This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation

Honeymoons and Joint Lunches: Effects of Retirement and Spouse's Employment on Depressive Symptoms

¹ Glennan Center for Geriatrics and Gerontology, Eastern Virginia Medical School, Norfolk.
² Polisher Research Institute, North Wales, Pennsylvania.

Address correspondence to Maximiliane E. Szinovacz, Glennan Center for Geriatrics and Gerontology, Eastern Virginia Medical School, Hofheimer Hall, Suite 201, 825 Fairfax Avenue, Norfolk, VA 23507-1912. E-mail: maxres{at}visi.net

	Abstract

TOP Abstract Methods Results Discussion References

 
With hypotheses derived from a life course perspective in conjunction with life event stress and role theories, we examine whether a spouse's employment and length of retirement affect a person's postretirement depressive symptoms and whether such effects differ by gender. Analyses use pooled data from Waves 1–4 of the Health and Retirement Survey, using a subsample of married individuals who either remained continuously employed over time or completely retired since the Wave 1 interviews (N = 2,695). Recently retired men seem to be negatively affected by their spouses' continuous employment when compared with men whose wives were continuously not employed. In contrast, spouses' joint retirement has a beneficial influence on both recently retired and longer-retired men. However, for recently retired men, the positive effect of wives' retirement seems to be contingent on spouses' enjoyment of joint activities. Among women, effects of spouses' employment occur only among very recently retired wives (0–6 months). These wives report more depressive symptoms if their spouses were already nonemployed prior to wives' retirement. These results demonstrate the complexity of retirement adaptation processes and suggest that marital context plays an important role in retirement well-being.

Increases in the labor force participation of middle-aged wives imply that, for many spouses, retirement has become a "couple event" (Szinovacz & Ekerdt, 1995). Dual-earner couples organize and experience the retirement transition not only in terms of their own but also in regard to their spouse's retirement. Although several earlier studies have addressed the question of how selected couple retirement patterns influence postretirement marital quality or well-being, many of these studies were limited because of their reliance on small or nonrepresentative samples such as data from retirees of specific firms (Kim & Moen, 2002; Moen, Kim, & Hofmeister, 2001), use of retrospective or cross-sectional data (Lee & Shehan, 1989; Skirboll & Silverman, 1992), or an exclusive focus on men's retirement (Myers & Booth, 1996). We revisit this topic by using longitudinal data from the Health and Retirement Survey (HRS; Juster & Suzman, 1995) and a theoretical framework that integrates assumptions from the life course perspective with assumptions pertaining to phases of retirement, gender role ideology, and marital context.

Theoretical Framework
Earlier studies on postretirement well-being focused either on retirees' activities and roles or on selected resources such as health, finances, or marital status (for reviews see Atchley, 1976; Kosloski, Ginsburg, & Backman, 1984; Ross & Drentea, 1998). In contrast, the theoretical framework informing our approach relies foremost on the life course perspective. According to life course theory, an understanding of life course transitions such as retirement requires consideration of context, linked life spheres, and timing, trajectories, and pathways (Settersten, 2003). Life transitions are contextually embedded; that is, adaptation to life transitions not only depends on the occurrence of the transition per se but also on the specific contexts under which the transition takes place. In the case of retirement, the most important contexts include former work roles and work history, health, finances, and marital status (Bossé, Aldwin, Levenson, & Workman-Daniels, 1991; Calasanti, 1996a; Kim & Moen, 2002; Richardson & Kilty, 1991; Szinovacz, 1989). Furthermore, the experience of life transitions is contingent on developments in other life spheres (linked lives). We focus on the marital relationship and especially on spouses' employment and retirement. The concepts of timing, trajectories, and pathways refer to development within and across life spheres. Timing of life events is important in regard to cultural norms and personal expectations; that is, off-time events tend to undermine well-being (Hagestad, 1990). Trajectories refer to development within a life sphere such as the work or retirement career. Specific work career trajectories (e.g., those ensuring adequate pension coverage and thus sufficient postretirement income; see O'Rand & Henretta, 1999) may enhance retirement well-being, whereas others may hinder it (e.g., unstable work patterns; see Marshall, Clarke, & Ballantyne, 2001). Pathways refer to the combination of trajectories from different life spheres and the timing of life transitions in relation to each other. For example, spouses may enter retirement jointly or separately. In addition to its emphasis on contexts, linked lives, and transitions, the life course perspective pays tribute to the gendered nature of life experiences. Consistent evidence suggests that spouses experience the marital relationship differently—Bernard (1973) speaks of a "his and her" marriage. Whereas evidence concerning gender differences in postretirement well-being is inconsistent (Carp, 1997; Slevin & Wingrove, 1995), studies indicate that both retirement contexts and retirement meanings vary by gender (Calasanti, 1996b; O'Rand & Henretta, 1999; Szinovacz, 1991).

Although the life course perspective provides a general framework for assessing effects of life transitions on well-being, it does not provide specific hypotheses about the influences of contexts, trajectories, or pathways. We thus rely on selected concepts and assumptions from other theories for hypothesis development. These include assumptions about phases of retirement adaptation, gender role ideology, and marital quality.

Retirement phases
Adaptation to life transitions may vary over time. Atchley (1976) first suggested five phases of retirement adaptation—honeymoon, disenchantment, reorientation, stability, and termination. Although the length of these phases may vary considerably, our analyses focus on retirement over one wave of the HRS, typically a period of 2 years, and thus are likely to reflect only the honeymoon and disenchantment phases (for a discussion of the HRS, see the Methods section). Some research provides support for a honeymoon phase, that is, an initial euphoric reaction to retirement, followed by some decline in well-being (Ekerdt, Bossé, & Levkoff, 1985; Richardson & Kilty, 1995), but other studies failed to sustain this pattern, especially for women (Kim & Moen, 2002; Richardson & Kilty, 1995). If retirement phases do indeed exist, short-term retirees should report fewer depressive symptoms than longer-term retirees. However, this effect may be contingent on marital context, especially spouses' employment and marital quality.

Gender role ideology and retirement patterns of couples
Previous research has shown that spouses strive toward joint retirement (Allmendinger, Brückner, & Brückner, 1991; Blau, 1998; Henkens, 1999; Szinovacz & DeViney, 2000; Zimmerman, Mitchell, Wister, & Gutman, 2000) and that continued employment of the wife after the husband's retirement is associated with reduced postretirement well-being (Moen et al., 2001; Myers & Booth, 1996; Szinovacz, 1996; Szinovacz & Schaffer, 2000). It has been argued that this timing pattern undermines the husband's status as the main provider and reduces his power in the marital relationship (Davey & Szinovacz, 2004; Kulik, 1996; Kulik & Bareli, 1997; Szinovacz & Harpster, 1993), both conditions that apparently hinder retirement adjustment. The disenchantment associated with the retired husband–employed wife pattern seems more pronounced among husbands. In light of the negative effect of the wife's continued employment on the husband's postretirement well-being, husbands in this situation are unlikely to experience a retirement honeymoon. We thus expect a retirement honeymoon effect, as reflected by relatively lower levels of depressive symptoms, for husbands whose wives were already not employed at baseline (Hypothesis 1a) or retired during the same time span as their husbands (Hypothesis 1b).

