|Home | About | Journals | Submit | Contact Us | Français|
Using data from the first seven waves of the Health and Retirement Study (1992 to 2004), the authors examined the extent to which joint retirement expectations were realized, the role of couple-level agreement in facilitating joint retirement, whether husbands' or wives' expectations were more likely to be realized in cases of disagreement, and factors associated with the realization of expectations. The results indicate that couples expecting joint retirement were over three times more likely to retire jointly than couples in which neither spouse expected to do so. However, the probability of joint retirement did not differ between couples in which both spouses expected to retire jointly and those in which only one spouse expected to do so. Wives' and husbands' expectations were equally strong predictors of joint retirement, and retirement age, health, spouses' relative earnings, and discussions of retirement were related to the likelihood of realizing joint retirement expectations.
Retirement has typically been viewed as an event experienced by men at the end of their careers. As such, most families experience only one retirement – that of the husband and breadwinner. However, this simple characterization of retirement is increasingly inappropriate as long-term increases in women's labor force participation across the life course have resulted in more couples with two retirements to coordinate. Joint, or synchronized, retirement of husband and wife has become a salient option for dual-worker couples and several studies have examined the trends, correlates, and consequences of joint retirement (Blau 1998; Gustman and Steinmeier 2000; Henretta, O'Rand and Chan 1993; Hurd 1988; Johnson 2004; Szinovacz 1989).
To date, however, little attention has been paid to couples' initial expectations regarding joint retirement and the extent to which these expectations are realized. As a result, there are several missing pieces in our understanding of joint retirement. We do not know whether it is an expected or planned retirement option, whether expectations about joint retirement are shared between spouses, and what factors are related to the realization of expectations. An increasing number of studies have examined relationships between retirement expectations and subsequent behavior at the individual level and conclude that expectations provide useful information for projecting future retirement outcomes (Anderson, Burkhauser and Quinn 1986; Bernheim 1987; Dwyer 2001). Do similar relationships hold for joint retirement? The relationship between expectations and behavior is more complex for joint retirement than at the individual level given the need to coordinate two retirements, each with its own institutionalized schedules. We are not aware of any existing research on the congruence between initial expectations regarding joint retirement and subsequent behavior. This is an important limitation given the increasing number of dual-worker couples approaching retirement together.
The second missing piece in existing research on joint retirement is an explicit recognition that this is a process that inherently involves two people whose expectations may differ. Currently, we do not know whether expectations of joint retirement are shared by both spouses or the extent to which shared expectations are related to outcomes. Are couples who share similar expectations more likely to realize those expectations? Alternatively, is joint retirement a relatively spontaneous outcome for which planning and couple-level agreement on expectations are largely irrelevant? If spouses do not share similar expectations, whose expectations are outcomes more likely to resemble? It is plausible that husbands' expectations are more likely to be realized because they have traditionally been the main breadwinner and their careers have typically been given priority within the family. At the same time, however, it is possible that wives' expectations may also be strong predictors of outcomes given that women who are at risk of experiencing joint retirement have also had substantial experience in the labor market (Henretta and O'Rand 1983).
Existing research on joint retirement also provides little evidence regarding factors that facilitate or hinder the realization of joint retirement expectations. From previous studies of individual outcomes, we know that unanticipated events such as health deterioration are related to differences between initial expectations and subsequent retirement behavior (Anderson, Burkhauser and Quinn 1986; Dwyer 2001). Are similar factors related to the likelihood of realizing joint retirement expectations? What factors influence the probability of realizing joint retirement expectations among couples who share similar expectations? Evidence that realization of retirement expectations is positively associated with subsequent subjective well-being (Herzog, House and Morgan 1991) highlights the importance of answering these questions.
In this paper, we address these gaps in the literature by examining joint retirement as a process involving two people with potentially distinct expectations. We address several research questions. At mid-life, what proportion of married workers expect joint retirement? What proportion of couples share similar expectations? What proportion of couples realize their initial expectations? Are couples with similar expectations more likely to realize their expectations? When spouses have different expectations, whose expectations do observed outcomes more closely resemble? Do expectations provide information about couples' future behavior, net of other relevant conditions such as pension benefits or health status? And, finally, what factors facilitate or hinder the realization of initial expectations?
