Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Retiring together, or not

By Grant Schellenberg and Yuri Ostrovsky

Full article in PDF format

Retirement continues to change in many ways.This is certainly evident in terms of its timing, given the declining rate of labour force participation among older men between the 1970s and mid-1990s and its reversal in more recent years (Marshall and Ferrao 2007). Likewise, the process appears to be changing, given phenomena such as post-retirement employment (Schellenberg et al. 2006), phased retirement and diverse pathways into retirement (Nouroz and Stone 2006). Retirement is also changing in the extent to which it is being navigated by dual-earner couples.

Throughout much of the 20th century, older couples faced only one retirement decision—the husband's. Women who had paid employment during their life typically left the workforce at an early age to care for children and work on an unpaid basis in the home. However, with the dramatic rise and sustained participation of women in the paid labour force since the 1970s, retirement transitions of married couples have been transformed. Increasingly, couples must make two decisions rather than just one and must balance the preferences and constraints of partners who both make substantial contributions to household income.

This has added new complexities to retirement decisions. Researchers generally agree that couples prefer to retire together, in large part because retirement is more enjoyable when it can be shared with a spouse (Gustman and Steinmeier 2004, An et al. 2004, Moen et al. 2001, Szinovacz and Davey 2005). However, the opportunity to retire 'jointly' may be constrained by factors such as age differences, health conditions, pension eligibility, job loss and career aspirations.

To date, evidence on spousal retirement transitions in Canada has been sparse. In the mid-1990s, about one-third of couples retired within one year of each other (Gower 1998). More recently, about half of couples approaching retirement intended to retire at the same time (Schellenberg et al. 2006). However, trends in the actual retirement outcomes of spouses have yet to be documented.

This article addresses several questions regarding retirements in dual-earner couples: the extent to which these spouses synchronize the timing of their retirements; the factors associated with taking one or another spousal pathway into retirement; and changes in spousal patterns of retirement through the 1990s (see Data sources and definitions).

With more wives employed, retirement becomes more complex

Most Canadians approaching retirement are married, which has changed little over the last 30 years. Between 1976 and 2006, the proportion of women aged 55 to 64 who were married or in a common-law relationship remained just over 70% (Table 1), while men in such relationships remained over 80%.3 The most noticeable changes in marital status have been the increase in separated or divorced and the decline in widowed. For individuals approaching retirement without a spouse, this is usually the case due to divorce or separation rather than widowhood or never having married. (The retirement characteristics of these individuals are outside the scope of this article.)

The employment histories of women in these couples have changed markedly. In 1976, almost one-half of married women aged 55 to 64 were not in the labour force after age 40 (Table 2). Over one-quarter (27%) had never worked in a paid job and another 19% had last worked before 40. Given their limited involvement in the paid labour force, these women did not retire in the usual sense nor was their paid employment a consideration in their husband's retirement decision.

In 2006 the situation was very different. Less than 4% of married women aged 55 to 64 had never worked and only 10% had last held a paid job before age 40. Instead, the vast majority (77%) were either currently employed (48%) or had held a paid job at age 50 or older (29%). In short, most married women now retire from paid employment and most married couples face the possibility of joint retirement.

Most wives retire after husbands

Among dual-earner couples in the 2001 retiring cohort, 29% of the spouses retired within two years of each other (Chart A). This includes couples in which both retired in 2001 (14%) and those in which one spouse retired in 2001 and the other the following year (15%).4 The incidence of joint retirement generally falls within a range of about 20% to 40% (Blau 1998, Hurd 1990, O'Rand and Farkas 2002, Johnson 2004, Gower 1998).5 Furthermore, the tendency for joint retirement is supported by the clustering of retirements within a two-year period. But while a significant proportion of couples retired jointly, the most prevalent pattern was for women to retire after their husband. In 28% of retiring couples, the husband retired in 2001 and the wife had still not retired by 2005. In another 15% of couples, the wife retired two to four years after the husband. However, more than a quarter (28%) of wives retired first. This includes 11% of couples in which the wife retired in 2001 and the husband retired two to four years later, and another 17% in which the husband had not retired by 2005.

