| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Glennan Center for Geriatrics and Gerontology, Eastern Virginia Medical School, Norfolk
b Department of Social Science, University of Maryland Eastern Shore, Princess Anne
c Department of Child and Family Development, University of Georgia, Athens
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 Theoretical BackgroundMethodsResultsDiscussionReferences |
|---|
This study examined whether economic and care obligations for family and kin salience influence retirement decisions and whether such influences differ by race, gender, and marital status.
Data.
Data from the first 2 waves of the National Survey of Families and Households were used. The sample consisted of individuals who were employed at baseline and aged 55–75 at follow-up (N = 897). Cox proportional hazard regressions were used.
Results.
Economic kin obligations impeded the likelihood of retiring. Individuals who made financial contributions to children outside the household and White women with resident children in the household were less prone to retire. Among Blacks, household composition effects were more complex and seemed to depend on the mix of care obligations, financial obligations, and financial contributions by resident kin. Kin salience also impinged on retirement decisions. Some groups who lacked family ties (e.g., nonmarried childless men) were less inclined to retire, whereas other subgroups (e.g., nonmarried men with monthly pre-retirement contacts with children) were more likely to retire.
Conclusions.
Our data suggest the importance of family obligations and relationships in retirement decisions and demonstrate considerable diversity in these processes. Models of retirement should pay greater attention to the interdependence of work and family spheres and to the diversity of retirement processes among various population groups.
PREDICTORS of retirement behavior or labor force withdrawal in the later years are typically derived from rational choice models that focus on individuals' own characteristics (Szinovacz and Ekerdt 1995
) and often rely exclusively on male samples (Hayward, Friedman, and Chen 1996, Hayward, Friedman, and Chen 1998). Using concepts from the life course perspective (Bengtson and Allen 1993; Elder 1995; Moen 1996; O'Rand 1996), we tested a model of retirement decisions that pays tribute to marital and familial contingencies of retirement processes and cultural embeddedness. Our objective was to assess whether selected family obligations and relationships influence retirement decisions and whether these influences differ by gender, race, and marital status. We used data from the National Survey of Families and Households (NSFH) for the analyses.
|
Theoretical Background |
|---|
|
TOPAbstractTheoretical BackgroundMethodsResultsDiscussionReferences |
|---|
; Elder 1995; Moen 1996). Interdependence of life spheres means that work and family spheres are linked throughout individuals' lives and that life experiences in one domain are connected with experiences in the other domain. This interdependence implies that retirement decisions derive not only from individual circumstances but also from the family situation. Contextual embeddedness draws attention to individuals' placement in the social structure. Both work and family experiences differ among societal subgroups, and these divergent experiences are likely to inform retirement decisions. In line with these assumptions, we based our theoretical model on the premise that economic and care obligations for family, as well as salience of kin ties, influence retirement decisions. Furthermore, we expected that the impact of these family obligations and ties would vary by gender, race, and marital status.
Interlinked life spheres.
Family considerations that may enter retirement decision processes include economic family demands, care responsibilities that compete with work obligations, and the salience of the family sphere. Past research has suggested that retirement decisions are driven by economic considerations such as benefit eligibility, financial assets, and the expected economic utility of remaining in or leaving the labor force (Burr, Massagli, Mutchler, and Pienta 1996; Fronstin 1999; Honig 1996, Honig 1998; Pienta 1999; Pienta, Burr, and Mutchler 1994; Wray 1996). Financial obligations for family members can also impinge on retirement decisions. Individuals with financial obligations for dependents tend to remain in the labor force and adopt less positive attitudes toward retirement (Clark, Johnson, and McDermed 1980; Hayward et al. 1998; Henkens 1999; Henkens and Tazelaar 1997; Turner, Bailey, and Scott 1994). Financial support to family members outside the household may further curtail the feasibility of retiring. Because individuals' main financial responsibilities in later life are toward their children, we focused on such obligations. Specifically, we expected a lower probability of retirement among individuals who either have children in the household or financially support children outside the household.