To answer these questions, we examine data from the first seven waves of the Health and Retirement Study (1992–2004). We begin by describing dual-worker couples' joint retirement expectations at the time of the first survey in 1992. We then follow couples for up to 12 years to determine whether or not they retired jointly. Here, we define joint retirement as a couple-level outcome in which both spouses retire within period of 12 months. We then estimate logistic regression models to assess whether couples sharing similar expectations are more likely to retire jointly, and whose expectations are more likely to be realized in cases where spouses' expectations differ. In these analyses, we control for several documented correlates of joint retirement, including demographic characteristics, economic conditions, preferences for joint leisure, and discussion of retirement. Finally, we restrict our sample to couples who shared expectations of joint retirement to examine the factors associated with the likelihood of realizing those shared expectations.
Steady increases in women's labor force participation across the life course have contributed to the increase in joint retirement. The labor force participation rate of women in 2000 was 60%, nearly a two-fold increase from 34% in 1950. Labor force participation rates of women between the ages of 55 and 64 also doubled from 27% in 1950 to 52% in 2000 (Toossi 2002). Importantly, this increase in period labor force participation rates reflects an increase in the stability of women's labor force attachment across the life course. As labor force exits associated with marriage and childrearing have become less common and shorter in duration, women's labor supply profiles have come to resemble those of men (Bureau of Labor Statistics 2005). In 2005, half of married couples were dual-worker couples (Bureau of Labor Statistics 2007) and an increasing number of late mid-life couples face two retirements to coordinate. In many of these couples, wives are entitled to Social Security benefits, private pension benefits, and post-retirement health insurance based on their own work histories, not as dependents of their husbands (O'Rand and Farkas 2002). This increasing symmetry in husbands' and wives' careers and retirement incentives presumably complicates the process of synchronizing retirements (O'Rand, Henretta and Krecker 1992).
Research on the family context of retirement has demonstrated the tendency for couples to synchronize their retirement timing. For example, research on individual retirement transitions find that the probability of retirement is higher for people with a retired spouse compared to those with a working spouse, net of economic and health conditions (Blau 1998; Pienta 2003; Szinovacz and DeViney 2000; Vistness 1994). Spouse's retirement and pressure to retire from a spouse who has already retired are commonly reported reasons for retirement, especially among women (Szinovacz 1989; Szinovacz and DeViney, 2000). Couple-level studies also highlight the importance of joint retirement. Using various definitions of joint retirement such as similar patterns of work and non-work over two-year period (Henretta and O'Rand 1983), both spouses not working during the same calendar year (Hurd 1988), joint exit from the labor force in the same quarter of a given calendar year (Blau 1998), and two self-reported retirements within 12 months (O'Rand and Farkas 2002), previous studies have shown that between twenty and thirty percent of dual-worker couples retire jointly.
Complementarity of leisure has been cited as a key reason for the trend toward joint retirement. The argument is that, rather than retiring earlier and enjoying the income generated by their spouse's labor force participation, the older spouse (typically the husband) may prefer to postpone his/her own retirement in order to better enjoy retirement by spending leisure time with his/her spouse (Coile 2003; Gustman and Steinmeier 2002). Several studies conclude that preferences for shared leisure are more important than economic conditions, health status, and other objective correlates of joint retirement in explaining the propensity for spouses to retire jointly (An, Christensen and Gupta 2004; Gustman and Steinmeier 1985; Gustman and Steinmeier 2000; Gustman and Steinmeier 2002).