Some trends are evident (Chart B). Between 1986 and 2001, the proportion of dual-earner couples in which both partners retired within two years of each other declined by 2 percentage points. Furthermore, the proportion of wives retiring two to four years after their husband declined by just over 4 points, as did the proportion of husbands retiring two to four years after their wife. This 11-point decline in the middle of the distribution was offset by increases at the ends. Between 1986 and 2001, the proportion of wives retiring five or more years after their husband increased by 7 points, while the proportion of husbands retiring five or more years after their wife increased by 4 percentage points. The same patterns were evident for the 1991 or 1996 retiring cohorts. Overall, this suggests that spousal retirement is becoming more disjointed.

Many factors influence spousal retirement

The probability of following one or another of these pathways into retirement is likely influenced by many factors, including age differences between spouses. To assess the relative importance of various factors, a multivariate model was constructed. The 1991, 1996 and 2001 retiring cohorts were pooled and a set of characteristics was introduced to determine their relationship to the likelihood of retiring in one way or another. The marginal effects of these characteristics show how much the predicted probability of taking a given spousal pathway into retirement changes when a specific characteristic is changed by a small amount (Table 3). For example, the model yields a predicted probability of joint retirement of 35.1% (that is, spouses retiring within two years of each other). A one-year increase over the average age of husbands increases the probability by 3.2 percentage points, with other characteristics remaining constant. (Pension contributions prior to retiring—an important variable—were not available for the 1986 cohort, so multivariate analysis was restricted to the 1991, 1996 and 2001 cohorts.)

The first three variables were the age of the husband, the age difference between the spouses, and whether the couple was legally married or in a common-law relationship.

The likelihood of spouses retiring jointly is greater among older than younger couples. A one-year increase in the age of the husband decreases the likelihood that either spouse retires five or more years after the other by about 2 percentage points, and increases the likelihood of joint retirement by 3.2 points. Perhaps not surprisingly, if a husband in a dual-earner couple is 65, his or his wife's retirement is likely to be accompanied (or closely followed) by the other's retirement. In contrast, if the husband is 55, his or his wife's retirement is less likely to be accompanied by the other's retirement. The other spouse is more likely to continue working.

The age difference between spouses also matters. A one-year increase in the age difference reduces the predicted probability of joint retirement by 2.1 percentage points and increases the probability of a wife retiring five or more years after her husband by 3.7 points. In short, a wife who is much younger than her husband is more likely to continue working after he retires than a wife who is about the same age as or older than her husband. The retirement patterns of couples in common-law relationships were not significantly different from those legally married.

Events en route to retirement may also influence spousal transitions. The loss of a job, for example, may force one spouse into retirement prematurely and reduce the prospects for joint retirement. Exposure to job loss was included in the model using a yes/no variable indicating if either spouse had received Employment Insurance (EI) benefits in the year prior to retirement.6 The receipt of EI benefits was significantly associated with spousal retirement patterns. Husbands and wives receiving such benefits were far more likely to retire before their spouse than those who did not receive them. For example, the predicted probability of a wife retiring five or more years after her husband increased by 11.1 percentage points if he received EI benefits prior to retiring. Likewise, the predicted probability of a husband retiring five or more years after his wife increased by 8.1 points if she received EI benefits prior to retiring. One interpretation is that when one spouse enters retirement via unemployment, the other continues working to shore up their financial resources. Indeed, in families with no working-age children, the earnings of wives increased following the layoff of their husband, offsetting approximately 22% of the husbands' earnings losses (Morissette and Ostrovsky 2008). Interestingly, while a husband's receipt of EI benefits decreased the likelihood of joint retirement, a wife's receipt of EI benefits increased the likelihood. It is not clear why this is the case.

Financial characteristics were also important. Average earnings of husbands and wives prior to retirement were correlated with spousal retirement transitions. Specifically, compared with those earning less than $15,000, husbands and wives earning $45,000 or more were significantly less likely to continue working five or more years after their spouse's retirement (decreases in predicted probabilities of 4.9 and 3.7 percentage points respectively). Conversely, husbands and wives with incomes of $45,000 or more were significantly more likely to retire jointly, with the predicted probability increasing by 5.7 and 4.4 points respectively.7 This is consistent with other studies (O'Rand and Farkas 2002) that found higher-income couples more likely to retire together.