Given that many couples aim for joint retirement, we expect wives to benefit from this retirement pattern as do husbands. Little is known about retirement adjustment of spouses after the wives' retirement, especially if the husbands continue to work. In contrast to the nonnormative character of husbands' retirement prior to their wives, this retirement pattern is consistent with the traditional provider role. Some research suggests that the retirement of wives prior to that of their husbands may profit husbands but not the wives themselves (Szinovacz & Schaffer, 2000), but other evidence indicates that husbands' and wives' adjustment to retirement is linked (Haug, Belgrave, & Jones, 1992). The positive impact of the retirement of wives prior to that of their husbands on postretirement well-being may reflect a resolution of disagreements surrounding wives' employment or a reduction in stress associated with employed wives' overload problems. We thus predict a retirement honeymoon, as reflected by relatively lower levels of depressive symptoms, for recently retired wives who retired jointly with their spouses (Hypothesis 2a) or whose husbands continued to work (Hypothesis 2b).

Life event stress and marital quality
The disruption in lifestyle and daily routines caused by life transitions has been linked to stress (Burke, 1991; Burke, 1996; Diehl, 1999; Kessler, Price, & Wortman, 1985; Pearlin, Lieberman, Menaghan, & Mullan, 1981). However, the extent of stress, and indeed whether stress occurs at all, is contingent on selected characteristics of the transition (Burke, 1991; Thoits, 1991; Wheaton, 1990; Wheaton, 1996). Most important for our analyses is the insight that the stress associated with life transitions depends on the salience of life spheres. Exits from salient spheres are more stressful, whereas availability of alternative salient life spheres reduces stress. Indeed, studies indicate that being married and having a high-quality marriage contribute to postretirement well-being (Atchley, 1992; Calasanti, 1996a; Kelly & Westcott, 1991; Niederfranke, 1989; Reitzes, Mutran, & Fernandez, 1996), whereas marital problems enhance perceptions of retirement-related hassles (Bossé et al., 1991). Implementation of a couple-oriented postretirement lifestyle most likely requires that spouses enjoy joint endeavors. Such endeavors are bound to intensify the salience of the marital relationship and have been found to be particularly important for postretirement marriages (Dorfman, Heckert, Hill, & Kohout, 1988; Vinick & Ekerdt, 1991). We thus expect the aforementioned hypothesized relationships to be contingent on couples' enjoyment of joint activities. Specifically, the predicted honeymoon effect after joint retirement or a wife's retirement prior to her husband's is expected to be stronger among couples enjoying joint activities than among those who do not enjoy joint activities (Hypothesis 3).

	METHODS

TOP Abstract Methods Results Discussion References

 
Sample
In our analyses we use data from the HRS, which was launched in 1992. The HRS is a longitudinal survey of households (the interviewed are the primary respondent and his or her spouse) that is conducted biannually. The data are in the public domain. Our analyses rely on Waves 1–4 (1992, 1994, 1996, and 1998). The primary original sample for the HRS consists of main respondents aged 51–61 years at Wave 1 and their spouses, regardless of the spouses' age (N = 12,652 respondents; 7,702 households). Selection of households was based on a multistage area probability design with oversamples for minorities and persons residing in Florida. The response rate was more than 80% (for further details see Heeringa & Rodgers, 1998, and Juster & Suzman, 1995).

The selection of the subsample for our analyses is guided by the assumption that divergent preretirement contexts influence postretirement well-being. From this perspective it is more important to restrict variability in the baseline population than to attain full generalizability of results. For the subsample used in our analyses, we first identified all HRS individuals aged 40 or older (the main respondents or their spouses) from couples with at least one age-eligible partner (i.e., age 51–61) who were employed 10 hr or more at baseline and did not define themselves as retirees. From this subsample we then selected persons who were either continuously employed throughout all four waves or who self-defined as "completely retired" at Waves 2, 3, or 4 and had left the labor force (the HRS retirement measures are discussed in the paragraphs that follow). The exclusion of Time 1 nonemployed or retired persons was necessary because we would lack baseline data for this population. We also excluded individuals working fewer than 10 hr at Time 1. This population may be underemployed or already in postretirement bridge jobs. Because our analyses focus on spouses' employment status and retirement, we further restricted the sample to persons who were married at baseline and reported no change in marital status or partners at Time 2, yielding a final subsample of 2,904 individuals. We lost an additional 122 persons whose partners did not participate in the surveys, and we lost some respondents (varying by analysis but fewer than 100) because there were missing values we could not impute. The final sample consists of 2,695 respondents, of whom 1,367 were men and 1,328 were women. After identifying these subsamples, we added information from spouses to each individual's file through the matching of records by household and partner identification. It should be noted that current retirees in the HRS tend to have retired early (i.e., before 62–65 years of age). Thus, our data may not be generalizable to those retiring in their mid to late 60s.

The analyses pool persons who retired between Waves 1 and 2, Waves 2 and 3, and Waves 3 and 4. Such pooling is necessary to achieve a sufficient number of cases for our complex analyses. It also creates a sample basis that addresses similar retirement transitions for all respondents, namely the transition from work to full-time retirement within the past 2 years. We recognize that many individuals do not follow this retirement path; they retire gradually. In line with our aim to explore clearly identifiable transition patterns (in the sense of Weberian ideal types), we focus on the full retirement pattern, which is also the most common in our sample. In each case, the earlier wave data serve as baseline and the later wave data as basis for the construction of change and outcome variables. For continuously employed workers, we randomly assigned respondents to waves to achieve relatively equal representation of waves in the analyses. We also controlled for wave to further exclude potential cohort or period effects. In the remainder of the article, we refer to the baseline data as Time 1 and the outcome wave as Time 2.

Losses of data caused by attrition or missing responses can bias results. We dealt with both issues. We adjusted for a potential attrition bias through weights that account for sample selection and nonparticipation biases as well as attrition. The HRS provides multistage sampling weights that correct for sample selection and nonresponse biases but do not account for attrition. To adjust for attrition, we estimated a logistic regression equation that predicted being in Wave 3. This used gender, race, marital status, various health measures (drinking, smoking, body mass index, self-rated health, depression, and limitations in activities of daily living, or ADLs), occupation, income, assets, pension, job satisfaction, point of origin (foreign), social contacts with neighbors, proxy interview, and participation in the Wave 2 interview as independent variables. The equation used the multistage sampling weights in the estimation of standard errors. From this equation we calculated the probability of being in Waves 2, 3, or 4 for each respondent. We then multiplied the inverse of this probability by the sampling weight to arrive at a final weight that includes the probability of being in the sample and being in all waves. The attrition weight is normalized so that weighted ns in the overall sample correspond to observed ns.