Retirement is increasingly viewed as a normative life event for which workers form expectations ahead of time (Ekerdt, Kosloski and DeViney 2000). A growing body of research on retirement expectations has examined the correlates of expected retirement dates and the subjective probability of working full-time beyond age 62 or 65. This work has found that expectations remain stable over extended periods of time and are associated with well-documented correlates of actual retirement behavior in expected ways (Benitez-Silva and Dwyer 2006; Chan and Stevens 2002; Honig 1996; Pienta and Hayward 2002). Studies examining the congruence between expectations and subsequent behavior have found that expected retirement timing is frequently consistent with actual timing observed in subsequent survey waves (Anderson, Burkhauser and Quinn 1986; Benitez-Silva and Dwyer 2006; Bernheim 1987). Bernheim's (1987:2) conclusion that people are “reasonably competent at forming relatively accurate expectations about the timing of retirement” using information available to them points to the importance of expectations as a source of reliable information for projecting future behavior.
However, unforeseen events such as health shocks and family caregiving obligations may cause behavior to deviate from initial expectations. Sudden health deterioration is a strong predictor of change in retirement expectations (McGarry 2004) and often prompts individuals to retire earlier than they had planned (Anderson, Burkhauser and Quinn 1986; Dwyer 2001). Family caregiving responsibilities, which are often unanticipated, may also affect actual retirement timing, especially for women (Zimmerman et al. 2000).
Married men and women (in particular) incorporate their spouses' retirement expectations and information about related factors such as health conditions and pension eligibility when forming their own expectations (Benitez-Silva and Dwyer 2006). Spouse's retirement or pressure from a spouse who has already retired are important motivations for retirement (Szinovacz 1989; Szinovacz and DeViney 2000) and people acknowledge spouse's influence on their own retirement decisions (Smith and Moen 1998). Inclusion of spousal and family characteristics explains variance in retirement expectations above and beyond that explained by individual characteristics (Pienta and Hayward 2002). Furthermore, it appears that there is a positive correlation in the degree to which husbands and wives plan for retirement (Moen et al. 2006).
In family studies, couples are often treated as a unit of analysis with united expectations or preferences, particularly in cases where there is just one outcome per couple such as fertility and joint retirement. However, research on fertility outcomes and family bargaining over the allocation of resources recognizes that husbands and wives may have different, sometimes conflicting, interests (Lundberg and Pollak 1996; Thomson 1989, 1997; Thomson, McDonald and Bumpass 1990). These studies raise important questions about the influence of disagreement on the outcome of interest and whose expectations are more likely to be realized. Although retirement is increasingly viewed as a coupled process involving two individuals (Lundberg 1999), no study of joint retirement has examined these questions to our knowledge. An understanding of the relationship between (dis)agreement and joint retirement outcomes may provide valuable insights into the joint retirement process while an understanding of how disagreements are resolved may provide insights into the gender dynamics of couples approaching retirement.
It is reasonable to expect that the retirement outcomes of couples with discordant expectations will differ from those of couples sharing similar expectations. Fertility research, for example, finds that disagreement results in a lower probability of achieving either spouse's individually desired number of children (Thomson, McDonald and Bumpass 1990). The fertility of couples in which one spouse wants a small number of children while the other wants a large number lies between that of couples who both want large numbers of children and that of couples with shared desire for a small number of children. Retirement obviously differs from fertility in that it does not require as much cooperation from the other spouse. In theory, people can realize their own joint retirement expectations without the cooperation of their spouse.
Discordant expectations also raise the question of gender symmetry and bargaining power: are husbands' or wives' expectations more likely to be realized? We consider the following two scenarios. The first is based on the traditional breadwinner-homemaker relationship in which the husband's career takes precedence and wives' work outside the home is of secondary importance to the household. Consistent with this scenario, several studies find that women are more likely to retire in response to pressure from their husbands (Szinovacz 1989; Szinovacz and DeViney 2000), while husbands' retirement is less likely to influenced by their wives (Benitez-Silva and Dwyer 2006; Moen et al. 2006). To the extent that couples are characterized by gender-asymmetric economic roles, we expect that patterns of spouses' retirement timing are more likely to be consistent with husbands' expectations. In these couples, wives' expectations are posited to have little or no predictive power beyond the expectations of their husbands. In contrast, our second scenario is based on a more egalitarian relationship in which the careers of husbands and wives are of similar importance. In this scenario, we expect greater variation in whether observed outcomes among couples with discordant expectations resemble husbands' or wives' expectations. Consistent with this scenario, research on fertility finds that outcomes reflect husbands' and wives' intentions or preferences equally (Thomson 1989; Thomson, McDonald and Bumpass 1990) while research on retirement finds that wives' characteristics are strongly associated with the retirement outcomes of both husbands and couples (Coile 2003; Gustman and Steinmeier 2002; Hurd and McGarry 1995; O'Rand, Henretta and Krecker 1992).