The wife's contribution to a couple's total earnings prior to retirement was correlated with spousal retirement patterns. Specifically, a one percentage point increase in the wife's contribution to pre-retirement earnings was associated with a 0.2-point increase in the predicted probability that she would retire five or more years after her husband. One might speculate that wives who contribute a larger share of income shoulder greater responsibility for the financial well-being of the household and hence have greater incentive to continue working. However, the LAD does not provide information to test this hypothesis.

Furthermore, whether husbands and wives in dual-earner couples made pension contributions prior to retirement was significantly correlated with retirement transitions. Specifically, compared with wives not making pension contributions, those who did so were significantly more likely to continue working after their husband's retirement and significantly less likely to retire first. For example, the predicted probability of a wife retiring two to four, or five years after her husband increased by 5.9 and 1.6 points respectively if she contributed to a pension.

Finally, the results of the multivariate model showed that, holding the characteristics discussed above constant, the predicted probability of dual-earner spouses retiring within two years of each other declined by 4.4 percentage points between 1991 and 2001, and the likelihood of retiring two to four years apart declined by about 2.5 points. In contrast, the likelihood of a wife retiring five or more years after her husband increased by 5.3 points and the likelihood of a husband retiring five or more years after his wife increased by 4.1 percentage points. This trend also held during the latter half of the decade, as husbands and wives were significantly more likely to retire five or more years after their spouse in 2001 than in 1996.


As a result of the widespread entry and sustained participation of women in the paid labour force, many Canadians now approaching retirement are part of a dual-earner couple. As such, the timing of their retirement can be assessed not only in terms their age, but also relative to the timing of their spouse's retirement. And just as the age of retirement has changed considerably, so too has the sequencing of retirement in dual-earner couples. Overall, evidence indicates that the retirements of such couples became increasingly disjointed through the 1990s.

From a research standpoint, one implication of this study is that identifying spousal retirement patterns simply as wife first, husband first or joint may miss an important part of the bigger picture. The growing disjointedness of spousal retirement is attributable to the declining proportions of husbands and wives retiring two to four years after their spouse and the increasing proportions retiring five or more years after. This shift would be obscured using broad wife-first/husband-first categories.

With the imminent retirement of the baby boom generation, much discussion focuses on how older workers might be encouraged to stay on the job (OECD 2006), and retirement incentives and constraints imposed by public programs, pension rules and workplace policies (OECD 2005). Spousal factors are another consideration, as an increasing proportion of older workers might take the plans and preferences of their partner into account when making their own retirement decisions. Such considerations could have either a positive or negative impact on labour supply. With husbands generally two to three years older than wives, a preference to retire jointly could be realized with wives leaving the labour force a few years earlier than they would have left otherwise, or with husbands working a few additional years. However, the potential impact of spousal considerations could be mitigated by the increasingly independent manner in which spouses in dual-earner couples appear to be retiring.

Data sources and definitions

This study uses a 20% version of the Longitudinal Administrative Database (LAD), which is derived from taxation data. LAD files provide detailed information about both individual and family income for those who filed income tax forms between 1982 and 2005. The 20% sample is randomly selected from all tax-filing Canadians and, once selected, individuals remain in the sample for as long as they appear on the annual T1 Family File (T1FF). Census families are formed from the personal data that filers provide on other family members. Filers are attached to their spouses (legal or common-law) by social insurance number or by matching age, sex, address and marital status. Baseline labour force information comes from the monthly Labour Force Survey, which covers the civilian, non-institutionalized population in the 10 provinces.

Dual-earner couples approaching retirement are defined by identifying those with a husband 55 to 69 years of age. The sample is limited to couples in which both partners derive their earnings primarily from paid employment rather than self-employment and have average annual earnings of $2,000 or more over at least three consecutive years prior to the retirement of one or both.1

Retirement can be defined in various ways, depending in part on the information available (Bowlby 2007). While the LAD provides a great deal of income detail, it contains limited information on demographic and labour market characteristics. Consequently, retired individuals are identified on the basis of changes in their income characteristics over time—more specifically, when their annual earnings decline to less than 10% of their average during the three previous years and remain below that level over the next five years. For example, an individual with annual earnings of $75,000 over three years would be identified as retired if annual earnings dropped below $7,500 and subsequently remained below that amount. The definition allows for the possibility that retirees might maintain some involvement in paid employment and also recognizes that some people might 'come out of' retirement.2