Imputation of missing values is recommended especially for longitudinal studies (Little & Rubin, 1989) to allow generalizations from the original sample. Except for missing values on the employment or retirement variables used to select our subsample, we used multiple imputations to correct for selective nonresponse. These imputations provide valid standard errors for all model parameter estimates (Davey, 2001; Davey, Shanahan, & Schafer, 2001; Schafer, 1997). Repeated draws were made using a standard noninformative prior and imputing every 2,000 iterations to construct five complete data sets. We analyzed each imputed data set separately, and we then pooled results to arrive at final estimates of parameters and their standard errors. Note that some variables were already imputed in the HRS or the data set distributed by RAND (e.g., financial information). Where available, we used the HRS or RAND imputations.

Measures
The main dependent variable is depressive symptoms. We measured depressive symptoms with selected items from the Center for Epidemiological Studies Depression (CES-D) scale (see Radloff, 1977). The items, wording, and scaling of measures from the CES-D scale differed between the first wave of the HRS and subsequent waves, and this is discussed extensively elsewhere (Steffeck, 2000). A random subsample of 808 individuals received both measures (the original Wave 1 measure and the new Wave 2 measure) at Wave 2. For Wave 1 ( =.80) and for the Wave 2 subsample receiving both measures ( =.81), the original CES-D categories were 0 = rarely or none of the time, 1 = some of the time, 2 = most of the time, and 3 = all or almost all of the time. For subsequent waves, responses were coded as yes (1) or no (0) in response to the item stem "Much of the time during the past week you felt ...."

To achieve comparable measures between the first and subsequent waves, we used the following procedure. Beginning with the eight comparable domains (felt depressed, everything was an effort, sleep was restless, felt happy, felt lonely, enjoyed life, felt sad, couldn't get going), we reverse coded positively worded items, and we summed the eight items according to their original coding. From these scales, we next estimated a regression equation to predict Wave 2 scale scores ( =.85) from Wave 2 scores on the module version, and from self-rated health, relative health (compared with a year ago), emotional health, limitations in activities of daily living, race (Black; other minority; White as reference category), age, age squared, and gender. The resulting regression equation predicted 71% of the variance in Wave 2 scale scores. We then used this equation to estimate Wave 1 rescaled scores. In order to allow for uncertainty in these estimates, we added a normally distributed random variate with a mean of 0 and variance equal to the mean square error from the Time 2 equation to all estimates. To verify that estimated rescaled scores behaved in a similar fashion to the newly scaled Wave 2 scores, we compared correlation coefficients between Wave 1 (original scale) scores and estimated Wave 1 rescaled scores with the correlation coefficients between Wave 2 module scores and Wave 2 scores. The former correlation coefficient was.82 and the latter was.83. Clearly, then, the behavior of our rescaled version of the Wave 1 CES-D scale was highly consistent with what was observed with the Wave 2 data.

The independent variables for the analyses are retirement, length of retirement, spouse's employment or retirement status at Times 1 and 2, and spouse's enjoyment of joint activities. Definitions of retirement can be ambivalent and often vary across studies. In the HRS, several indicators of retirement are available (Gustman & Steinmeier, 1995). We use a combination of labor force status (based on answers to this question: Are you doing any work for pay at the present time? Yes or no) and the respondent's self-definition as retiree (derived from answers to this question: Do you consider yourself partly retired, completely retired, or not retired at all?). Answer categories for this question reflect the three retirement statuses implied in the question (partly, completely, or not retired). For our analyses, retirees are those who had worked 10 hr or more for pay in the previous wave, stopped working between waves, and considered themselves as completely retired at Time 2. Persons defined as continuous employees worked throughout all four waves considered here and self-defined as "not at all retired."

We defined length of retirement as the time since the respondent stopped working. We dichotomized this variable into those who were retired up to 1 year at Time 2 and those retired more than 1 year. Because the length of retirement phases is not clear, we also subdivided the first group into those retired up to 6 months and those retired 7–12 months. Subdivision of those retired over 1 year is not possible because of the smaller number of cases in this subgroup. We created dummy variables for retirement status and length of retirement. Continuously employed persons serve as reference in the analyses.

For spouses we rely exclusively on employment status (any pay for work). We labeled spouses who did not work during both waves as "not employed," those who worked at Time 1 but had stopped working by Time 2 as "retirees," and those who worked at Time 2 as "employed." The distribution of couples by respondent's own and spouse's employment status is shown in Table 1. The sample selection of persons aged 51 to 61 at Wave 1 leads to a relatively young sample so that most individuals remain continuously employed. Men are somewhat more likely than women to have nonemployed spouses, whereas the proportion of respondents whose spouses retired is slightly higher for women. In line with earlier research, there is a pronounced association between respondent's and spouse's employment status: ²(4) = 62.38 for men and 57.19 for women, p <.01.

Controls
Postretirement well-being has been linked to numerous predictors. We attempted to include variables that have been shown most consistently to influence postretirement well-being in the analyses, provided data were available. The first set of controls consisted of demographic background variables, namely, age and race (all analyses were performed separately by gender). Race was entered in the form of three dummy variables, that is, Black, Hispanic, and other race (Whites serve as reference).

Our hypotheses refer to change in depressive symptoms after retirement. It is, therefore, essential that we control for preretirement psychological distress. We include depressive symptoms at Time 1, as well as whether respondents ever had psychiatric problems measured at baseline.

Foremost among other predictors of postretirement well-being are health and economic situation (Bossé et al., 1991; Calasanti, 1996a; Hardy & Quadagno, 1995). For respondents' health we use four indicators, namely self-rated health and limitations in ADLs, both measured at baseline, as well as changes in these indicators between waves. We measured self-rated health with a single item: Would you say your health is...? Answer categories ranged from poor (1) to excellent (5). We assessed limitations in ADLs across a variety of domains, 16 of which were consistent across waves and were performed by sufficient numbers of participants to be useful. These domains were as follows: walk several blocks; walk one block; walk across a room; sit for about 2 hr; get up from a chair after sitting for long periods; get in and out of bed without help; climb several flights of stairs without resting; climb one flight of stairs without resting; lift or carry weights over 10 lb; stoop, kneel, or crouch; pick up a dime from a table; bathe or shower without help; reach or extend your arms above shoulder level; push or pull large objects such as a living room chair; eat without help; and dress without help) However, there were some differences in item scaling across waves. To ensure comparability, we recoded items to 1 if the individual reported having any difficulty with the tasks, 0 if he or she did not, and missing if the individual reported that he or she did not perform a task. We then used the mean number of ADLs (multiplied by 16 to retain the original scaling) in subsequent analyses. The final score could thus range from 0 to 16. The coefficient alpha for Waves 1 and 3 was =.86 and for Wave 2 was =.87. We assessed changes in health with simple change scores, that is, the difference between Time1 and Time 2 scores. In the case of self-rated health, positive change scores indicate an increase in self-rated health. For limitations in ADLs, positive change scores reflect increases in limitations.