We use data from the first seven waves of the Health and Retirement Study (HRS) (1992–2004), a nationally representative longitudinal survey focusing on health, economic status, and the retirement process of men and women who were between the ages of 51 and 61 in 1992 (HRS, 1995). The HRS is well-suited to our analyses for several reasons. First, joint retirement expectations of both spouses were ascertained in the first wave of the survey. Second, extensive information about retirement status and related family, health, and financial characteristics has been collected at each wave. Third, the HRS collects data directly from both husbands and wives. Parallel information from both spouses is essential for analysis of discordant expectations regarding joint retirement and for examining the correlates of congruence between expectations and outcomes, given that retirement outcomes are influenced by the characteristics and expectations of each spouse (Coile 2003; Gustman and Steinmeier 2002; Hurd and McGarry 1995; O'Rand, Henretta and Krecker 1992). Finally, the HRS provides recent information on joint retirement whereas many previous studies have focused on the behavior of much earlier cohorts (Blau 1998; Henretta and O'Rand 1983; Hurd 1988). Evidence of cohort differences in couples' retirement planning (Moen et al. 2006) and the fact that the large baby boom cohorts are now approaching retirement highlights the importance of using the most recent data.
Our analytic sample consists of couples who met the following conditions at the first wave in 1992: both spouses were at least 50 years old, were currently working for pay, and considered themselves to be “not retired at all.” We set a lower age limit of 50 in the first wave because these spouses would not yet be eligible for early receipt of Social Security benefits by the seventh wave in 2004. There are 1,144 couples that meet these conditions. The question about joint retirement expectations is only asked of respondents who reported being “not retired at all” and provided a planned retirement date. It is not asked of those who reported that they “will never retire” or “haven't though about when to retire.” We therefore exclude 436 couples in which one or both spouses were not asked about joint retirement expectations. We also exclude 59 couples in which either one or both spouses answered `don't know,' refused to answer the question, or responded that their spouse was not working. These conditions reduce the sample to 649 couples. We follow this initial sample through the seventh wave in 2004 to identify which couples retired jointly. We exclude 85 couples in which neither spouse retired by the seventh wave, 19 couples whose marriage ended by divorce or widowhood before their joint retirement status could be identified, 43 couples for whom retirement dates were missing, 39 couples with missing data on any other variables used in the analyses, and 35 couples lost to sample attrition. We also exclude 8 couples in which one spouse retired within 11 months of the last interview date while the other spouse remained in the labor force. Because we define joint retirement as two retirements separated by no more than 12 months, we cannot identify the joint retirement status of these couples. We are left with an analytic sample of 420 couples.
Joint retirement expectations were ascertained only in the first survey in 1992. Respondents who reported not being retired and provided an expected retirement year were asked: “Do you expect your spouse to retire at about the same time that you do?” Based on responses to this question, we classify couples into four categories: (a) couples in which both spouses expect to retire jointly, (b) couples in which the wife expects joint retirement while the husband does not, (c) couples in which the husband expects joint retirement while the wife does not, and (d) couples in which both spouses do not expect joint retirement.
We determine retirement timing based on subjective retirement status and self-reported date of retirement. At each wave, the HRS asks respondents: “At this time do you consider yourself partly retired, completely retired, or not retired at all?” Respondents who report being partly or completely retired are then asked: “In what month and year did you retire?” We define joint retirement as cases in which these self-reported retirement dates of husband and wife differ by no more than 12 months.