Using these criteria, the retirement patterns of dual-earner couples in which at least one spouse retired in 1986, 1991, 1996 or 2001 were identified. For the 2001 retiring cohort, for example, both spouses had earnings of $2,000 or more in 1998, 1999 and 2000 (i.e. they were a dual-earner couple) and in 2001 the earnings of at least one spouse fell below the 10% threshold. Retirees whose spouse's earnings fell below the 10% threshold in 2001 or later were also identified.


  1. Identifying whether self-employed individuals are retired, based on changes in their net self-employment incomes, could be problematic. For example, a self-employed individual who is actively working may report negative or zero net self-employment income resulting from business losses or expenses. Hence, a year-over-year decline in net self-employment income might not necessarily signal labour force exit. Conversely, an individual receiving net income from self-employment might no longer be actively engaged in the workforce.
  2. Two additional definitions of retirement were used in the early stages of analysis. Individuals were identified as retired when their annual earnings declined to zero following at least three consecutive years of $2,000 or more. Once identified as retired, earnings were not tracked in subsequent years to determine if they became positive again (i.e. the individual came out of retirement). Individuals were also identified as retired when annual earnings declined to less than 10% of their average during the three years prior to retirement. This definition allowed for the possibility of some limited involvement in paid employment after their initial retirement. Once individuals were identified as retired, their earnings were not tracked in subsequent years. Trends in the timing of spousal retirement based on these definitions were comparable to those in the body of the article.
  3. The Labour Force Survey does not provide information on marital history so the proportion who had divorced and remarried cannot be determined.
  4. The precision of the estimate of relative timing of each spouse's retirement is limited by the annual earnings data. If one spouse retires at the end of January, his or her annual earnings for the year will likely fall below the 10% threshold and that spouse will be identified as a retiree. If the other spouse retires at the end of March the same year, his or her annual earnings will not likely fall below the 10% threshold, until the following year. Consequently, the spouses will be identified as having retired in two consecutive years, when in fact the dates were only two months apart.
  5. Differences within this range are attributable to factors such as different data sources, sample selection criteria, age cohorts, reference periods and definitions of retirement. Blau tracks the labour force exits of a sample of persons born between 1906 and 1911 and estimates that 30% to 40% of spouses in dual-earner couples retired within one year of each other. Hurd examines a slightly younger cohort and estimates the incidence of joint retirement at about 25%, while O'Rand and Farkas track women in their 50s and early 60s from 1989 to 1997 and estimate the incidence of joint retirement at 33% to 39%. Johnson uses 1992 to 2002 and estimates the incidence of joint retirement at 19%. In one of the few Canadian studies, Gower estimates that about one-third of dual-earner spouses leave the labour force within one year of each other.
  6. Individuals actively seeking employment are counted as unemployed by the Labour Force Survey. This could include older workers who come out of retirement or who look for work after leaving a career job. Such individuals may not have applied or qualified for Employment Insurance benefits, so they would be missed by the EI benefit variable. Consequently, the EI benefit variable is likely a weak proxy for unemployment. Furthermore, EI eligibility rules were tightened in the early 1990s so the receipt of benefits may be a better proxy for job loss in 1996 and 2001 than it was in 1991.
  7. Simple cross tabulations show that wives with earnings of $45,000 or more are far more likely to retire five or more years after their husband than wives in lower earnings categories. However, this bivariate correlation disappears when other characteristics, such as pension coverage and receipt of EI benefits, are taken into account.



The authors are with the Business and Labour Market Analysis Division. Grant Schellenberg can be reached at 613-951-9580, Yuri Ostrovsky can be reached at 613-951-4299 or both at

You need to use the free Adobe Reader to view PDF documents. To view (open) these files, simply click on the link. To download (save) them, right-click on the link. Note that if you are using Internet Explorer or AOL, PDF documents sometimes do not open properly. See Troubleshooting PDFs. PDF documents may not be accessible by some devices. For more information, visit the Adobe website or contact us for assistance.