We used two measures to capture financial status—couples' income and net worth. The former variable includes income of both spouses from all income sources. Net worth reflects couples' assets minus their debts. We rely on RAND imputed income data, which are comparable over waves. In addition, we computed changes in assets and income between waves. Preliminary analyses revealed that asset changes and income had no influence on results, and these variables were dropped in the final models. We retained assets at baseline and change in income between waves. We logged all income measures to adjust for skewed distributions.

The other socioeconomic status variable included in the analyses was education (coded in years). We also considered current or last occupation, but this variable had no effect on results and was dropped from the final models. We further included selected job characteristics. Respondents were asked whether their jobs required lots of physical effort; lifting heavy loads; stooping, kneeling, or crouching; good eyesight; intense concentration; or skill in dealing with people. Answer categories for these items ranged from 4 (almost all of the time) to 1 (none or almost none of the time). A second item battery referred to job difficulty (my job requires me to do more difficult things than it used to; my job involves a lot of stress). Answers could range from 4 (strongly agree) to 1 (strongly disagree). On the basis of factor analyses (available from A. Davey), we created three job characteristic variables: stress (job more difficult, job involves stress; for the full sample ranges from.93 to.95 over the four waves); concentration (job requires intense concentration, requires people skills, requires good eyesight; ranges from.96 to.98); and physical effort (physical effort, lifting, stooping; ranges from.93 to.94). Items for each subscale were summed. Concentration had little effect on results and was dropped from the final models.

As postretirement well-being may be linked to individuals' work commitment, we also included indicators for this concept, namely, number of years in the current or last job (this is the best work history variable available in the HRS) and perception of the importance of work. We derived the latter variable from the question concerning whether individuals considered work mostly important as a source of income or whether they valued their work in itself; we coded it as 1 = only income important, 2 = both important, and 3 = work itself important.

Another important predictor of postretirement well-being is whether retirement is perceived as voluntary (Hardy & Quadagno, 1995). Interviewers asked retirees whether retirement "was something you wanted to do or something you felt you were forced into?" The answer categories were as follows: wanted to, forced into, and part wanted, part forced. Because the last category was not implied in the question, relatively few individuals chose that answer. We thus created one dummy variable in which those reporting exclusive "forced" retirement were coded as 1 and the other two groups were coded as 0.

The final set of controls refers to marital characteristics and family and friend relations. For marital characteristics, we included length of marriage (in years, based on RAND computations) as well as spouse's limitations in ADLs and change in spouse's ADLs over time. The latter two variables were coded the same way as those for respondents. The family–friend variables included whether respondents had grandchildren (own or spouse's; 1 = yes and 0 = no), as well as an index of social integration based on responses (1 = yes and 0 = no) to questions concerning whether respondents had children or stepchildren living within 10 miles, whether they had other relatives in the neighborhood, and whether they had friends in the neighborhood. The social integration index summed scores over these three variables and could range from 0 to 3. We measured the family–friend variables at Time 2. For length of marriage we used the baseline wave. Means and standard deviations of all variables are shown in Tables 2 and 3.

The analyses were set up to test the hypotheses separately for men and women. In Table 2 we present results for men who retired or remained continuously employed (reference) with specific wife employment status categories as predictors, whereas in Table 3 we show findings for women who either retired or remained continuously employed (reference) with husbands' employment status categories as predictors. To test specific hypotheses pertaining to spousal employment status, we designed models with divergent spouse employment status categories as reference categories. In each case, we first estimated baseline models that included all predictor and control variables without any interaction terms. Model 1a (in Tables 2 and 3) uses "spouse continuously not employed" as the reference group, and Model 1b uses "spouse employed" as the reference group; they are otherwise identical. To each of these models, we next added the interaction terms for Own retirement x Spouse's employment status in order to test Hypotheses 1a and 1b and 2a and 2b, again using "spouse continuously not employed" and "spouse employed" as the reference group in Models 2a and 2b, respectively. Finally, we included the interaction between Own retirement x Spouse's employment x Enjoyment of joint activities. Tests for multicollinearity suggested that this was not a problem for our analyses (mean variance inflation was <2.00 for all analyses excluding the interaction terms).

	RESULTS

TOP Abstract Methods Results Discussion References

 
The subsamples used for our analyses, as well as the estimated models, are geared to test the hypothesized relationships. Although results pertaining to other predictors of postretirement depressive symptoms may be of interest, we consequently focus our discussion exclusively on effects of retirement, length of retirement, spouse's employment status, and enjoyment of activities with the spouse. However, the full models including all covariates are shown in Tables 2 and 3 along with the relevant interaction effects. Preliminary analyses indicated that the effects of men's retirement were similar for 0–6 and 7–12 months postretirement, whereas those for women were different for these periods. We consequently present models for men that distinguish solely between recent (1 year) and longer-term (>1 year) retirees (Table 2), whereas the analyses for women refer to three separate postretirement periods (0–6 months, 7–12 months, and >1 year; see Table 3). All analyses control for depressive symptoms at Time 1. The results thus reflect depressive symptoms at Time 2 net of Time 1 symptoms. To ease interpretation of results, we also present the significant interaction effects in figures that were derived from the Tables and are adjusted for mean levels on all covariates.

Our initial analyses, for which we presented no hypotheses, indicated that none of the independent variables (length of retirement, spouse's employment status, and enjoyment of joint activities) exerted a significant main effect on men's depressive symptoms at Time 2 (Table 2, Models 1a and 1b). For women (Table 3, Model 1b) we find a negative effect of spouse's retirement; this suggests that women, regardless of their own employment status, reported fewer depressive symptoms if their spouse retired than if he continued working.

Next, we tested our study hypotheses. For men we hypothesized fewer depressive symptoms at Time 2 if they were recently retired and their wives were either continuously not working (Hypothesis 1a) or retired between waves (Hypothesis 1b). To test the first hypothesis (the hypothesized coefficients are set in boldface in the tables), we contrasted recently retired husbands whose wives were continuously nonworking (reference) with those whose wives remained employed (Table 2, Model 2a). The finding supports the hypothesis. As shown in Figure 1 (first two bars for recent retirees), recently retired men reported more depressive symptoms when their spouses worked (b =.90, p <.01) than when their spouses were continuously not employed (reference). No similar effect of spouse's employment shows for long-term retired husbands (b = –.10, ns in Model 2a; first two bars for long-term retired men in Figure 1).