Drawing on previous research on retirement timing in general and joint retirement in particular, we model joint retirement as a function of demographic and family characteristics, economic conditions, and indicators of the complementarity of leisure and discussion of retirement. Detailed definitions of these variables are presented, along with descriptive statistics, in Table 1. Demographic and family characteristics include husband's age at couple's first retirement, age difference between spouses, subjective health status of both husband and wife, educational attainment of both husband and wife, length of current marriage, and couples' family caregiving responsibilities. These variables are measured in 1992 except for health status and family caregiving responsibility which are time-varying. Age is one of the most important predictors of joint retirement. Couples who retired jointly are more likely to do so at later ages, because in most couples husbands are older than their wives and husbands tend to wait for their wives to reach the age of eligibility for retirement benefits (Gustman and Steinmeier 2002; Szinovacz 1989). In this study, we measure age at retirement as the husband's age when either spouse retired for the first time. Large age differences between spouses should increase the cost of joint retirement either by requiring the older spouse (the husband in most cases) to stay in the labor force longer and/or by pressuring the younger spouse (the wife in most cases) to retire before reaching full eligibility for retirement benefits (Szinovacz 1989). We measure the age difference between spouses by subtracting wife's age from husband's age. Previous studies have shown that people with higher educational attainment expect to retire later (Hall and Johnson 1980) and have lower probability of retirement, net of other occupational characteristics (Hayward 1986; Hayward et al. 1989). We use a four-category measure of educational attainments: less than high school, high school, some college, and college and above. Based on evidence that marital duration is negatively related to the probability of retirement (Pienta 2003), we include the number of years in the current marriage. Couples married to each other for a longer period of time are more likely to have a more substantial shared labor force participation history (Henretta, O'Rand and Chan 1993) and may thus be more likely to share similar expectations regarding retirement options.
Previous studies posit that unforeseen changes in health and family care responsibility contribute to deviations between retirement expectations and realization. Poor health is associated with expectations for earlier retirement (Hall and Johnson 1980; McGarry 2004) and unanticipated health shocks trigger labor force exit earlier than planned (Anderson, Burkhauser and Quinn 1986; Coile 2004; Dwyer 2001). In this study, we measure whether spouses have fair or poor health at two time points, at the first wave in 1992 and at the wave after either spouse retires for the first time. Family caregiving is associated with earlier retirement, especially for women (Dentinger and Clarkberg 2002), and unanticipated family care responsibilities may result in retirement at an earlier age than expected (Zimmerman et al. 2000). In this study, we include a couple-level measure distinguishing those couples in which at least one spouse provided more than 100 hours of care for grandchild(ren) or parent(s) over past 12 months. As in the health measures, we measure family caregiving responsibility at two time points, at the first wave in 1992 and at the wave subsequent to the couple's first retirement.
We include five measures of economic conditions: total net wealth of the household, concerns about retirement income sufficiency, couples' relative earnings, and husbands' and wives' participation in private pension plans. In general, we expect that the financial impact of synchronized retirement will be lower for couples in better economic circumstances prior to retirement and that these couples will therefore be more likely to retire jointly (Adams et al. 2002; Hall and Johnson 1980; O'Rand and Farkas 2002). Household net wealth is measured as the sum of all wealth components less all debt. We assume that joint retirement implies a larger loss of regular income than non-joint retirement and thus expect that couples with greater household wealth will be better able to afford joint retirement (Blau 1998; O'Rand, Henretta and Krecker 1992). If wealth does indeed facilitate joint retirement, we also expect that concerns about financial well-being after retirement should be negatively associated with joint retirement. Concern about retirement income sufficiency is measured by the degree to which people worry about having enough income to get by. The relative earnings of couples are measured by the ratio of the difference between husband's earnings and wife's earnings and their combined earnings (Sorensen and McLanahan 1987). This variable measures the wife's relative contribution to the couple's earnings. Assuming that wives' relative earnings are positively associated with the economic cost of joint retirement, we expect that this measure will be negatively related to the probability of joint retirement. Participation in private pension plans is positively related to retirement wealth and should thus lower the cost of joint retirement. At the same time, however, pension plans are also expected to discourage joint retirement because eligibility for benefits is typically linked to age and these institutional incentives for husbands and wives to retire at a particular point in time may not coincide. Previous studies suggest that if wives are eligible for their own pension benefits, they are more likely to remain in the labor force rather than to retire together with their husbands (Blau 1998; O'Rand, Henretta and Krecker 1992). All of these indicators of economic status were measured at the first wave in 1992.