Women, especially, often find it difficult to simultaneously attend to family and work obligations (Ferree 1990; Lopata 1995). Whereas care obligations for children compete with work demands during parents' early life and midlife, care obligations for other family members (e.g., ill parents or spouses) may take precedence in the later years. Earlier research showed effects of care for ill spouses on labor force withdrawal, but the direction of this effect remains unclear (Hayward et al. 1998; Honig 1996; Szinovacz and DeViney 2000; Talaga and Beehr 1995). Similarly, care obligations for kin may impinge on retirement decisions. Previous research linking care obligations for ill relatives to employment decisions yielded contradictory evidence (Hatch and Thompson 1992; Pavalko and Artis 1997; Ruhm 1996; White-Means and Chollet 1996). Most of these studies included caregivers at different life stages. For persons who are close to or have reached retirement age, withdrawal from the labor force for caregiving purposes may be more feasible than for younger caregivers. However, care for ill relatives sometimes involves financial costs that may render retirement more difficult (Ruhm 1996). Thus, the overall impact of caregiving on retirement decisions most likely depends on the balance between required care effort and the financial costs of the care arrangement.
Past research has demonstrated the viability of disposition factors for retirement decisions. Individuals' propensity to retire has been linked to work commitment, workplace attributes, and perceived benefits and costs of retirement for personal or marital pursuits (George, Fillenbaum, and Palmore 1984; Henkens 1999; Henkens and Tazelaar 1997; Honig 1998; Ruhm 1996; Szinovacz and DeViney 2000; Taylor and Shore 1995). Family ties may influence disposition toward retirement as well. Employment curtails how much time individuals can devote to family activities, but it also offers chances for social contacts and self-fulfillment outside the family context. We expected that salient kin ties would motivate individuals to retire, especially if their work obligations precluded frequent kin interactions. On the other hand, individuals lacking salient family ties may be less inclined to retire anticipating that withdrawal from the labor force will curtail opportunities for social and personal fulfillment.
Contextual embeddedness.
Family ties and obligations differ among diverse subpopulation groups. Most important among these are variations by gender, race, and marital status.
Although men continue to be viewed as main economic providers, women carry major responsibility for noneconomic family responsibilities and the management of kin relationships (Ferree 1990; Rossi and Rossi 1990; Silverstein, Lawton, and Bengtson 1994). This gender difference is likely to influence how kin obligations and relationships affect men's and women's retirement decisions. We expected that men's retirement decisions would reflect financial kin obligations (see also Pienta 1999; Pienta et al. 1994; Talaga and Beehr 1995), whereas care obligations and salience of kin ties would be more influential for women's retirement decisions.
In view of data limitations (the small number of ethnic minorities other than African Americans), we were able to assess differences between Whites and Blacks only. Many studies have indicated that the kin networks of African Americans are particularly close and characterized by mutual economic and support exchanges, within and often across households (Chatters and Jayakody 1995; Scott and Black 1994). Furthermore, African American kinship networks seem to be matrifocal, that is, organized through female kinship ties (Scott and Black 1994). We therefore expected a stronger influence of family obligations and relationships for Blacks than for Whites and that this relationship would prevail among women.
Because kin relationships are typically managed by women, men's access to kin is often mediated through their wives' efforts (Antonucci 1990; DiLeonardo 1987), and unmarried men sometimes lack close kin ties (Cooney and Uhlenberg 1990; Szinovacz 1992). Although unmarried women rarely lose contact with kin, they may develop a strong commitment to their work and the social contacts such employment affords, and they often lack financial resources (O'Rand 1996). Consequently, we expected kin obligations and relationships to have a stronger impact on married than on nonmarried individuals' retirement.
|
Methods |
|---|
|
TOPAbstractTheoretical BackgroundMethodsResultsDiscussionReferences |
|---|
). Weights adjust for nonresponses to the survey at Time 1 and attrition between waves.
Because labor force withdrawal and self-definitions of retirement can diverge (Ekerdt and DeViney 1990; Flippen and Tienda 2000; Gibson 1991; Hayward, Hardy, and Liu 1994; Mutchler, Burr, Pienta, and Massagli 1997; Szinovacz and DeViney 1999), we relied on a subsample that consisted of respondents at "risk" of retirement and excluded individuals who were looking for a job and those whose retirement was ambiguous (i.e., who left the labor force but self-defined as not retired or who were employed and self-defined as retiree). We considered individuals at risk of retirement if they were employed 20 h/week or more at Time 1 and aged 55–75 at Time 2. We chose 20 work hours at Time 1 to eliminate individuals engaged in sporadic employment activities. These criteria led to exclusion of 13% of cases. Sensitivity analyses (comparison of results based on different sample specifications) indicated that effects of the family variables considered in our analyses were unaffected by sample selection. Because of these exclusions, our findings pertain to retirement and may not be applicable to displaced workers or to persons with sporadic employment.