View larger version (51K): [in this window] [in a new window]   Figure 1. Men's depressive symptoms at Time 2, by own employment–retirement, length of retirement, and spouse's employment–retirement

In addition to the hypothesized effects, the data also revealed that longer-retired men whose spouses retired between waves reported significantly fewer depressive symptoms than longer-retired husbands whose wives were continuously not employed (Model 2a, b = –.76, p <. 05, first and last bar for long-term retired men in Figure 1).

Thus, in line with Hypotheses 1a and 1b, recently retired men seem to be negatively affected by their spouses' continued employment as compared with husbands whose wives were already nonworking prior to the husband's retirement or retired jointly with him. The longer-term effects of spouse's employment status seem to depend on whether the wife was continuously nonemployed during the husband's retirement transition or retired during the same period. Longer-retired men fare particularly well if they retired jointly with their spouses but less well if their wives either remain employed or were already nonworking prior to the husband's retirement.

Among women, effects of the spouse's employment or retirement generally seem more short lived than among men; that is, differences by spouse's employment status occurred foremost among women retired 6 months or less. Our hypotheses for wives (2a and 2b) predict that short-term retired wives will report fewer depressive symptoms if they either retired jointly with their husbands (2a) or if their husbands continued to work (2b). To test these hypotheses, we first contrasted women retirees whose husbands had retired with those whose husbands were continuously not employed (Table 3, Model 2a). The data provide support for Hypothesis 2a. Very recently retired women (within 6 months) whose spouses retired between waves had significantly fewer depressive symptoms than women whose husbands were continuously not employed (Table 3, Model 2a, b = –.97, p <.05; see first and third bars for recently retired women in Figure 2). To test Hypothesis 2b, we contrasted recent women retirees whose spouses were continuously employed with those whose spouses were continuously not employed. The data again support the hypothesis. Recently retired women whose husbands remained employed reported significantly fewer depressive symptoms than those whose husbands were continuously not employed (Table 3, Model 2a, b = –1.09, p <.01; see first and second bars for recently retired women in Figure 2). In contrast to the findings for men (but in line with our hypotheses), very recent women retirees whose husbands retired jointly with them do not differ from those whose husbands remained employed (b =.11, ns in Table 3, Model 2b; second and third bars for recently retired women in Figure 2). These data confirm a honeymoon effect especially for recently retired women who either retired before or jointly with their spouses, consistent with Hypotheses 2a and 2b. We find no significant effects of spouses' employment status for women retired 7 months or more.

View larger version (45K): [in this window] [in a new window]   Figure 2. Women's depressive symptoms at Time 2, by own employment–retirement, length of retirement, and spouse's employment–retirement

View larger version (32K): [in this window] [in a new window]   Figure 3. Short-term retired men's depressive symptoms at Time 2, by spouse's employment retirement, and enjoyment of joint activities with spouse

	DISCUSSION

TOP Abstract Methods Results Discussion References

Generally, our research supports a life course approach to retirement adaptation and indicates that the retirement experience is quite heterogeneous and contingent on the specific contexts under which retirement occurs. Furthermore, the influence of such contexts may change over time or, in the case of couples, encourage retirement of the partner. This heterogeneity of retirement experiences raises questions about earlier studies that only assess the direct effect of retirement on well-being and ignore contingencies of the retirement experience (Calasanti, 1996a; Charles, 2002; Gall, Evans, & Howard, 1997). Such research will tend to show little influence of retirement on well-being as positive and negative experiences are averaged and thus provide few insights to guide retirement preparation and planning. Indeed, our data offer little support for a general effect of length of retirement on well-being, precisely because responses to retirement (including the experience of a honeymoon) seem to be dependent on selected retirement contexts. Furthermore, if a retirement honeymoon period exists at all, it is not clear, as Atchley (1976) pointed out, how long this phase lasts. Our results suggest that the honeymoon may last longer among men than among women. True tests of retirement phases probably also require more frequent interviews over time, as the analyses of Richardson and Kilty (1995) suggest as well. If, as our data show, variability in the retirement experience is particularly pronounced in the immediate aftermath of retirement, then study designs with a relatively long time span between waves (i.e., greater than 6 months) may not capture such variation.

Rather than supporting a general honeymoon effect, the results offer some support for the hypothesis that length of retirement and spouse's employment status interact in their influences on postretirement well-being. Recently retired husbands report somewhat higher depressive symptoms if their wives continue to work outside the home. This finding is consistent with earlier evidence (Moen et al., 2001; Myers & Booth, 1996; Szinovacz, 1996) as well as with our hypothesis that wives' continued employment runs counter to traditional gender role ideology and may therefore undermine retired husbands' perceived status in the marriage. Other mechanisms that may contribute to adjustment problems among recently retired husbands with still employed wives include conflicts surrounding the division of household labor and the specific reasons for husbands' retirement prior to their wives. Although retired husbands may take over some of their still-employed wives' housework, their wives still tend to carry the main responsibility for household work (Szinovacz, 2000). Although wives may accept such inequity in the division of household labor as long as their husbands are employed, they may no longer accept it after husbands' retirement, leading to marital conflicts that are reflected in husbands' depressive symptoms. In addition, as one anonymous reviewer suggested, recently retired husbands with employed wives may feel lonely and bored at home until they develop postretirement activities and friendships. It is also conceivable that husbands who retire prior to their wives do so as a result of adverse circumstances such as unemployment or poor health. Although we control for "forced" retirement and health in the analyses, these controls may not capture the full range of adverse retirement circumstances. Clearly, more detailed data are needed to understand the precise mechanisms leading to the negative influence of wives' employment on their retired husbands' postretirement well-being.

The negative effect of wives' continued employment seems to persist over time, though it is somewhat weaker and only approaches significance for longer-retired men. Couples may resolve potential conflicts that arise from this arrangement and thus reduce its negative influence. Furthermore, retired husbands who are most troubled by their wives' employment may succeed in persuading their wives to retire. Other research has shown that retired husbands whose wives are still employed exert pressure on their wives to retire as well (Skirboll & Silverman, 1992; Szinovacz, 1989). The latter explanation would imply that couples remaining in the single-retired husband status for some time are self-selected; that is, they underrepresent couples in which husbands are particularly opposed to wives' continued employment.

We further find that husbands whose wives also retired between waves are quite happy in retirement. The positive effect of wives' retirement is particularly pronounced among longer-retired husbands. This finding suggests that husbands' retirement honeymoon may be triggered by their wives in conjunction with or after their own retirement. Perhaps husbands who initially opposed their wives' continued employment are particularly happy in the aftermath of their wives' retirement.

As expected, the positive effect of both spouses' retirement on husbands' well-being is limited to couples who enjoy joint activities. Though hardly surprising, this finding once again demonstrates the contextual nature of retirement adaptation processes. It also confirms earlier research showing the importance of high marital quality and joint leisure activities for retirees' well-being (Bossé et al., 1991; Dorfman et al., 1988; Vinick & Ekerdt, 1991). However, this effect once again subsides over time. This suggests, on the one hand, that spouses who do not enjoy joint activities are able to establish satisfactory life routines after some time, but this seems to take them longer than couples who favor joint endeavors. On the other hand, the euphoria associated with joint retirement activities among couples enjoying such endeavors may also be curtailed over time. It might be difficult to maintain high involvement in particularly enjoyable joint activities such as travel over long time periods. Couples may then be disappointed that high expectations for joint retirement leisure are no longer met, and such unmet expectations have been linked to reduced postretirement well-being (Vinick & Ekerdt, 1992).