As mentioned above, earlier studies have stressed the importance of the complementarity of leisure as a primary motivation for joint retirement. In this study, we measure the preference for shared leisure by respondents' reported value of spending time with their spouse after retirement. Couple-level discussion of retirement can be viewed as facilitator of joint retirement. Frequent discussion of retirement between spouses during the pre-retirement period is expected to contribute to shared expectations regarding the relative timing of retirements (Smith and Moen 1998) and/thus to facilitate couples' efforts to realize their expectations. In this study, we measure the couple-level discussion by how often couples discuss retirement. Both complementarity of leisure and discussion of retirement are measured at the first wave in 1992.
Our analysis consists of two parts. In the first part, we estimate two logistic regression models for joint retirement. The first model estimates the baseline associations between the log-odds of joint retirement and the variables just described. In the second model, we include initial expectations. If the coefficients of initial expectations are significant and their inclusion improves model fit, we can conclude that expectations provide important information about future retirement behavior above and beyond the established correlates of joint retirement. We also compare the log-odds of joint retirement for two types of couples with discordant expectations (“the husband expects joint retirement but the wife does not” and “the wife expects joint retirement but the husband does not”). If the coefficients for these two categories do not differ, husbands' and wives' expectations are equally relevant predictors of subsequent retirement behavior. Similarly, by comparing the estimated coefficients for these disagreeing couples with that for couples who share joint retirement expectations, we can evaluate the extent to which disagreement is associated with the likelihood of joint retirement. In the second part of our analysis, we restrict our analytic sample to couples in which both spouses initially expected joint retirement. This restriction allows us to identify the characteristics of couples most likely to realize expectations of joint retirement.
Expectations of joint retirement are widespread, with 63% men and 60% of women in our sample reporting that they expect to retire at the same time as their spouse. At the couple-level, almost half of dual-worker couples (48%) expect to retire at about the same time, while one quarter (25%) anticipate that they will retire at different times. The proportion of couples with discordant expectations is small: 15% of our sample is comprised of couples in which husbands, but not wives, expect joint retirement and 12% is comprised of couples in which wives, but not husbands, expect joint retirement.
The observed proportion of couples who retire jointly is relatively small considering that nearly half of couples expected to retire together. Slightly less than a third of couples (29%) retired within one year of each other. Joint retirement is strongly associated with initial expectations. The proportion retiring jointly was 36% for couples with shared expectations of joint retirement, 31% for couples in which the husband, but not the wife, expected joint retirement, and 26% for couples in which the wife, but not the husband, expected joint retirement. If neither spouse expected joint retirement, the proportion retiring within a year of each other is only 15%.
In Table 2, we present exponentiated values of the estimated coefficients from the two logistic regression models for joint retirement. These odds ratios describe change in the odds of joint retirement associated with a one-unit change in the correlates of interest. In Model 1, most of the correlates are related to joint retirement behavior in expected ways. Joint retirement is more common when husbands retire at older ages (odds ratio = 1.22) and had fair or poor health in 1992 (odds ratio = 2.53). Joint retirement is less common if wives have a college degree or more (odds ratio = 0.45) and if couples were more concerned about income sufficiency after retirement (odds ratio = 0.86). Discussion of retirement is positively associated with joint retirement, suggesting that joint retirement is facilitated by careful preparation (odds ratio = 1.26).
Results of Model 2 indicate that initial expectations are significantly associated with subsequent behavior, net of well-established correlates of joint retirement. The odds of joint retirement are more than four times higher for couples in which both spouses expected joint retirement relative to otherwise similar couples in which neither spouse expected to retire jointly. Couples with discordant expectations also have significantly higher odds of joint retirement - 2.9 times higher for couples in which only husbands expected joint retirement and 2.4 times higher for couples in which only wives expected joint retirement. Controlling for expectations, husbands' education and wives' poor health before retirement are no longer significant, suggesting that these factors influence joint retirement behavior through initial expectations, not directly.