Missing cases may lead to biased results, especially when cases are not missing completely at random, as was the case for our data (Little's MCAR test p < .001; Little and Rubin 1987). To avoid such bias, we created nine complete-case data sets using the imputation regression method provided by SPSS (SPSS 1997). The results were based on multiple imputation inference using NORM (Schafer 1997). Comparison of results derived from the imputation regressions with those we obtained using expectation-maximization imputation suggested that imputation method did not alter findings.
Dependent Variable
The dependent variable was time to exit from the labor force, measured in months, from Wave 1 to the date the respondent stopped working. If he or she did not retire, time was measured from the Wave 1 to the Wave 2 interview. A dummy variable for stopped working was used as right-hand censuring variable. In our weighted sample, 35% of respondents retired between waves.
Family Obligations and Relationships
Presence of children in the household served as an indicator of financial obligations. It was derived from household rosters and was captured by two dummy variables—those with one child and those with more than one child in the household. Those without children in the household served as reference. We also constructed other measures of children's presence in the household (by children's gender and age). Results pertaining to these measures are not shown in the tables but are addressed in the text.
At baseline respondents were asked whether they or their spouse regularly provided financial assistance to children aged 19 or more. Respondents indicated each child to whom such assistance was given. Because very few respondents financially supported more than one or two children, we created a dichotomous measure (1 = "financial assistance given," 0 = "no financial assistance given").
Care obligations were based on questions about whether and for whom respondents provided care inside the household and on questions pertaining to spouse's health. We constructed two dichotomous measures, differentiating between care for spouses and care for other family members. Spousal care was coded 1 for respondents who reported that they cared for a disabled spouse in the household or whose spouses indicated that they either had limitations in activities of daily living or that their health was poor or very poor. Care for other family members was derived from questions about respondents' care obligations. Those providing care were coded 1. Note that care for spouses and other family members could overlap although this was very rare.
We relied on three indicators to assess salience of kin ties: whether respondents had children, whether their parents were still alive, and the frequency of contacts with children outside the household. Childlessness and both parents' deaths (1 = "yes," 0 = "no") were both represented as dichotomous variables. The NSFH contains questions about frequency of face-to-face contacts ("How often do you see ... ") with each adult child outside the household (1 = "not at all," 6 = "several times a week"). We constructed two dummy variables: respondents reporting at least weekly visits with any child and those reporting fewer than weekly but at least monthly visits (1 = "has weekly visits," 1 = "has monthly visits"). Respondents who saw their children less frequently (several times a year or less) served as reference category.
Controls
Numerous studies (for reviews, see Beehr 1986; Feldman 1994; Leonesio 1996; Ruhm 1996) have demonstrated that retirement decisions are influenced by the economic feasibility of retiring (i.e., whether continued employment is economically necessary or beneficial), barriers to continued employment (illness, job displacement), and disposition toward retirement (psychological and social benefits and costs of leaving the labor force). The NSFH contains some indicators for each of these domains but lacks detailed information on pension eligibility and anticipated pension wealth. Because pension eligibility and wealth tend to be closely associated with industry (core versus periphery), occupational characteristics, and work history (DeViney and Solomon 1995; O'Rand 1986; O'Rand and Maclean 1986), we initially performed models that included indicators for core and periphery industry, major Census occupation groups (managerial, technical, service, farm, crafts, labor), earnings, work hours, years in the labor force, and nonwage income and savings in addition to the work and economic variables maintained in the models and described later. These variables had no significant effect on likelihood of retirement in our models, and their inclusion or exclusion did not influence the effect of the family variables. We consequently dropped them from the final analyses to enhance parsimony and reduce potential multicollinearity. It is, of course, conceivable that other variables not contained in the NSFH (e.g., pension eligibility and wealth) may alter the impact of the family variables considered here.
Individuals' economic ability to leave the labor force was captured by indicators of benefit eligibility, socioeconomic status, and variables related to employment history. The NSFH does not contain direct measures of Social Security eligibility. Instead, we employed age categories (60–61, 62–64, 65 and over). Persons under age 60 were the reference. Concerning pensions, NSFH respondents reported whether they were enrolled in a pension plan and whether they expected to or were then receiving a pension. The wording of the last two items is such that it was impossible to distinguish individuals who had reached pension eligibility but deferred their claims from those who had not reached eligibility. However, we could distinguish persons who were enrolled in pension plans from those who were not (1 = "no pension").
In addition to benefit eligibility, we included indicators of work context and employment history. Occupational prestige influences work commitment (Hayward et al. 1998) and also helps define pension eligibility (DeViney 1995; O'Rand 1996). Occupational prestige was obtained from descriptions of respondents' current or most recent job and was based on Stevens and Cho 1985 recoding of the 1980 Census occupational classification scheme. In view of the restricted pension measure and reference to current rather than main career job, we further included several work context variables that may help distinguish individuals in bridge jobs from those who were still in their main career job at Time 1. These variables were work disruptions between waves (but prior to retirement if it occurred) and self-employment. The measure of work disruptions relied on detailed work histories and was coded as a dichotomous variable (1 = "did disrupt employment"). Self-employment was based on whether respondents derived income from self-employment (1 = "yes"). Because spouse's earnings or pensions contribute to an individual's economic situation and have previously been linked to retirement decisions (Anderson, Clark, and Johnson 1980; Burkhauser, Couch, and Phillips 1996; Clark et al. 1980; Fronstin 1999), we further added indicators for spouses' employment (1 = "employed at Time 1") and enrollment in a pension plan (1 = "yes") to our models.
Health in general and disability in particular exert a strong influence on people's decisions to remain in the labor force (Bound, Schoenbaum, and Waidmann 1995; Burr et al. 1996; Hayward et al. 1998; Pienta et al. 1994; Pienta 1999; Santiago and Muschkin 1996; Wray 1996). Self-reported health was measured with the question: "Compared to other people your age, how would you describe your health?" (1 = "very poor," 5 = "excellent"). Disability was measured by reporting of health problems that limit employment (1 = "yes"). Some individuals may have encountered health problems between waves. We could not date such changes, and their inclusion as a predictor of retirement would hence be problematic. However, we tested whether inclusion of change in health between waves affected the influence of the family variables and found no effect.
Past research has further suggested that individuals' work commitment may motivate them to remain in the labor force (George et al. 1984; Hayward and Grady 1990; Hayward et al. 1998; Taylor and Shore 1995). This possibility was measured with an indicator of job dissatisfaction. Respondents were asked if they agreed or disagreed with the following statement: "The job I do is one of the most satisfying parts of my life" (1 = "strongly agree," 5 = "strongly disagree").
Background
Background characteristics were marital status (1 = "not married"), race (two dummy variables, one for Blacks and one for all other minorities; White was the reference category), and gender (1 = "female"). Some individuals' marital status changed between waves. However, inclusion of such changes had no effect on the influence of the family variables. We assessed all independent variables except age at baseline to ensure accurate causal attributions. Means and standard deviations of the independent variables are shown in Table 1 .
|
; Yamaguchi 1991). Cox proportional hazard models (often called event history analysis) are a statistical method for the analysis of duration or survival data. Cox models allow for censored data—that is, not all events (e.g., retirement) need to occur for all persons—and account both for the occurrence of an event (retirement) and when it occurred (time of retirement for those who retired). In Table 2 we present hazard ratios (the exponentiated regression coefficient). A ratio greater than 1 means that a predictor increases the chance of retirement, and a ratio less than 1 denotes a decreased chance of retirement.
|
) between the family variables and gender, race, and marital status, starting with three-way interactions (either Gender x Race x Family Variable or Gender x Marital Status x Family Variable). We had too few cases to test for Race x Marital Status interactions or for higher order interaction terms. We tested interaction terms with each family variable in separate models to limit multicollinearity. If the three-way interaction was not significant, we then tested (again in separate models) for lower order interactions (e.g., Gender x Family Variable, Race x Family Variable). Because higher interaction terms are often difficult to interpret from mere examination of the regression coefficients, we further calculated linear predictions from the estimated models for the pertinent subgroups. These data are not shown but we used them to explain the reported interaction effects.
All of the analyses were performed with sampling weights. The NSFH Wave 2 weights accounted for oversamples, nonresponses at baseline, and attrition. Use of weights, in combination with robust estimates of variance specifically designed for complex samples, allows adjustment for dependency within the primary sampling units in the NSFH and for stratification (Stata 1999). Variance inflation tests (Fox 1991) suggested that multicollinearity was not a problem in our analyses.
|
Results |
|---|
|
TOPAbstractTheoretical BackgroundMethodsResultsDiscussionReferences |
|---|
22 difference = 168.94, df = 9, p < .001 for all respondents). This finding supported our main assumption that family experiences contribute to retirement decisions. The only significant main effect, however, was that persons providing financial help to children outside the household were less prone to retire (Table 2 , Model 2). This trend seemed to hold across subpopulations (i.e., there were no significant interactions). Although only one main effect emerged, several interaction effects were detected. The effect of children in the household differed by race and gender (see Table 2 , Model 3). Black women with one child in the household were more likely to retire, whereas White women and Black men with one child in the household were less inclined to retire. Additional analyses (not shown) further revealed that Blacks (regardless of gender) with two or more children in the household demonstrated a particularly high propensity toward retirement. Although the latter interaction by race was significant (b = 1.23, p < .01), inclusion of this interaction term did not contribute to model fit.
Because care obligations for ill relatives (other than spouses) inside the household were quite rare, we could only test for simple interactions (by gender, race, marital status). Nonetheless, the data (not shown) suggested that effects of care inside the household varied by race. Blacks showed a decreased tendency to retire when they had care obligations for relatives inside the household (interaction b = -1.56, p < .05). However, inclusion of this interaction term did not contribute to model fit.
An interaction effect involving childlessness, gender, and marital status was also identified (Table 2 , Model 4). Women with children were more prone to retire, whereas childless married women and childless nonmarried men were less inclined to retire. Similarly, there were interaction effects comprising contacts with children, gender, and marital status (Table 2 , Model 5). Men who did not have weekly contacts with their children were more and women lacking weekly contacts with children were less likely to retire. Furthermore, nonmarried men with monthly contacts were particularly prone to retire, whereas women (regardless of marital status) and nonmarried men who did not see their children on a monthly basis were less inclined to retire.
|
Discussion |
|---|
|
TOPAbstractTheoretical BackgroundMethodsResultsDiscussionReferences |
|---|
; Szinovacz and Ekerdt 1995) and suggests that individual-oriented models of retirement processes are incomplete.
There also is evidence for gender, race, and marital status variations in the influence of selected family variables, confirming the diversity of retirement processes across population groups (Calasanti 1993, Calasanti 1996; Szinovacz 1991). Specifically, Black women with one child in the household tend toward retirement, whereas White women and Black men with one child in the household are less prone to retire. Additional findings (not shown) further reveal that household composition effects may differ by co-resident children's gender and age. Because gender, age, and number of children in the household correlate, we cannot determine which of these factors is most relevant. The complexity of the household composition effects suggests that children's presence in the household reflects a variety of conditions with diverse consequences for retirement decisions. It can imply an economic liability but may also denote care obligations (for younger or disabled children and, perhaps, co-resident grandchildren) or even economic help from children to parents. Clearly, more research is needed to unravel the meaning of household composition for retirement decisions.
There is little support for our assumption that care for ill spouses or other disabled family members relates to retirement decisions. This finding needs to be viewed with some caution. Other analyses based exclusively on NSFH couples have suggested that men with care responsibilities for their wives are more prone to retire (Szinovacz and DeViney 2000). As far as care for other family members is concerned, our analyses are seriously restricted by the small number of caregivers in the sample and the lack of information on extent of care responsibility. These limitations may also account for contradictory findings concerning the impact of caregiving on employment in previous studies (Hatch and Thompson 1992; Pavalko and Artis 1997; Ruhm 1996; White-Means and Chollet 1996). Furthermore, some care may involve financial costs, thus hindering rather than promoting withdrawal from the labor force, and these effects may cancel each other out.
Salience of family ties seems to play some role in retirement decisions. Effects are most pronounced for ties to children, most likely due to individuals' greater stake in relationships with their offspring than with other kin (Giarrusso, Stallings, and Bengtson 1995). Once again, these influences seem to be subgroup-specific. Effects of childlessness differ by gender and marital status. That nonmarried childless men are particularly unlikely to retire is in line with our expectation that lack of salient family ties may induce individuals to remain in the labor force. Of course, childlessness could itself derive from very high work demands or, alternatively, long-term exposure to marginal economic positions that reduce retirement benefits and hence options for voluntary earlier retirement (Hayward et al. 1998). Childless married women may be particularly career oriented and consequently inclined to postpone their retirement. Thus, lack of salient family ties could influence retirement decisions either through its direct impact on disposition toward retirement or through its association with and reflection of career opportunities and decisions over the life course.
There also is some support for our contention that frequency of contacts with children relates to retirement decisions. The reduced likelihood of retirement among nonmarried men who have no monthly contacts with children may again be indicative of this group's special situation both in regard to kin salience and work history. Similarly, women who lack weekly or monthly contacts with children are less inclined to leave the labor force. These women may be faced with undue barriers (e.g., geographical distance, estrangement) to enhance contacts after retiring and thus be less disposed toward retirement. On the other hand, some men (those who do not have weekly contacts with children and unmarried men with monthly contacts) may see in retirement an opportunity to catch up with their children.
Taken together, our analyses suggest that kin relations play a role in retirement decisions. However, data limitations and the complexity of findings (especially subgroup differences) preclude unambiguous causal interpretations. Specifically, the relatively small sample size for the present analyses and consequently small cell frequencies for selected subgroups render testing of interactions difficult. Multiple hypothesis testing involving the search for significant interactions can result in Type I error. Although we restricted our interpretation to interactions that significantly contributed to model fit, the potential for such error cannot be fully discarded. In addition, the NSFH lacks information on several established predictors of retirement, especially pension eligibility and wealth. Although we find that inclusion of pension-related variables in the models does not alter results pertaining to the family variables, better indicators for the potential economic consequences of retirement may be related to and modify effects of family variables on retirement decisions. The intersection of work and family spheres over the life span also may evolve into distinct pre-retirement conditions that can either deter or promote labor force withdrawal. Future research should address these limitations. Furthermore, detailed measures of family economic liabilities, care responsibilities, and fulfillment of contact needs with close relatives are necessary for researchers to adequately assess family influences on retirement processes. It may also be useful for researchers to investigate economic transfers from relatives as possible anticipated sources of retirement income. Especially underprivileged populations still rely to some extent on family strategies to secure old age economic well-being, as was the case prior to the establishment of Social Security (Haber and Gratton 1994).
Despite these caveats, our findings have implications for retirement research and policy. They provide evidence that inclusion of both marital and family characteristics in retirement models can help enhance model specification. It is important to note, however, that addition of family variables has no impact on the effects of other variables typically considered in retirement models. Thus, models excluding marital and family variables are not misspecified but incomplete; that is, additional refinement of retirement models can be achieved through inclusion of marital and family characteristics. Similarly, our findings underline the need for greater acknowledgment of diversity in retirement processes among diverse subpopulations. They indicate that this diversity cannot be fully captured by reference to health and economic limitations among minorities but requires more complex understanding of life course processes and family transactions. Recognition of these complex processes will not only enhance theory development but is also necessary for policy. Retirement policies rely on the predictability of individuals' retirement behaviors. Refined model specification can enhance such predictability and thus the efficiency of measures to secure old age provisions into the future. Because Social Security constitutes the main basis of retirement income for economically underprivileged groups (O'Rand 1996; Pampel 1998), the consequences of inadequate old age provisions will fall most heavily on these groups.
|
Acknowledgments |
|---|
Received for publication November 1, 1999. Accepted for publication August 9, 2000.
|
References |
|---|
|
TOPAbstractTheoretical BackgroundMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. Shacklock, Y. Brunetto, and S. Nelson The different variables that affect older males' and females' intentions to continue working Asia Pacific Journal of Human Resources, April 1, 2009; 47(1): 79 - 101. [Abstract] [PDF]
|
|
|
|
 |
|
 R. W. Johnson Family, Public Policy, and Retirement Decisions: Introduction to the Special Issue Research on Aging, March 1, 2009; 31(2): 139 - 152. [PDF]
|
|
|
|
|
|
J. M. Raymo, J. Liang, E. Kobayashi, Y. Sugihara, and T. Fukaya Work, Health, and Family at Older Ages in Japan Research on Aging, March 1, 2009; 31(2): 180 - 206. [Abstract] [PDF]
|
|
|
|
 |
|
 T. H. Brown and D. F. Warner Divergent Pathways? Racial/Ethnic Differences in Older Women's Labor Force Withdrawal J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2008; 63(3): S122 - S134. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. van Solinge and K. Henkens Involuntary Retirement: The Role of Restrictive Circumstances, Timing, and Social Embeddedness J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2007; 62(5): S295 - S303. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Raymo and M. M. Sweeney Work-family conflict and retirement preferences. J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2006; 61(3): S161 - S169. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. A. Cotter, J. M. Hermsen, and R. Vanneman Gendered Opportunities for Work: Effects on Employment in Later Life Research on Aging, November 1, 2002; 24(6): 600 - 629. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