It is notable that we do not find a similar interaction with enjoyment of joint activities for wives. Two explanations for this result come to mind. First, wives are often responsible for planning couples' postretirement activities and thus may be able to gear the couple's retirement endeavors into those pursuits they themselves enjoy. In this case, couples' previous enjoyment of joint activities would have little influence on wives postretirement well-being. Second, husbands more than wives depend on the spouse for friendships and other social supports (Huyck, 1995). Thus, wives whose marriages provide few outlets for joint activities may be better able than husbands to find alternative social arrangements. Such arrangements seem to have a similar protective function against high postretirement depressive symptoms as joint spouse endeavors.

A husband's retirement into a marriage in which the wife has not been employed for some time has quite different ramifications for his well-being than joint retirement. Under this scenario, husbands show relatively few depressive symptoms in the short term, but they report more depressive symptoms after 1 year of retirement. This seems to suggest that after a brief honeymoon, husbands whose wives were not employed face some adjustment problems. On the basis of previous research, we suggest that interference and insufficient distance problems could be responsible for this finding. Wives of retired husbands sometimes feel that their husbands are "underfoot" and impinge on their realm (Ekerdt & Vinick, 1993; Schäuble, 1989; Vinick & Ekerdt, 1991). The often-cited saying "I married him for better or for worse but not for lunch" exemplifies such impingement (Vinick & Ekerdt, 1991b). Wives with retired husbands may also complain about curtailment of their privacy and the need to plan retirement activities for their husbands (Gilford, 1986; Niederfranke, 1991). Women who retired well before their husbands (or who are housewives) are most likely to have established household routines and a lifestyle that incorporated their working husbands' consistent absence from the home. These couples may have problems finding the right balance of distance and togetherness (Caradec, 1994), and wives' dissatisfaction with their husbands' impingement may undermine both spouses' well-being (linked lives). It is also possible that separate retirements are more typical for less happily married spouses who would then lack another salient life sphere after retirement.

The results for wives show more variability in postretirement depressive symptoms over short time periods of 6 months or less. These recently retired wives report fewer depressive symptoms either if they retired jointly with their husbands or if their husbands remain employed, relative to wives whose husbands had already exited the labor force. This result supports a joint retirement honeymoon experience, but in the case of wives this experience seems to be quite short lived. Wives whose husbands stopped working before the wife retired tend to report more depressive symptoms. This couple retirement pattern is most likely to induce husbands to exert pressure on their wives to retire as well (Skirboll & Silverman, 1992; Szinovacz, 1989). Such pressure may curtail wives' control over their retirement decision and eventually undermine postretirement well-being. It is also conceivable that this retirement pattern occurs most commonly among couples in which the husbands were unable to remain in the labor force (displaced or disabled men). As already noted, our controls for husbands' ADLs may be insufficient to capture all negative effects deriving from husbands' unemployment or disability. Further analyses differentiating more clearly among truly retired and displaced or disabled husbands may shed more light on this issue.

Taken together, the results from this study support a theoretical framework that combines the life course approach with selected assumptions from other perspectives. They demonstrate that it is often not the retirement transition per se that influences postretirement well-being but rather the specific circumstances or contexts under which retirement occurs or that closely follow retirement. Our analyses demonstrate such contextual effects for spouses' employment and marital quality. Although several of the hypotheses derived from our theoretical framework were supported by the data, large data sets such as the HRS rarely include the detailed information necessary for unambiguous interpretations. Thus, more research (perhaps with qualitative data) is needed to definitively explain the observed trends.

Our data further suggest that postretirement well-being is not static. This process-oriented view of retirement adaptation is consistent with an assumption of retirement phases. However, such phases may be defined not only by time but also by contextual changes. Because our data are restricted to two time points, conclusions pertaining to postretirement changes in terms of length of retirement or the relative timing of spouses' retirement are tentative. Data comparing retirees over several waves are needed to address retirement adaptation processes over time. Use of later waves of the HRS should offer such an opportunity. Furthermore, the first four waves of the HRS are restricted to mostly younger and often early (exit before age 62 or age 65) retirees. Consequently, our findings can only be generalized to this group, and couples retiring later may indeed have different retirement experiences. Once again, later waves of the HRS can be used to explore this possibility.

Another limitation of our research pertains to the measure of marital quality (enjoy time together) and number of cases. The marital quality measures included in the HRS are quite basic and highly skewed. Given that interaction effects such as those tested here require quite large samples, especially to test effects for more marginal groups (e.g., long-married spouses who do not enjoy time together), we may very well have found stronger effects of this variable had we not recoded the data. However, this would have led to extremely low cell counts and unstable results. Once more HRS respondents have made the retirement transition, it should become possible to explore postretirement experiences among such marginal groups.

Despite these caveats, our results lend themselves to some suggestions for couples, policy makers, and practitioners. Couples need to plan not only for their postretirement finances but also for their postretirement lifestyle. The current problems associated with the wife's continued employment after the husband's retirement or with husbands who pressure employed wives to retire may become less pronounced as less gender-role traditional cohorts approach retirement. Nevertheless, careful planning may help couples to achieve retirement timing patterns that accommodate both spouses. Not only spouses' own actions but retirement policies and programs that support couples' needs are essential to further postretirement well-being. Because husbands are typically somewhat older than their wives, current Social Security benefit regulations (and many pension plans) reinforce a retirement pattern in which husbands retire first. Couples tend to respond by some adjustments in retirement age (Szinovacz, 2002), which can lead to benefit reductions especially for wives. Proposals that emphasize a life course perspective of labor force attachments and leaves would be more in line with retiring (and younger) couples' needs.

Similarly, couples need to learn, perhaps through retirement preparation programs or in more extreme cases marital therapy, how to develop mutually enjoyable endeavors and how to find the best distance in their postretirement relationship. To be more useful to couples, retirement preparation programs require some changes. The nearly exclusive emphasis on financial issues may distract couples from other potential retirement problems. Couples would benefit more from these programs were they to adopt a broader and research-based perspective.

Our research demonstrates the validity of the life course perspective and the importance of marital context for postretirement well-being: How well spouses fare in retirement is contingent on spouses' employment status as well as marital quality. They also show that her and his retirement experiences are indeed different. It is only by capturing such complexities that we will gain full understanding of retirement adaptation processes.