In terms of the resolution of discordant expectations, husbands' and wives' expectations appear to be equally powerful predictors of joint retirement. There is no significant difference in the log-odds of joint retirement for couples in which only the husband or only the wife expected to retire at the same time as their spouse. Furthermore, there is no statistical difference between the log-odds of joint retirement for couples in which both spouses expected joint retirement and couples in which only one spouse expected joint retirement. These results suggest that agreement about the relative timing of husbands' and wives' retirement is not necessary. As long as one spouse expects to retire jointly, the odds of that outcome being realized are significantly increased.
In the second part of our analysis, we restrict the sample to couples in which both spouses expected joint retirement in order to examine which factors are related to the realization of joint retirement expectations. Results of the logistic regression analysis, presented in Table 3, suggest that husbands' age at couples' first retirement, relative earnings of spouses, worries about retirement income sufficiency, and discussion of retirement are significantly related to the realization of initial expectations. A one year increase in husbands' age at couples' retirement is associated with a 24% higher odds of joint retirement. This relationship is consistent with results from previous studies on joint retirement (Gustman and Steinmeier 2002; Szinovacz 1989). Two measures of economic circumstances are significantly related to the realization of joint retirement expectations. Wives' contribution to couples' earnings is inversely related to the likelihood of realizing expectations to retire jointly (odds ratio = 0.38). Presumably, this reflects the higher cost of joint retirement for couples in which husbands' and wives' economic contributions are similar. Concerns about retirement income sufficiency are negatively associated with the realization of joint retirement expectations (odds ratio = 0.82). This relationship is consistent with previous findings that joint retirement is a “pattern of choice” for socioeconomically advantaged couples rather than a “pattern of constraint” (O'Rand, Henretta and Krecker 1992:97). Finally, we see that couples who discuss retirement more often are more likely to realize their initial expectations (odds ratio = 1.39).
The main purpose of this study was to examine relationships between joint retirement expectations and subsequent behavior among dual-worker couples in late-mid life. Using the first seven waves of the HRS data, we found that half of couples share expectations of synchronized retirement. Compared to these initial expectations, however, the proportion of couples who actually retire jointly is relatively low. Less than one-third of couples retire jointly, and less than two in five couples with shared expectations of joint retirement realized those expectations.
We found that initial expectations are strong predictors of subsequent behavior net of other well-established correlates of joint retirement. Couples in which both spouses expected joint retirement are over four times more likely to retire jointly compared to couples in which neither spouse expected to do so. These results suggest that expectations provide valuable information for projecting future behavior. In terms of spousal disagreement, we found that wives' and husbands' expectations are equally strong predictors of joint retirement. There was no statistical difference in the odds of joint retirement for couples in which wives, but not husbands, expected joint retirement and couples in which husbands, but not wives, expected joint retirement. These results are consistent with previous studies of retirement timing showing that both wives' and husbands' characteristics are associated with couples' retirement timing (Coile 2003; Gustman and Steinmeire 2002; Hurd and McGarry 1995; O'Rand, Henretta and Krecker 1992). Our results thus provide no support for a gender-asymmetric scenario in which husbands' expectations are prioritized in couples' decisions about retirement timing.
Another goal of our study was to examine the influence of couples' disagreement on joint retirement outcomes. We found no significant differences in the odds of joint retirement for couples in which both spouses expected joint retirement and couples with discordant expectations. Unlike fertility, where spousal disagreement is associated with a lower probability of realizing either spouse's intended number of children (Thomson, McDonald and Bumpass 1990), disagreement regarding joint retirement is not significantly related to lower odds of joint retirement. Not surprisingly, joint retirement requires less couple-level cooperation than fertility.
Among couples in which both spouses initially expected joint retirement, the likelihood of realizing those expectations was higher among couples who retired at later ages, couples in which wives' relative earnings were lower, couples who were less worried about retirement income sufficiency, and couples who discussed retirement more often. Later age at retirement is consistent with evidence that joint retirement is often achieved by husbands postponing their own retirement until their wives are ready (eligible) to retire (Szinovacz 1989). The inverse relationship between wives' relative earnings and the realization of joint retirement expectations suggests that the economic impact of joint retirement is lower for couples in which the wife's earnings are supplementary and thus implies that realization of joint retirement expectations may be more difficult among subsequent cohorts characterized by more gender-symmetric economic relationships. Finally, the positive relationship between discussion of retirement and realization of joint retirement expectations suggests the importance of making realistic plans based on mutual understanding.
This study provides several useful insights for future research on joint retirement. First, it is clear that a substantial proportion of couples fail to realize their expectations for joint retirement. Given that failure to realize expectations or aspirations is negatively related to psychological well-being (Carr 1997; Gallo et al. 2006; Herzog, House and Morgan 1991), the relatively high likelihood of failing to realize joint retirement expectations may lead to lower levels of retirement satisfaction and mental well-being. Consistent with this speculation, recent research has found that joint retirement is related to higher retirement satisfaction and lower depression (Szinovacz and Davey 2004; Szinovacz and Davey 2005). Together, these results suggest that careful examination of linkages between joint retirement expectations, their realization, and subsequent retirement satisfaction may be a valuable source of insights in subsequent research on subjective well-being at older ages. Second, although retirement has become a normative life event that most people expect to experience at the end of their careers (Ekerdt, Kosloski and DeViney 2000), retirement timing and pathways have become increasingly heterogeneous (Han and Moen 1999). Work after retirement, partial retirement, and multiple labor force transitions at older ages have increased simultaneously with continued growth in early retirement (Herz 1995; Kim and Moen 2002; Mutchler et al. 1997). This growing variety of retirement options presumably makes it increasingly possible for couples to coordinate their retirements. Husbands may wait for their wives to retire while working part-time or in a bridge job. Alternatively, wives can choose earlier retirement packages. Third, results from the second part of our analysis provide insights for measures facilitating the realization of joint retirement expectations. Evidence that husbands' later age at retirement and frequent discussion of retirement facilitate the realization of expectations to retire together suggests public policy targets. For example, policies designed to facilitate extended labor force participation and efforts to encourage early planning and discussion of retirement among couples may help couples to achieve their preferred retirement scenario.
Our study also suggests several avenues for developing future research on joint retirement. First, increasing variability in the retirement process suggests the value of considering alternative definitions of retirement. Although there is substantial overlap between self-defined retirement status and working status (80% of respondents who report themselves as being retired were not working), the distribution of joint retirement experience may differ depending on the definition of retirement. At the same time, it is also possible that the definition of “joint” retirement may differ from person to person. Although we define joint retirement as spouses retiring within 12 months of each other, respondents may interpret the question, “spouse retiring at about the same time,” as retiring in the same month or in the same calendar year, or perhaps retiring within a few years of each other. Even spouses who share similar joint retirement expectations may have different definitions of joint retirement. Tabulations of spouses' expected years of retirement (not shown) indicate that, among those couples who share expectations of joint retirement, slightly less than one-third report expecting to retire in the same calendar year while the majority (57%) reported adjacent calendar years. Second, our analytical sample immediately precedes the large baby boom cohorts now entering prime retirement ages. Moen and colleagues (2006) conclude that there are important cohort differences in the interdependence of retirement planning between spouses. The baby boom cohorts also differ from the main HRS cohort in our study in terms of women's long-term career commitment and the prevalence of dual-worker couples (Dailey 1998), higher levels of wealth (Lusardi and Michell 2006; Rander 1998), and more diversity in demographic and economic characteristics within the cohorts (Dailey 1998). It is important that our results are reevaluated for these younger cohorts once data are available. Finally, the high likelihood of failing to realizing expectations for joint retirement highlights the importance of examining the psychological implications of congruence (or lack thereof) between joint retirement expectations and subsequent outcomes.
This research was supported by grants to Center for Demography and Ecology (R24 HD047873) and the Center for Demography of Health and Aging (P30 AG17266) at the University of Wisconsin-Madison.