Szinovacz, DeViney and Davey2001), as well as to post-retirement work specifically (Pleau, Reference Pleau2010; Pleau & Shauman, Reference Pleau and Shauman2013). Pleau (Reference Pleau2010) found that American women who are married tend to be less likely than unmarried women to engage in post-retirement work. Pleau and Shauman (Reference Pleau and Shauman2013) further observed that married men in the United States are more likely than are unmarried men to return to the workforce after retirement. Pleau (Reference Pleau2010) attributed the higher likelihood of return to work – by American women who are not married and have retired – to their precarious financial circumstances.
Despite this notable work, there is a relative lack of research that investigates how this interaction is associated with return to work after retirement. Yet men and women experience marriage (or cohabitation) and singleness differently, which in turn, may lead to different post-retirement work outcomes as well.
Recent times have seen convergence in the roles performed by men and women (Crompton & Harris, Reference Crompton and Harris1998; Powell & Mainiero, Reference Powell and Mainiero1992). Blair-Loy (Reference Blair-Loy2003) cited research showing that, although in the early 1950s a substantial majority of American families conformed to expectations that men would be breadwinners while women took care of the home, this proportion dropped to about a quarter of families by the end of the 1990s. This trend over time suggests that present-day older populations that came of age earlier in the 20th century might still hold values and show expectations that conform to the more traditional and less equal gender relations (Crompton & Harris, Reference Crompton and Harris1998).
These expectations for men to be the breadwinners of the family have created a system in which men who have never married tend to compare unfavourably in their occupational success with those who have married (Leslie, McClure, & Oaxaca, Reference Leslie, McClure and Oaxaca1998; Nakosteen & Zimmer, Reference Nakosteen and Zimmer1997), perhaps indicating lower levels of motivation. However, concerning the relative success of men who have married compared with those who have not, a debate exists between those researchers who attribute this relative success to processes of selection into marriage and those who attribute this relative success to the advantages that marriage provides to men’s lives. One theory holds that married men earn more because high-earning men hold greater value in the marriage market and are thus more likely to marry (Nakosteen & Zimmer, Reference Nakosteen and Zimmer1997). Nakosteen and Zimmer (Reference Nakosteen and Zimmer1997) found that single American men who earn more are more likely to marry over time, and that married men who earn more are less likely to divorce. This study suggests that processes of selection of men of higher ability into marriage are at least a part of the explanation for the earnings premium of married men. Some research in the United States shows that, although in the present day higher levels of education increase the probability of both men and women being married, this difference by level of education is greater for men (Aughinbaugh, Robles, & Sun, Reference Aughinbaugh, Robles and Sun2013; England & Bearak, Reference England and Bearak2012). In their study of gender, financial position, and divorce in the Netherlands, Kalmijn and Poortman (2006) stated:
Several authors have suggested a negative income effect on divorce. … A poor financial position can often be attributed to the husband being unable to provide for his family. Financial problems are a potential threat to the husband’s role as breadwinner. And just as wife’s employment may be disapproved of by the husband, the husband’s inability to provide for his family may be disapproved of by the wife. (p. 204)
The other theory suggests that married men earn more because they have more time and energy to devote to paid work activities as their wives undertake the duties based on housework and care of family members (Chun & Lee, Reference Chun and Lee2001). In support of this idea, Chun and Lee (Reference Chun and Lee2001) found that the wage premium of married American men is explained by the degree to which roles within the household are specialized.
Other studies focus on how, within marital relationships, the decisions taken in retirement, as well as the motivations behind the decisions taken, differ by gender. Some scholars expect that the retirement decisions of American men are more likely to be impacted by financial obligations to their family members, whereas the decisions taken by women during retirement are more likely to be related to their sense of social obligation to their kin (Szinovacz et al., Reference Szinovacz, DeViney and Davey2001). Another study of gender, marital characteristics, and retirement in the United States, by Szinovacz and DeViney (Reference Szinovacz and DeViney2000), presents the argument that power dynamics within a marriage can encourage retired men to return to work while discouraging the post-retirement paid work of women. A man might proceed with paid work after retirement in order to maintain his status within his marriage. A woman, on the other hand, might hasten her exit from paid work if her continued paid employment might be seen by her husband as a threat to his dominance.
Much evidence suggests that women, on the other hand, who do not marry are more career-oriented, prioritizing socioeconomic and career success over commitment to family. Leslie et al. (Reference Leslie, McClure and Oaxaca1998) found that, among American women, having never married is associated with a higher likelihood of being employed full-time in the physical sciences and engineering professions and a lower probability of being in part-time employment, or of being either unemployed or out of the labour force. This is the opposite of what they found concerning men. Ellemers, van den Heuvel, de Gilder, Maass, and Bonvini (Reference Ellemers, van den Heuvel, de Gilder, Maass and Bonvini2004) also delved into the underrepresentation of women in science in their study of this subject concerning the Netherlands. They posited that women with careers in science are likely to be single and without children, and to explain this, they emphasized normative expectations that women should be devoted to their homes and caregiving, including care of children and elders, rather than to paid employment. Scholars have noted that American women’s childcare and home upkeep responsibilities serve as obstacles to their continuous full-time employment (Kim & Feldman, Reference Kim and Feldman2000), implying that some women who have decided to dedicate themselves to their professional careers might opt to not get married and take on family or care responsibilities. In her study of women executives in the United States, Blair-Loy (2003) contrasted “conformists”, who adapt to societal expectations by choosing either to devote themselves fully to work or to devote themselves fully to family, with “innovators”, who seek to cause the larger society to change through their creative combinations of devotion to work and to family. Numerous women who decide to focus on their careers might thus “conform” and choose to avoid family responsibilities entirely by not getting married.
On a related note, some studies have investigated the “independence” hypothesis, according to which women with greater financial resources have less to gain from a marital relationship and are thus more likely to divorce (Boyle, Kulu, Cooke, Gayle, & Mulder, Reference Boyle, Kulu, Cooke, Gayle and Mulder2008; Dribe & Lundh, Reference Dribe and Lundh2012; Kalmijn & Poortman, Reference Kalmijn and Poortman2006; Lyngstad & Jalovaara, Reference Lyngstad and Jalovaara2010). In their study on marital dissolution in Sweden, Dribe and Lundh (2012) elaborated further that the traditional system of men dedicated to paid work and women dedicated to the home increased what both spouses gained from marriage. Women’s economic dependence on their husbands served to discourage union dissolution. Risk of dissolution might be highest when earnings between partners are equal, because then the level of dependency within the marriage is lowest. Aughinbaugh et al. (Reference Aughinbaugh, Robles and Sun2013) presented, within their study concerning patterns of marriage and divorce in the United States, the optimistic finding that these gendered trends that disadvantage women might be gradually fading away. These authors explained that, although women with a college education were less likely to marry than those without a college education in the past, women with a college education at present are at least as likely to get married as women with lower levels of education.
Of relevance to this discussion, Mirowsky and Ross (Reference Mirowsky and Ross2003) explored differences between American men and women in factors contributing to a sense of control. They stated that because of differences between men and women in earnings, autonomy, and responsibility for family and home, employment is much more beneficial for the sense of control by married men than it is for that by married women. Applying this line of research to return to work, it would seem that married men would be more likely to return to work after retirement than married women.
Methods
The source of the data used in this study was the 2007 General Social Survey (GSS-21) (Statistics Canada, 2007a), a cross-section of people aged 45 years and older who live in Canada, with the exclusion of the residents of the Yukon, the Northwest Territories, and Nunavut, as well as with the exclusion of those who are full-time residents of institutions. The sample utilized in the present study was composed of individuals between 50 and 74 years of age, who retired at least once, and whose first retirement occurred after having turned 50 years of age and within 15 years of when the data were collected. This amounts to 4,700 respondents. Our final analytical sample, involving the listwise deletion of missing data within any of the variables, included 4,414 respondents. All analyses within our study were based on this sample of 4,414 respondents. Because 93.9 per cent of the original sample is maintained, and because all variables with missing data included in the present study have less than 2.6 per cent missing, the missing data were sparse enough to be dealt with through listwise deletion. Notably, this was a selective sample that did not allow for comparisons between those who continued to work versus those who retired, even though their circumstances may have been very similar.
The age restrictions on our sample were chosen to ensure that we were studying an older population whose first retirement occurred during later life. We chose an upper bound of 74 years of age to avoid too much of a selection effect based on the unique characteristics of those living until very old age. We also restrained our sample to those having retired within 15 years of when the data were collected to avoid too much of a period effect based on change over historical time in retirement transitions and behaviours.
Variables
The main dependent variable in this study, return to work after retirement, was based on responses to the following question within the GSS-21 (Statistics Canada, 2007a, p. 210): “Did you/he/she do any paid work at any time at a job or business after your/his/her retirement?” Note that proxy interviews were allowed in the GSS-21 (Statistics Canada, 2007a) whenever a respondent could not answer survey questions because of inability to speak English or French, or because of a physical condition, a mental condition, or a health problem. Of the respondents within our analytical sample, 30.5 per cent (n = 1,346) returned to work after a first retirement.
Our central independent variables were gender (reference category = men) and relationship status (reference category = in a relationship). Relationship status is a binary variable based on yes/no responses to the following question pertaining to respondents’ first retirement: “Were you married or in a common-law relationship when you retired?” This central independent variable was effective for the present study because of its inclusiveness; same-sex relationships were included and present-day trends of high levels of cohabitation were integrated within this choice of variable. For the sake of simplicity, in this article we refer to whether or not a respondent was/is in a “marital/common-law relationship”. These two variables, gender and relationship status, were tested in interaction, and we refer to this interaction term as the “focal interaction”.
Adequacy of household income and investments at the time of first retirement to maintain one’s standard of living (financial adequacy), highest level of education obtained, and involvement over the past 12 months in formal and informal groups/organizations/networks/associations (community involvement) served as non-focal independent variables. The latter variable includes civic, community, and service clubs that perform volunteer work for the benefit of others. The financial adequacy variable was composed of the categories of those claiming their household income and investments at retirement to be more than adequate (reference category), adequate, barely adequate, or inadequate to maintain their standard of living. The highest level of education variable was based on three categories. The first category comprised those who had not received a high school diploma (reference category), the second represented those who obtained a high school diploma but who had not obtained a university degree, and the third represented those who graduated from university. The community involvement variable was a binary variable based on whether a respondent had been thus involved, and for which non-involvement served as the reference category. We employed education and community involvement (which includes volunteer work) in this study as indicators of the levels of engagement and motivation of the study’s respondents.
Current age and age at first retirement were combined into a single control variable based on the following categories:
-
– 50–54 years of current age / 50–54 years of age at first retirement
-
– 55–59 years of current age / 50–54 years of age at first retirement
-
– 55–59 years of current age / 55–59 years of age at first retirement
-
– 60–64 years of current age / 50–54 years of age at first retirement
-
– 60–64 years of current age / 55–59 years of age at first retirement
-
– 60–64 years of current age / 60–64 years of age at first retirement
-
– 65–69 years of current age / 55–59 years of age at first retirement
-
– 65–69 years of current age / 60–64 years of age at first retirement
-
– 65–69 years of current age / 65–69 years of age at first retirement
-
– 70–74 years of current age / 60–64 years of age at first retirement
-
– 70–74 years of current age / 65–74 years of age at first retirement (reference category)
The reference category contained those first retiring at 65 to 74 years of age in order to include an adequate cell size. Health at the time of retirement compared to others of that age also served as a control variable, divided into excellent (reference category), very good, good, fair, and poor.
Table 1 presents the cell sizes and proportions for each category within each of these categorical variables. Furthermore, as this study was focused on the interaction between gender and marital/common-law relationship status, cell sizes are presented here for all four possible combinations of these two variables.
Table 1: Descriptive statistics: Cell sizes and proportions (1), total sample size = 4,414
Analysis
We based the present study on a series of logistic regression analyses. In this type of regression analysis, dependent variables are based on binary categories, in which one of the two categories is the outcome being modelled, and the other serves as the reference category. Odds ratios were the main outcomes of the logistic regression equations we studied because they provide an effective assessment of how each independent variable category, when compared with its reference category (zero), affects the probability that a respondent falls within the modelled dependent variable category rather than the reference category of the dependent variable, holding all other variables constant. In our analysis, the odds ratios presented for gender and for marital/common-law relationship status individually pertained to these probabilities when the other component of the focal interaction was set to its reference category (zero).
The first step of the analysis involved the regression of return to work after retirement on the focal interaction, financial adequacy, current age and age at first retirement, as well as health at retirement. The second step included the addition of highest level of education to the logistic regression equation, with attention directed to the consequent changes in the odds ratios for the focal interaction and for financial adequacy. The third step included the further addition of community involvement to the logistic regression equation, with a focus on the consequent change in the odds ratio for the focal interaction. This progression of three logistic regression models allows for an assessment of whether the differences between men and women in how marital/common-law relationship status is associated with post-retirement work have to do with differing levels of engagement, motivation, and career orientation, as measured through highest level of education and through involvement in community groups.
Highest level of education and community involvement were also regressed upon the focal interaction and the set of control variables corresponding with each respective further step in the progression of three logistic regression models. Thus, although levels of education were controlled in the logistic regression model corresponding with involvement in the community, involvement in the community was not controlled in those models in which levels of education served as dependent variables. The parallel lines assumption was broken when highest level of education (a three-category ordinal variable) served as dependent variable. Therefore, we investigated this variable in two steps (no high school diploma vs. at least a high school diploma; no university degree vs. a university degree) for this segment of the analysis. These three additional models were tested to help determine whether marital/common-law relationship status had different connotations for the levels of engagement, motivation, and career orientation of men and women.
We conducted data analyses with the Stata 13 statistical software package. The logistic regression models were followed with the use of the “margins” command to reveal the predicted percentages of men who were in a marital/common-law relationship, men who were not in a marital/common-law relationship, women who were in a marital/common-law relationship, and women who were not in a marital/common-law relationship, who had engaged in paid work after retirement, who had obtained a high school diploma, who had obtained a university degree, and who had been involved in community groups. We conducted all analyses with the use of bootstrap weights provided for the GSS-21 (Statistics Canada, 2007a).
Results
Table 2 shows that gender, parental status, financial adequacy, current age and age at retirement, and health at retirement were all significantly related to the probability of return to work after retirement. Being a woman among those who were in a marital/common-law relationship (OR = 0.52, p < .001) was associated with a lower likelihood of involvement in post-retirement paid work. Being a parent (OR = 1.27, p < .05) was associated with a higher probability of return to work after retirement. Inadequate finances at retirement (OR = 1.59, p < .05) were associated with a higher probability return to work after retirement. Lower current ages and lower ages at first retirement were both associated with higher likelihoods of involvement in paid work after retirement (e.g., 55–59 years of current age / 50–54 years of age at first retirement: OR = 4.06, p < .001). Worse health at retirement was associated with a lower likelihood of post-retirement work (e.g., poor health at retirement: OR = 0.24, p < .001).
Table 2: Logistic regression analysis of return to work after retirement: Model 1, n = 4,414
Note: CI = confidence interval.
^ p < .10, * p < .05, ** p < .01, *** p < .001.
Most notably, the interaction between being a woman and not being in a marital/common-law relationship held an odds ratio of 1.52 that was significant at the p < .05 level. In Table 3, it can be seen that this was manifested in the fact that not being in a marital/common-law relationship implied a lower likelihood that a man would return to work after retirement (32% vs. 36.8% among those who were in a marital/common-law relationship), while implying a higher likelihood that a woman would return to work after retirement (27.4% vs. 24% among those who were in a marital/common-law relationship).
Table 3: Gender x relationship status probabilities of return to work after retirement: Model 1, n = 4,414
Table 4 shows the results obtained when highest level of education was added to the model. A higher level of education was associated with a higher likelihood of return to work after retirement. Having a high school diploma (OR = 1.91, p < .001) and, especially, holding a university degree (OR = 3.26, p < .001) were associated with higher likelihoods of return to work after retirement. It is also notable that a suppressor effect was revealed as the three odds ratios for lower levels of financial adequacy were now substantially higher, with the categories of “barely adequate” (OR = 1.58, p < .01) and “inadequate” (OR = 2.11, p < .001) both achieving substantially higher levels of statistical significance. Less financial adequacy implied a higher likelihood of post-retirement work for financial reasons. However, it also implied lower levels of education; higher levels of education were associated with a greater propensity to engage in paid work after retirement. With education controlled, the full extent to which financial difficulty was associated with a retired older person’s return to work was revealed.
Table 4: Logistic regression analysis of return to work after retirement: Model 2, n = 4,414
Note: CI = confidence interval.
^ p < .10, * p < .05, ** p < .01, *** p < .001.
Of most importance is the fact that the addition of highest level of education to the model reduced the odds ratio for the interaction between being a woman and not being in a marital/common-law relationship (OR = 1.43, p < .05) as it related to the likelihood of returning to work. This was manifested in the fact that, for both men and women, the difference in the probability of post-retirement work for those who were in a marital/common-law relationship and those who were not was now smaller. Table 5 shows that men who were in a marital/common-law relationship had a probability of return to work after retirement of 36.5 per cent, whereas men who were not in a marital/common-law relationship showed a likelihood of 32.5 per cent. The corresponding probabilities for women were 24.3 per cent and 27.2 per cent.
Table 5: Gender x relationship status probabilities of return to work after retirement: Model 2, n = 4,414
Table 6 shows the set of odds ratios obtained when community involvement was added to the model. Community involvement was very significantly related to return to work after retirement, showing an odds ratio of 1.41 (p < .001). The odds ratio and level of significance of the focal interaction can be seen to have decreased even further (OR = 1.36, p < .10); this interaction held only marginal statistical significance. This was manifested in even closer probabilities of return to work after retirement between men who were in a marital/common-law relationship and men who were not. Table 7 shows that while men who were in a marital/common-law relationship showed a likelihood of return to work of 36.6 per cent, those not in a marital/common-law relationship showed a probability of return to work of 33.5 per cent. The corresponding likelihoods for women were 24.2 per cent and 26.9 per cent, showing minimal further contraction of the difference in probability of return to work after retirement between the two groups.
Table 6: Logistic regression analysis of return to work after retirement: Model 3, n = 4,414
Note: CI = confidence interval.
^ p < .10, * p < .05, ** p < .01, *** p < .001.
Table 7: Gender x relationship status probabilities of return to work after retirement: Model 3, n = 4,414
Because the parallel lines assumption did not hold for the three ordinal categories of highest level of education, tables 8 and 9 involve education divided between those with and those without a high school diploma, while tables 10 and 11 involve education divided between those with and those without a university degree. We will here discuss only those results pertaining to our focal interaction.
Table 8: Logistic regression analysis of possessing a high school diploma, n = 4,414
Note: CI = confidence interval.
^ p < .10, * p < .05, ** p < .01, *** p < .001.
Table 9: Gender x relationship status probabilities of possessing a high school diploma, n = 4,414
Table 10: Logistic regression analysis of possessing a university degree, n = 4,414
Note: CI = confidence interval.
^ p < .10, * p < .05, ** p < .01, *** p < .001.
Table 11: Gender x relationship status probabilities of possessing a university degree, n = 4,414
Among our central independent variables and focal interaction, Table 8 shows that only our focal interaction holds statistical significance. The interaction between being a woman and not being in a marital/common-law relationship achieved an odds ratio of 1.60, significant at the p < .05 level. Accordingly, Table 9 shows that, concerning having achieved a high school diploma, whereas men who were in a marital/common-law relationship held a likelihood of 72.7 per cent, men who were not in a marital/common-law relationship held a likelihood of 68.6 per cent, and although women who were in a marital/common-law relationship held a likelihood of 75.4 per cent, women who were not in a marital/common-law relationship held a likelihood of 79.4 per cent.
Table 10 reveals that among those who were in a marital/common-law relationship, women were less likely to hold a university degree (OR = 0.60, p < .001). Among men, those who were not in a marital/common-law relationship showed a marginally lower likelihood of holding a university degree (OR = 0.75, p < .10). The interaction between being a woman and not being in a marital/common-law relationship achieved marginal statistical significance (OR = 1.47, p < .10). This is manifested in the results in Table 11 that revealed that among men, being in a marital/common-law relationship was associated with a higher likelihood of holding a university degree (30.5% vs. 25.1% among those who were not in a martial/common-law relationship), whereas among women, being in a marital/common-law relationship was associated with a somewhat lower probability of being a university graduate (21.2% vs. 22.8% among those who were not in a marital/common-law relationship).
Table 12 reveals that among those who were in a marital/common-law relationship, women were marginally more likely than were men to be involved in the community (OR = 1.16, p < .10). Not being in a marital/common-law relationship among men was associated with a lower likelihood of community involvement (OR = 0.58, p < .001). The interaction between being a woman and not being in a marital/common-law relationship achieved a very significant odds ratio of 1.89 (p < .001). This is manifested in Table 13 that shows that among men, being in a marital/common-law relationship suggested a probability of community involvement of 50.8 per cent, whereas not being in a marital/common-law relationship implied a likelihood of 38.2 per cent that one would be involved in community groups. The respective likelihoods for women showed scarcely any effect of relationship status (54.2% and 56.4%).
Table 12: Logistic regression analysis of community involvement, n = 4,414
Note: CI = confidence interval.
^ p < .10, * p < .05, ** p < .01, *** p < .001.
Table 13: Gender x relationship status probabilities of community involvement, n = 4,414
Discussion and Conclusion
This study has investigated correlates of return to work after retirement among Canadians aged 50 to 74 in 2007 who retired at least once, and its results can inform policy and programs concerning retirement, pensions, and work in the later years of life. This study adds to past research that has examined the relationship between gender and return to work after retirement by considering how gender interacts with relationship status at the point of retirement in framing the probability that individuals will engage in post-retirement paid work.
Discussion of Results Pertaining to Control Variables
Our results confirm the work of other scholars on how financial circumstances, current age, age at retirement, health, education, and parental status are related to post-retirement work. Although worse financial circumstances are related to the financial need to engage in paid work after retirement, they are also linked with lower levels of education. Higher levels of education are associated with the desire to engage in paid work after retirement as a means of remaining active and engaged, as well as being linked with skills that make one qualified for a diverse array of jobs and with access to contacts that facilitate the finding of employment (Griffin & Hesketh, Reference Griffin and Hesketh2008). Controlling for levels of education leads to lower levels of financial adequacy being associated with substantially higher odds ratios that achieve higher levels of statistical significance. In other words, a suppressor effect is revealed: were it not for the higher levels of education associated with higher levels of financial adequacy, higher levels of financial adequacy would be associated with an even lower likelihood of post-retirement paid work. Related to this discussion, Cahill et al. (Reference Cahill, Giandrea and Quinn2006) have argued that there is a U-shaped relationship between income level and bridge employment. Those of low income tend to engage in bridge employment for financial reasons, whereas those of high income tend to perform bridge work for intrinsic life satisfaction (Cahill et al., Reference Cahill, Giandrea and Quinn2006). The results of simultaneously studying financial circumstances and levels of education can be related to this U-shaped relationship. Those at the lower end who require a return to paid work after retirement for financial reasons are identified by the financial adequacy variable. Those at the upper end who desire a return to paid work after retirement in order to remain active and in order to maintain high levels of life satisfaction are identified by the education variable.
It should be borne in mind, however, that our measure of financial adequacy is based on a rather subjective assessment. In answering the GSS-21 (Statistics Canada, 2007a) question about financial adequacy at the point of retirement, respondents might have compared themselves with others in similar circumstances. As such, because of measurement difficulties, our inclusion of the financial adequacy variable may be an incomplete control.
In accordance with scholars who have found that high levels of health are linked with a higher propensity to be involved in paid work (Cahill et al., Reference Cahill, Giandrea and Quinn2006; van der Giezen, Bouter, & Nijhuis, 2000; Griffin & Hesketh, Reference Griffin and Hesketh2008; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011; Maestas, Reference Maestas2010; Statistics Canada, 2006), we have found that higher levels of health are linked with a greater likelihood of return to work after retirement. Our study also confirms the claim made by Statistics Canada (2006) that those who first retire before turning 60 years of age are more likely to return to paid work, and that this might be because they are not adequately psychologically prepared for retirement. Our study also confirms the finding that younger chronological age is linked with a higher probability of post-retirement paid work (Davis, Reference Davis2003; Lefebvre et al., Reference Lefebvre, Merrigan and Michaud2011). This might be linked with the fact that the norms of recent times encourage a “busy ethic” among older persons, according to which older persons should stay vigorous and active (Warr, Butcher, & Robertson, Reference Warr, Butcher and Robertson2004). Because younger segments of the older population are typically quite capable of being active and productive, present-day norms might provide them with strong encouragement to return to paid work after retirement.
We found that parental status is associated with a higher likelihood of return to work after retirement. Even though their children might be living independently, older parents might be inclined to return to paid work after retirement in case an unexpected event results in their adult children’s requiring financial support. In this regard, it seems that after children have left the home, a big part of the responsibility that parents feel towards children is financial.
Discussion of Results Pertaining to Gender and Relationship Status
The present study makes some notable contributions to the literature on the divergent work patterns undergone by men and women. Its design allows for the isolation of important factors within these divergent patterns: labour market engagement, motivation, and career orientation. Our results present much evidence that levels of engagement among men are higher among those who are in a marital or common-law relationship, and that levels of engagement among women are higher among those who are not married or in a common-law relationship. Because financial adequacy has been controlled, our study investigated probabilities of older persons returning to work after retirement for reasons other than financial need. Alternative motivations for post-retirement work are likely to include the desire to remain active and involved. The fact that we have found community involvement to be highly positively related to post-retirement paid work suggests that some of the same motivations underlie these two categories of activity. The present study thereby shows how gendered truths are revealed through an analysis of a very specific employment phenomenon: return to paid work after retirement.
The present study concurs with the work of scholars who have emphasized that the experience of family circumstances differs by gender (Calasanti, Reference Calasanti2004; Kim & Feldman, Reference Kim and Feldman2000; Pleau, Reference Pleau2010; Powell & Mainiero, Reference Powell and Mainiero1992; Szinovacz & DeViney, Reference Szinovacz and DeViney2000; Szinovacz et al., Reference Szinovacz, DeViney and Davey2001). Among men, those who are not in a marital or common-law relationship are less likely to engage in post-retirement paid work. One reason for why this is the case is that depending on the circumstance, men who are not in a serious relationship might have less need to work for pay because they are supporting a one-person household. As we find that men who are not in a serious relationship are less likely to return to work after retirement even after controlling for financial adequacy, we draw on scholarship that has identified other possible explanations for this empirical finding. Numerous scholars emphasize that men who have never married tend to be of lower levels of occupational success than those who have married (Leslie et al., Reference Leslie, McClure and Oaxaca1998; Nakosteen & Zimmer, Reference Nakosteen and Zimmer1997). These assertions suggest that men not in a marital or common-law relationship, as a group, might be less motivated to achieve status and financial rewards within the larger social structure. These lower levels of motivation could imply a lower likelihood of returning to paid work after retirement.
Our review of the literature elaborated upon the debate concerning whether men who are married are more successful than those who are not married because of the advantages that marriage brings (Chun & Lee, Reference Chun and Lee2001) or because of the type of men who are selected into marriage (Nakosteen & Zimmer, Reference Nakosteen and Zimmer1997). Our findings can be interpreted as offering support for either argument. We find men who are married or in a common-law relationship to be more likely to return to work after retirement, to be involved in community groups, and to have higher levels of education. Furthermore, the gap between men who were in a marital or common-law relationship at the time of first retirement and men who were not in any such relationship when they first retired in likelihood of engagement in post-retirement work narrows after the control of levels of education and of community involvement, suggesting that different levels of engagement are a substantial source of these divergent trends in post-retirement paid work. Although the effects of controlling for levels of education support the selection theory of Nakosteen and Zimmer (Reference Nakosteen and Zimmer1997), it might also be the case that men in a marital or common-law relationship at the point of retirement might experience the benefits of their partners taking care of the home, allowing these retired men more time and energy for involvement in the community and in paid work.
Studies that have linked lower levels of income among men with likelihood of divorce (Kalmijn & Poortman, Reference Kalmijn and Poortman2006) provide another explanation for our findings concerning men. Men with higher occupational success and higher incomes are less likely to divorce, and by extension, to lose a common-law partner. This implies that higher earning men might have a greater likelihood of reporting being in a serious relationship at the time of retirement. The fact that we have found men in a marital or common-law relationship to be more likely to return to work after retirement than men not in a marital or common-law relationship even after the control of financial adequacy can be explained by the fact that men’s financial circumstances substantially improve after the dissolution of either a marital or cohabiting relationship (Sayer, Reference Sayer, Fine and Harvey2006). This is largely an outcome of the fact that after the dissolution of serious relationship, men are often not required to support others to the same extent as before. This is in stark contrast to women whose financial circumstances tend to worsen following the end of a serious relationship. This is largely an outcome of the fact that despite women’s economic progress over time, women continue to be substantially dependent on their male partners for economic support (Sayer, Reference Sayer, Fine and Harvey2006). A divorced man might report good financial adequacy at the point of first retirement even though earlier financial difficulties precipitated the dissolution of his most recent marriage. Despite not requiring a return to work after retirement for financial reasons, he might be less likely to show the levels of engagement and motivation that earlier in life might lead one to attain higher levels of education, and later in life might lead one to become involved in community groups and to return to paid work after retirement.
Sayer (Reference Sayer, Fine and Harvey2006) did not distinguish between younger and older adults when she stated that marital dissolution worsens women’s financial circumstances while improving the financial circumstances of men, and the explanations she provided for these gendered effects allow these effects to be generalized to both younger and older adults. Although Sayer (Reference Sayer, Fine and Harvey2006) highlighted how declines in divorced mothers’ financial circumstances were affected by their greater likelihood of keeping custody of their children, she emphasized how mothers’ economic declines subsequent to divorce were also linked with broader gender inequalities in income and employment opportunities, as well as with the obstacles to labour force participation throughout women’s adult life course created by having children. Divorced women often can no longer depend on financial support from their former husbands who might have attained higher levels of economic success throughout their adult life course (Sayer, Reference Sayer, Fine and Harvey2006). As such, Sayer’s (2006) analysis of how divorce affects women’s financial circumstances is applicable to women who have never had children, as well as to younger mothers taking care of dependent children and older mothers whose children have begun their independent lives. Although Sayer (Reference Sayer, Fine and Harvey2006) explained that men’s improved financial circumstances after divorce were partly because their economic support of children post-dissolution tended to be lower than pre-dissolution, she stated that these improved financial circumstances also resulted because men profit from socioeconomic advantages throughout the adult life course. No longer being required to share their wealth with their former wives improves the financial circumstances of many divorced men (Sayer, Reference Sayer, Fine and Harvey2006). Therefore, Sayer’s (2006) analysis of how divorce tends to be linked with improved economic circumstances for men can be applied to men who have never had children, as well as to younger fathers whose children have yet to reach adulthood and older fathers whose children are no longer dependent.
There is an alternative interpretation of how it is that we found men in a marital or common-law relationship at the point of retirement to be more likely to return to paid work even when financial adequacy was controlled. Kalmijn and Poortman (Reference Kalmijn and Poortman2006) extended their ideas by linking financial difficulties to economic strains that led to higher levels of conflict in a marriage that can precipitate marriage dissolution. The fact that we have found men in a marital or common-law relationship to be more likely to return to work after retirement than men not in any such relationship even with financial adequacy controlled suggests that the characteristics of men who are able to achieve stability in a serious relationship might extend beyond the ability to secure financial resources. Other personal traits, perhaps indicated by variables such as extent of community involvement (altruism and engagement) and levels of formal education (cognitive and communication skills), might play an important role in whether a man is able to achieve stability, and thus continuity, within a serious relationship.
A further possible contributor to this finding is that men in a marital or common-law relationship at the point of first retirement might feel the need to return to paid work in order to maintain their status within their relationship (Szinovacz & DeViney, Reference Szinovacz and DeViney2000). This same drive for status could motivate one earlier in life to achieve higher levels of education, and later in life to achieve status in the community through involvement in community organizations.
Unlike men not in a marital or common-law relationship, women who were not in marital or common-law relationship at the point of first retirement were found to be more likely to return to paid work after retirement. Numerous scholars have attributed the higher rates of return to work after retirement of non-married women to their precarious financial circumstances (Davis, Reference Davis2003; Perkins, Reference Perkins1994; Pleau, Reference Pleau2010). However, we have found that women who were not in a marital or common-law relationship when they first retired are more likely to return to work after retirement even after controlling for financial adequacy at retirement. Although financial circumstances certainly play a big role in the return of non-married and non-cohabiting women to paid work after retirement, we show that there are other forces at work. Our results concur with those of scholars who have emphasized how family involvements are associated with lower career success among women (Blair-Loy, Reference Blair-Loy2003; Ellemers et al., Reference Ellemers, van den Heuvel, de Gilder, Maass and Bonvini2004; Kim & Feldman, Reference Kim and Feldman2000; Leslie et al., Reference Leslie, McClure and Oaxaca1998). Thus, as Blair-Loy (Reference Blair-Loy2003) suggested, those women who are most career motivated and ambitious might be more inclined to avoid marriage and parenthood. This commitment to socioeconomic success might find expression in return to work after retirement.
Furthermore, the “independence” hypothesis proposes that more financially successful women will be less dependent on their relationships for economic reasons, and thus more likely to exit a serious relationship if it is not bringing them the emotional satisfaction that they desire (Boyle et al., Reference Boyle, Kulu, Cooke, Gayle and Mulder2008; Dribe & Lundh, Reference Dribe and Lundh2012; Kalmijn & Poortman, Reference Kalmijn and Poortman2006; Lyngstad & Jalovaara, Reference Lyngstad and Jalovaara2010). However, we have found that women not in a marital or common-law relationship are more likely to return to paid work after retirement even after the control of financial adequacy at the time of first retirement. This suggests that a woman’s level of independence extends beyond financial independence. Some women may not feel dependent on a marital or cohabiting relationship even if they experience financial difficulties upon ending their relationship. This internal sense of independence, which extends beyond feelings of financial adequacy, might be related to educational attainment earlier in life, and to propensity to become involved in the community later in life.
Our results show that women not in a marital or common-law relationship at first retirement are slightly more likely to hold a high school diploma, to hold a university degree, and to be involved in community groups. Furthermore, control of levels of education and of community involvements slightly narrows the gap between women who were in a marital or common-law relationship at the time of retirement and women who were not in any such relationship when they first retired in likelihood of involvement in post-retirement paid work. As is the case for men, this suggests that levels of engagement and motivation are among the factors leading to the differing likelihoods of return to work after retirement for women in a marital or cohabiting relationship when they first retired and for women not in a marital or common-law relationship at the time of first retirement.
An alternative explanation is that retired women in a serious relationship might be pressured by their partners to not return to paid work and to not be involved in community groups in order for their partners to maintain their status within their relationship (Szinovacz & DeViney, Reference Szinovacz and DeViney2000). Women not in a serious relationship do not experience this pressure, and thus have greater freedom to return to paid work and to involve themselves in community organizations. Knowledge of these gender-based relationship power dynamics might be a factor in the decision of some ambitious and career-oriented women to avoid marriage.
The following passage from Mirowsky and Ross (Reference Mirowsky and Ross2003) fits well with our results:
Men have higher autonomy and earnings, less responsibility for household work, and lower amounts of other household income. Because of the differences in these factors, employment increases the expected sense of control most for married males, followed by nonmarried males, then nonmarried females, and finally married females. … Adjusting for household income, nonmarried women have a significantly greater sense of control than both men and married women: ordered from lowest to highest sense of control are married females, nonmarried males, married males, and nonmarried females. (p. 192)
The present study shows that the findings of Mirowsky and Ross (Reference Mirowsky and Ross2003) concerning gender and paid and unpaid work earlier in life hold for return to work after retirement.
Policy Implications
Our results hold some notable policy implications. Efforts to retain older individuals in the workforce as a means of preventing shortages of skilled workers as the population ages are shown here to be more likely to succeed if current trends towards widespread high levels of education continue. If effective health care and public health policies succeed in maintaining the health and vitality of older persons, we can expect older individuals to continue their paid work endeavours until later in life. Most significantly, the present study suggests the need to establish greater equality between the genders. Our results indicate that present-day society still pressures women to choose between commitment to employment and commitment to family. They also suggest that men, on the other hand, have more freedom to enjoy the full benefits of both. If there were no significant interaction between gender and relationship status in propensity to return to work after retirement, in levels of education, and in involvement in community groups, this would indicate that a greater level of equality between women and men has been achieved in both the public and private spheres.
Implications for Future Research on Intimate Ties
Our study also encourages researchers to consider cohabiting couples as well as married couples when assessing how intimate ties among adults are related to various personal and social outcomes. The present-day situation is marked by cohabitation having become much more common. Future research should consider studying marriage and cohabitation together or studying them both separately. With increasing rates of adults living alone, we have cause to contrast, as the present study has done, those in an intimate relationship with those not involved in a romantic union.
Limitations
Some limitations of our study deserve mention. It is worth repeating that because of measurement difficulties, our control for financial adequacy may have been incomplete. In answering the GSS-21 (Statistics Canada, 2007a) question about financial adequacy at the point of retirement, respondents might have subjectively compared themselves with reference groups composed of persons in similar life situations.
A further limitation is that our study does not address retirement legislation. Notably, decisions to retire, to remain retired, or to return to work are influenced by structural, personal, and policy-related issues. Public pension policies in Canada, for instance, determine the age of eligibility for benefits, and this influences decisions about retirement and return to work. Furthermore, the abolition of mandatory retirement in Canada came around the time that the data used in this study were being collected. Because this legislation was implemented in different years in different provinces, our findings could have varied regionally. However, the GSS-21 (Statistics Canada, 2007a) does not include any variable that indicates region of residence at the point of retirement. Although there is a variable for region of birth, it is a far stretch to assume that a respondent lived in the same region from birth to retirement. Regional analyses are thus impossible. Future investigations would benefit from integrating regional differences in legislation within the research design.
Another limitation is that our study did not address possible relationship transitions subsequent to first retirement. It is possible that a relationship transition after first retirement, such as a divorce, might have affected whether an older individual decided to return to paid work. This is a limitation we could not avoid with the data source we were using. Future research would benefit from a more detailed delineation of relationship trajectories both before and after first retirement, preferably with longitudinal analysis, which could capture the timing and transition of marriage, widowhood, divorce, and other serious relationship formation and dissolution, in relation to the timing of retirement and return to work.
Data limitations precluded investigation of the characteristics of the career employment that respondents left, nor did we examine the characteristics of the employment undertaken by respondents after their first formal retirement. As occupation before retirement was not available in our data source, levels of formal education served as its proxy. It is important to consider whether a respondent’s previous workforce involvement had been fragmented and filled with part-time work, or stable and characterised by long duration full-time work. It is also important to consider whether the employment undertaken by respondents after they first retired was a casual part-time job of short duration versus a more substantial and long duration return to paid employment. It would be of interest to conduct an analysis of the gendered implications of patterns of workforce involvement before and after one’s self-identified first retirement.
Paths for Future Research
A good path for future research would be to conduct an analysis similar to the one presented here concerning both those returning to work after retirement and those who have chosen not to retire despite having passed the typical retirement age. At the typical retirement age, those who had decided to keep working might have been in work circumstances similar to the immediate pre-retirement work circumstances of those individuals who chose to retire. Accordingly, a selectivity effect based on decision to retire rather than remain employed might have affected our results. Still, we suspect that our results might be similar if we were to repeat the present analysis with those who chose to remain employed past the typical retirement age. Both those who returned to paid work after retirement and those who continued to work past the typical retirement age might have been engaged in paid work for financial reasons. Because we controlled for financial adequacy at the point of retirement in our study, our results might be at least partially driven by the fact that for many older persons, retirement did not provide the extent of stimulation and engagement that they had expected. Then, too, it is likely that many individuals who continued to work past the typical retirement age did so because they enjoyed the stimulation and challenge of paid work. As such, for both financial and intrinsic reasons, we expect that return to work after retirement and continued paid work beyond the typical retirement age might be similar in their predictors. Addressing this possibility would be an interesting future project.
Furthermore, studies such as the present one should be repeated in the future in order to test whether the gender differences here revealed are lessening over time as the participation of women in paid work and as the benefits that they thereby acquire converge with those of men (Hofmeister & Blossfeld, Reference Hofmeister, Blossfeld, Blossfeld and Hofmeister2006). The respondents in the present study were at least 50 years of age, and so the patterns of gender inequality here revealed might result from the fact that these respondents came of age during more gender inegalitarian times. It is expected that future cohorts of retired men and women will be more similar in their patterns of bridge employment (Pleau, Reference Pleau2010).