	Acknowledgments

 
This study was funded by the National Institute on Aging under Grant R01 AG13180 (Maximiliane E. Szinovacz, principal investigator). The analyses rely on data from the HRS public release and imputed data files (as of December 2002). Selected variables were taken from the data set compiled by RAND (first version).

The HRS is managed by the University of Michigan. Detailed information on the data set is available from their Web site (http://www.umich.edu/hrswww/).

	Footnotes

 
Decision Editor: Margie E. Lachman, PhD

Received for publication July 3, 2002. Accepted for publication March 15, 2004.

	References

TOP Abstract Methods Results Discussion References

 

• Allmendinger, J., Brückner, H., Brückner, E. (1991). Auswirkungen von Familien- und Erwerbsarbeit auf die Situation im Ruhestand [Effects of family labor and employment on the retirement experience]. In K. U. Mayer, J. Allmendiger, & J. Huinink (Eds.), Vom Regen in die Traufe [Out of the frying pan into the fire] (pp. 423–459). Frankfurt, Germany: Campus.
• Amemiya, T. (1984). Tobit models: A survey. Journal of Econometrics, 24,3-61.
• Atchley, R. C. (1976). The sociology of retirement. New York: Wiley.
• Atchley, R. C. (1992). Retirement and marital satisfaction. In M. Szinovacz, D. Ekerdt, & B. H. Vinick (Eds.), Families and retirement (pp. 145–158). Newbury Park, CA: Sage.
• Bernard, J. (1973). The future of marriage. New York: Bantam.
• Blau, D. M. (1998). Labor force dynamics of older married couples. Journal of Labor Economics, 16,595-629.
• Bossé, R., Aldwin, C. M., Levenson, M. R., Workman-Daniels, K. (1991). How stressful is retirement? Findings from the Normative Aging Study. Journal of Gerontology: Psychological Sciences, 46,P9-P14.
• Burke, P. J. (1991). Identity processes and social stress. American Sociological Review, 56,836-849.
• Burke, P. J. (1996). Social identities and psychosocial stress. In H. B. Kaplan (Ed.), Psychosocial stress. Perspectives on structure, theory, life-course, and methods (pp. 141–174). San Diego, CA: Academic Press.
• Calasanti, T. M. (1996a). Gender and life satisfaction in retirement: An assessment of the male model. Journal of Gerontology: Social Sciences, 51B,S18-S29.
• Calasanti, T. M. (1996b). Incorporating diversity: Meaning, levels of research, and implications for theory. The Gerontologist, 36,147-156.[Abstract]
• Caradec, C. (1994). Le problème de la "bonne distance" conjugale au moment de la retraite (Finding the right conjugal distance once retired). Revue Française de Sociologie, 35,101-124.
• Carp, F. M. (1997). Retirement and women. In J. M. Coyle (Ed.), Handbook on women and aging (pp. 112–128). Westport, CT: Greenwood Press.
• Charles, K. K. (2002). Is retirement depressing? Labor force inactivity and psychological well-being in later life (Working Paper w9033). Washington, DC: National Bureau of Economics.
• Davey, A. (2001). An analysis of incomplete data. In L. M. Collins & A. G. Sayer (Eds.), New methods for the analysis of change (pp. 379–383). Washington, DC: American Psychological Association.
• Davey, A., Shanahan, M. J., Schafer, J. L. (2001). Using multiple imputation to correct for selective attrition in the National Longitudinal Study of Youth. Journal of Human Resources, 36,500-519.
• Davey, A., Szinovacz, M. E. (2004). Dimensions of marital quality and retirement. Journal of Family Issues, 24,431-464.
• Diehl, M. (1999). Self-development in adulthood and aging: The role of critical life events. In C. D. Ryff & V. M. Marshall (Eds.), The self and society in aging processes (pp. 150–186). New York: Springer.
• Dorfman, L. T., Heckert, D. A., Hill, E. A., Kohout, F. J. (1988). Retirement satisfaction in rural husbands and wives. Rural Sociology, 55,25-39.[Medline]
• Efron, B., Tibshirani, R. (1986). Bootstrap methods for standard errors, confidence intervals, and other measures of statistical accuracy. Statistical Science, 1,54-75.
• Ekerdt, D. J., Bossé, R., Levkoff, S. (1985). An empirical test of phases of retirement: Findings from the normative aging study. Gerontology, 40,95-101.
• Ekerdt, D. J., Vinick, B. H. (1993, November). The husband underfoot: Explaining resilient ideas about retirement's effect on marriage. Paper presented at the Annual meeting of The Gerontological Society of America, New Orleans, LA.
• Fahrmeir, L., Tutz, G. (1994). Multivariate statistical modeling based on generalized linear models. New York: Springer-Verlag.
• Gall, T. L., Evans, D. R., Howard, J. (1997). The retirement adjustment process: Changes in the well-being of male retirees across time. Journal of Gerontology: Psychological Sciences, 52B,P110-P117.
• Gilford, R. (1986). Marriages in later life. Generations, 10,16-20.
• Greene, W. H. (1990). Econometric analysis. New York: MacMillan.
• Gustman, A. L., Mitchell, O. S., Steinmeier, T. L. (1995). Retirement measures in the health and retirement study. The Journal of Human Resources, 30,S57-S83.
• Hagestad, G. O. (1990). Social perspectives on the life course. In R. H. Binstock & L. K. George (Eds.), Handbook of aging and the social sciences (pp. 151–168). San Diego, CA: Academic Press.
• Hardy, M. A., Quadagno, J. (1995). Satisfaction with early retirement: Making choices in the auto industry. Journal of Gerontology: Social Sciences, 50B,S217-S228.
• Haug, M. R., Belgrave, L. L., Jones, S. (1992). Partners' health and retirement adaptation of women and their husbands. Journal of Women and Aging, 4,5-29.
• Heeringa, S. G., Rodgers, W. L. (1998, November). Information loss in longitudinal surveys. Types, patterns, simple compensation methods. Paper presented at the annual meeting workshop of The Gerontological Society of America, Philadelphia, PA.
• Henkens, K. (1999). Retirement intentions and spousal support: A multi-actor approach. Journal of Gerontology: Social Sciences, 54B,S63-S74.
• Huyck, M. H. (1995). Marriage and close relationships of the marital kind. In R. Blieszner and V. H. Bedford (Eds.), Handbook of aging and the family (pp. 181–200). Westport, CT: Greenwood Press.
• Juster, F. T., Suzman, R. (1995). An overview of health and retirement study. The Journal of Human Resources, 30,S7-S56.
• Kelly, J. R., Westcott, G. (1991). Ordinary retirement: Commonalities and continuity. International Journal of Aging and Human Development, 32,81-89.[Medline]
• Kessler, R. C., Price, R. H., Wortman, C. B. (1985). Social factors in psychopathology: Stress, social support, and coping processes. Annual Review of Psychology, 36,531-572.[Medline]
• Kim, J. E., Moen, P. (2002). Retirement transitions, gender, and psychological well-being: A life-course, ecological model. Journal of Gerontology: Psychological Sciences, 57B,P212-P222.
• Kosloski, K., Ginsburg, G., Backman, C. W. (1984). Retirement as a process of active role transition. In V. L. Allen & E. V. d. Vliert (Eds.), Role transitions: Explorations and explanations (pp. 331–341). New York: Plenum Press.
• Kulik, L. (1996). Perceptions of conjugal relations among Israeli males—pre-retirement versus early and late retirement phases. Contemporary Family Therapy, 18,115-128.
• Kulik, L., Bareli, H. Z. (1997). Continuity and discontinuity in attitudes toward marital power relations: Pre-retired vs. retired husbands. Aging and Society, 17,571-595.
• Lee, G. R., Shehan, C. L. (1989). Retirement and marital satisfaction. Journal of Gerontology: Social Sciences, 44,S226-S230.
• Little, R. J. A., Rubin, D. B. (1989). The analysis of social science data with missing values. Sociological Methods and Research, 18,292-326.
• Long, J. S. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks, CA: Sage.
• Marshall, V. W., Clarke, P. J., Ballantyne, P. J. (2001). Instability in the retirement transition. Research on Aging, 23,379-409.[Abstract/Free Full Text]
• Moen, P., Kim, J. E., Hofmeister, H. (2001). Couples' work/retirement transitions, gender, and marital quality. Social Psychology Quarterly, 64,5-71.
• Myers, S. M., Booth, A. (1996). Men's retirement and marital quality. Journal of Family Issues, 17,336-358.[Abstract]
• Niederfranke, A. (1989). Bewältigung der vorzeitigen Berufsaufgabe bei Männern [Men's adaptation to early retirement]. Zeitschrift für Gerontologie, 22,143-150.[Medline]
• Niederfranke, A. (1991). Lebensentwürfe von Frauen beim Übergang in den Ruhestand [Life designs of retiring women]. In C. Gather, U. Gerhard, K. Prinz, & M. Veil (Eds.), Frauen—Alterrsicherung [Women—old age security] (pp. 279–291). Berlin: Sigma.
• O'Rand, A. M., Henretta, J. C. (1999). Age and inequality: Diverse pathways through later life. Boulder, CO: Westview Press.
• Pearlin, L. I., Lieberman, M. A., Menaghan, E. G., Mullan, J. T. (1981). The stress process. Journal of Health and Social Behavior, 22,337-356.[Medline]
• Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1,385-401.[Abstract]
• Reitzes, D. C., Mutran, E. J., Fernandez, M. E. (1996). Preretirement influences on postretirement self-esteem. Journal of Gerontology: Social Sciences, 51B,S242-S249.
• Richardson, V., Kilty, K. M. (1991). Adjustment to retirement: Continuity vs. discontinuity. International Journal of Aging and Human Development, 33,151-169.[Medline]
• Richardson, V., Kilty, K. M. (1995). Gender differences in mental health before and after retirement: A longitudinal analysis. Journal of Women and Aging, 7,19-36.
• Ross, C. E., Drentea, P. (1998). Consequences of retirement activities for distress and the sense of personal control. Journal of Health and Social Behavior, 39,317-334.[Medline]
• Schafer, J. L. (1997). Analysis of incomplete multivariate data. London: Chapman & Hall.
• Schäuble, G. (1989). Die schönsten Jahre des Lebens?. [The best years of life?]. Stuttgart: Enke.
• Settersten, R. A. (2003). Invitation to the life course: The promise. In R. A. Settersten (Ed.), Invitation to the life course: Toward new understanding of later life (pp. 1–12). Amityville, NY: Baywood.
• Skirboll, E., Silverman, M. (1992). Women's retirement: A case study approach. Journal of Women and Aging, 4,77-90.
• Slevin, K. F., Wingrove, C. R. (1995). Women in retirement—A review and critique of empirical research since 1976. Sociological Inquiry, 65,1-20.
• Stata. (2003). Stata user's guide. [Version 8]. College Station, TX: Stata Press.
• Steffeck, D. E. (2000). Documentation of affective functioning measures in the Health and Retirement Study (HRS/AHEAD Documentation Report DR-005). Ann Arbor, MI: University of Michigan, Survey Research Institute.
• Szinovacz, M. (1989). Decision-making on retirement timing. In D. Brinberg & J. Jaccard (Eds.), Dyadic decision making (pp. 286–310). New York: Springer.
• Szinovacz, M. (1991). Women and retirement. In B. B. Hess & E. W. Markson (Eds.), Growing old in America (4th ed., pp. 293–304). New Brunswick, NJ: Transaction Books.
• Szinovacz, M. (1996). Couple's employment/retirement patterns and marital quality. Research on Aging, 18,243-268.[Abstract]
• Szinovacz, M. (2000). Changes in housework after retirement: A panel analysis. Journal of Marriage and the Family, 62,78-92.
• Szinovacz, M. E. (2002). Couple retirement patterns and retirement age. International Journal of Sociology, 32,(2), 30-54.
• Szinovacz, M., DeViney, S. (2000). Marital characteristics and retirement decisions. Research on Aging, 22,470-489.[Abstract/Free Full Text]
• Szinovacz, M., Ekerdt, D. J. (1995). Families and retirement. In R. Blieszner & V. H. Bedford (Eds.), Handbook on aging and the family. Westport, CT: Greenwood Press.
• Szinovacz, M., Harpster, P. (1993). Employment status, gender roles and marital dependence in later life. Journal of Marriage and the Family, 55,927-940.
• Szinovacz, M., Schaffer, A. M. (2000). Effects of retirement on marital conflict management. Journal of Family Issues, 21,367-389.[Abstract/Free Full Text]
• Thoits, P. A. (1991). On merging identity theory and stress research. Social Psychology Quarterly, 54,101-112.
• Vinick, B. H., Ekerdt, D. J. (1991). The transition to retirement: Responses of husbands and wives. In B. B. Hess & E. W. Markson (Eds.), Growing old in America (4th ed., pp. 305–317). New Brunswick, NJ: Transaction Books.
• Vinick, B. H., Ekerdt, D. J. (1992). Couples view retirement activities: Expectation versus experience. In M. Szinovacz, D. J. Ekerdt, & B. H. Vinick (Eds.), Families and retirement (pp. 129–144). Newbury Park, CA: Sage.
• Wheaton, B. (1990). Life transitions, role histories, and mental health. American Sociological Review, 55,209-223.
• Wheaton, B. (1996). The domains and boundaries of stress concepts. In H. B. Kaplan (Ed.), Psychosocial stress (pp. 29–70). San Diego, CA: Academic Press.
• Zimmerman, L., Mitchell, B., Wister, A., Gutman, G. (2000). Unanticipated consequences: A comparison of expected and actual retirement timing among older women. Journal of Women and Aging, 12,109-128.

This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation