Later-life Employment Preferences and Outcomes: The Role of Mid-life Work Experiences

Abstract

In this paper, we evaluate relationships between mid-life work experiences and the realization of preferences for full-time employment, part-time employment, and complete retirement at age 63–64. Using rich data from the Wisconsin Longitudinal Study, we demonstrate that the likelihood of achieving one's preferred employment status is related to earlier work experiences including employment stability in mid-life and self-employment, part-time employment, and private pension coverage across the life course. Despite large gender differences in work experiences across the life course, relationships between earlier work experiences and the likelihood of realizing later-life employment preferences are generally similar for men and women. We also find that these relationships are only partially mediated by economic and employment circumstances in late mid-life, suggesting the need for further evaluation of the cumulative pathways linking mid-life work experiences to the realization of later-life employment preferences.

As the population ages, there is a growing consensus that older Americans will need to work longer in order to ensure their own economic well-being in later life and to limit projected strains on public transfer programs (Munnell and Sass 2008). Despite evidence that the long-term trend toward earlier retirement has slowed or reversed (Friedberg 2007), the mean and modal ages of retirement (63 and 62, respectively) remain below the age of full Social Security eligibility (Burtless and Quinn 2002). At the same time, however, recent attitudinal surveys indicate that a majority of middle-aged Americans plan to work at older ages, with many interested in reduced hours prior to retirement, phased retirement, and part-time work after retirement (AARP 1998, 2004; Merrill Lynch 2006; U.S. General Accounting Office 2001). It is likely that intentions to work beyond retirement have continued to increase in response to the recent economic crisis and associated reductions in pension wealth. The extent to which growing preferences for continued work at older ages will be realized is unclear in light of existing barriers to partial or phased retirement such as pension incentives to completely retire at specific ages and employers' reluctance to hire older workers (Penner, Perun, and Steuerle 2002). Several studies have found that many older Americans find themselves limited to a choice between complete retirement or continued full-time work (Abraham and Houseman 2005; Moen and Roehling 2005).

Although a substantial increase in extended labor force participation seems unlikely based on the experiences of recent cohorts, it is important to recognize that the context of retirement is changing rapidly. Crisp and unidirectional retirements have become less common as “blurred” retirement transitions, characterized by reductions in work commitment, bridge jobs, and reemployment after retirement have become more frequent (Herz 1995; Mutchler, Burr, Pienta, and Massagli 1997; Ruhm 1990). Indeed, the majority of older Americans who retire from a career job now do so in stages rather than in a single well-defined transition out of the labor force (Cahill, Giandrea, and Quinn 2006). Collectively, these changes have been described as a “deinstitutionalization” or “individualization” of retirement (Guillemard and Rein 1993; Han and Moen 1999; Moen and Altobelli 2007) or as a “new retirement paradigm” (Clark and Mitchell 2005).

Growing individual responsibility for planning and navigating the retirement process increases the importance of understanding preferences for continued work at older ages and the factors that facilitate or hinder the realization of those preferences. Existing research tells us very little, however, about what kinds of people are the most and the least likely to successfully realize their later-life employment preferences. An improved understanding of these relationships will inform public and private efforts to promote extended labor force attachment while also providing potentially important insights into the implications of retirement experiences for variation in well-being at older ages. Previous studies finding lower levels of psychological well-being among those who either retired early as a result of involuntary job loss (Gallo, Bradley, Siegel, and Kasl 2000) or felt that they had little control over their retirement (Herzog, House, and Morgan 1991) highlight one potentially important implication of later-life employment preferences and their realization in the rapidly changing context of retirement.

Although research focusing explicitly on the realization of later-life employment preferences is extremely limited, several studies have considered the closely related, but distinct, concept of retirement expectations. Some of these studies have examined the correspondence between expected and actual age of retirement (e.g., Anderson, Burkhauser, and Quinn 1986; Bernheim 1989; Dwyer 2001; Dwyer and Hu 2000), others have described stability and change in individuals' retirement expectations (e.g., Benitez-Silva and Dwyer 2005; Ekerdt, DeViney, and Kosloski 1996), and one examines relationships between expectations of preretirement reduction in work hours or job change and subsequent outcomes (Abraham and Houseman 2005).

Although expectations are of clear importance for projecting future behavior, they may differ from preferences to the extent that they also reflect institutional, structural, and normative constraints. Existing research on retirement expectations is also limited in a number of other important respects. For example, a sizable proportion of men and women in their fifties and early sixties are unable to articulate clear expectations about when or how they will retire. Among employed respondents to the first wave of the Health and Retirement Study (HRS), 12% said they did not know when they would retire and another 13% said they would never retire (Ekerdt, Hackney, Kosloski, and DeViney 2001). “Don't know” was the modal response to questions about expectations regarding preretirement job change or reduction in work hours (Abraham and Houseman 2005). The difficulty of ascertaining expectations in the context of a retirement process characterized by increasing heterogeneity and substantial uncertainty thus precludes examination of correspondence between expectations and outcomes for a substantial minority of respondents. In contrast, nearly all preretirement employees are willing and able to state preferences regarding employment at peak retirement ages (Raymo and Sweeney 2006).

A second, important, limitation of existing research on the realization of retirement expectations is that it has focused primarily on individual characteristics observed in late mid-life while paying little attention to the role of earlier work experiences. Life course scholars have long emphasized the importance of understanding individual outcomes and circumstances at a given age as the product of earlier experiences and exposures accumulated across the life course (Elder 1995). They also recognize that, because this process of cumulation results in increasing variation with age in individual life circumstances (Crystal and Waehrer 1996; O'Rand 1996; O'Rand and Henretta 1999), attention to experiences across the life course is particularly important for understanding variation in later-life outcomes, such as work and retirement.

In this study, we address these limitations by using uniquely rich data from the Wisconsin Longitudinal Study (WLS) to examine the ways in which mid-life work experiences are related to the realization of preferences for continued employment or complete labor force withdrawal at peak retirement ages. As the only survey to repeatedly collect information on multiple aspects of work experiences across the life course of a large sample of men and women who are currently in the midst of the retirement process, the WLS is an ideal source of data for our purposes. In addition to detailed information about all jobs that respondents have held since age 35–36, the WLS contains information on preferences for continued work or retirement at age 63–64 as reported at age 53–54, and detailed information on individual circumstances in late mid-life.

These data allow us to describe correspondence between employment preferences and subsequent outcomes and to examine the extent to which congruence between preferences and outcomes is related to earlier work experiences across the life course. Building upon earlier studies of retirement expectations and outcomes as well as research on career influences on retirement, we address the following questions. Are employment stability and the characteristics of jobs held across the life course associated with the realization of preferences for continued work or retirement at age 63–64? To what extent are these relationships mediated by established correlates of retirement outcomes including financial resources, retirement benefits, health, employment circumstances, and job loss? Do these relationships differ for men and women?

Background

The changing context of retirement

The growing consensus that older Americans will need to work longer than in the recent past reflects not only evidence that many are financially unprepared for early retirement, but also projected growth in the absolute and relative size of the elderly population and related increases in longevity. Efforts to promote later retirement or extended labor force participation are seen as the most obvious, and perhaps most feasible, solution to projected labor force shortages, loss of skilled workers, difficulties in financing pay-as-you go public pension and health care systems, and ensuring the financial well-being of older Americans (Burtless and Quinn 2002; Munnell and Sass 2008). Early retirement remains common, but public and private policy efforts have sought to encourage and facilitate extended labor force attachment at older ages (e.g., Burtless and Quinn 2001; Morton, Foster, and Sedlar 2005) and recent surveys indicate that increasing proportions of Americans approaching retirement plan to continue working beyond current peak ages of retirement (AARP 1998, 2004; Merrill Lynch 2006). Any effort to understand the likelihood that these plans will be realized should be based on an understanding of the ways in which the context of retirement in the U.S. is being reshaped.

Beginning in the 1950s, a combination of institutional, economic, and social forces contributed to the emergence of a highly institutionalized retirement process for American men (Atchley 1982; Moen and Roehling 2005). The mean age of retirement for men declined steadily from 70 in 1950 to 63 after 1985 (Burtless and Quinn 2002), with large peaks at 62 and 65, the ages for early and full Social Security eligibility. This trend toward earlier retirement, and concurrent declines in rates of labor force participation (Costa 1998), reflected mandatory retirement policies, increases in pension wealth, and strong incentives to retire at particular ages built into both private defined benefit pension plans and Social Security (Quadagno and Quinn 1997). During this period, clear societal and workplace norms about appropriate retirement ages emerged (Ekerdt 1998), and retirement became a widely anticipated, normative stage of the life course (Ekerdt, Kosloski, and DeViney 2000; Moen and Altobelli 2007).

This description of retirement has become progressively less accurate over the past 20 years. The elimination of mandatory retirement from most occupations, rapid growth in defined contribution pension plans, associated weakening of institutional incentives to retire at relatively young ages (Beller and Lawrence 1992; Munnell and Perun 2006), increases in self-employment (Karoly and Zissimopolous 2003), increases in involuntary job loss (Couch 1998), and the increasing tendency for spouses to coordinate their retirements (Blau 1998) have all contributed to a decline in crisp, unidirectional retirements concentrated at specific ages. Retirement is increasingly viewed as a process rather than an event (Moen 2004) and, for a large proportion of older Americans, this process involves partial retirement characterized by reductions in work hours or commitment, often in “bridge jobs” with a different employer or in a different industry. For many, retirement from career jobs is followed by “second acts” (Moen and Roehling 2005) in jobs with more flexible schedules and lesser responsibility, often in self-employment (Haider and Loughran 2001), or perhaps in volunteer work (Kim and Feldman 2000).

As established norms and guidelines for planning and navigating the retirement process become less relevant (Guillemard and Rein 1993) and as more workers retire “under terms of their own choice and at their own pace” (Han and Moen 1999:196), it is perhaps not surprising that men and women now approaching retirement age appear to have very different attitudes about retirement than earlier cohorts. Widespread plans to remain in the labor force beyond current normative retirement ages likely reflect inadequate financial preparation for retirement (Moore and Mitchell 2000; Munnell, Golub-Sass, and Webb 2007) and an increasing desire to remain physically active and productive at older ages (Pitt-Catasouphes and Smyer 2005). To the extent that the realization of retirement preferences is associated with subsequent financial and psychological well-being (e.g., Herzog, House, and Morgan 1991), these attitudinal shifts suggest that success or failure in realizing preferences for continued employment may be an increasingly important dimension of variation in well-being among older Americans in years to come. This may be particularly true in light of the recent economic downturn.

Research on the realization of retirement expectations

We are not aware of any existing research on individual characteristics associated with the realization of preferences for employment and retirement, but several earlier studies have examined correspondence between retirement expectations and outcomes based on the Retirement History Study (Anderson, Burkhauser, and Quinn 1986), the Normative Aging Study (Ekerdt, Vinick, and Bosse 1989), and the Health and Retirement Study (Dwyer and Hu 2000). These studies have typically found that approximately 60% of respondents retired when expected and those who did not were likely to retire earlier than they had expected. In one of the few studies to consider the realization of expectations for outcomes other than retirement timing, Abraham and Houseman (2005) found that expectations of reduced hours prior to retirement were much less likely to be realized than were expectations of complete retirement or continued full-time work. These earlier studies have also demonstrated that the likelihood of realizing retirement expectations is positively associated with financial resources, pension benefits, health insurance coverage, and good health, but negatively associated with unforeseen events such as policy changes, macroeconomic fluctuations, changes in health, and family responsibilities.

Despite a substantial body of research linking careers to the timing and nature of retirement (discussed below), existing research on the realization of retirement expectations has paid little attention to the role of earlier work experiences. This is an important limitation in that the temporally proximate correlates of congruence between retirement expectations and outcomes are not assigned exogenously in late mid-life but are the product of experiences cumulated across the life course (Elder and Pavalko 1993; Han and Moen 1999; Hayward, Friedman, and Chen 1998; O'Rand 1996). This process of “cumulative stratification” (O'Rand 1996; O'Rand and Henretta 1999) may be particularly important for cohorts that are now approaching an increasingly individualized retirement process with increasingly heterogeneous work histories (AARP 2004; Han and Moen 1999; Moen and Roehling 2005). Heterogeneity and inequality in work experience has increased at all stages of the life course, including the transitions from school to work and to stable employment (e.g., Kerckhoff 2002), but work experiences across mid-life (which we define roughly as ages 35–55) are of particular relevance for studies of retirement (Moen 2004; Moen and Roehling 2005). For insights into the ways in which work experiences across the life course may contribute to variation in the realization of later-life employment preferences, we draw upon studies of career influences on retirement outcomes.

Research on mid-life work experiences and retirement

Previous research has placed particular emphasis on the ways in which the stability of employment and employment circumstances across the life course influence retirement via established temporally proximate antecedents such as economic resources and benefits, health, and employment circumstances. This work provides insights into potential linkages between work experiences across the life course and the realization of later-life employment preferences. For example, retirement outcomes for those who experienced stable careers in higher status occupations differ from those who had less stable careers characterized by intermittent employment, self employment, or employment in lower status occupations (Han and Moen 1999; Hayward, Friedman, and Chen 1998; O'Rand and Henretta 1982). Based on the results of these studies, we expect that earlier work experiences associated with either later retirement or post-retirement employment should also be associated with the realization of preferences for continued employment. Similarly, work experiences across the life course that have been linked to earlier retirement should be associated with the realization of preferences for full retirement. To the extent that spousal influences on employment preferences and outcomes are stronger for women than for men (Moen, Huang, Plassman, and Dentinger 2006; Szinovacz and Deviney 2000), we expect that work experiences across the life course may be less salient for understanding variation in the realization of women's later life employment preferences.

Employment stability

While many follow an orderly career, progressing through the ranks of a single employer, work histories characterized by multiple transitions in and out of the labor force and jobs with many different employers are also common (Han and Moen 1999; Moen and Roehling 2005; Moen and Spencer 2006). For women, career interruptions during the childbearing years are associated with later age at retirement (O'Rand, Henretta, and Krecker 1992; Pienta, Burr, and Mutchler 1994), reflecting lower levels of accumulated pension wealth (Mehdizadeh and Luzadis 1994; Yabiku 2000) and desire for continued involvement in late-starting careers (Moen and Roehling 2005). For men, spells of unemployment across the life course are associated with earlier retirement (Hayward, Friedman, and Chen 1998), reflecting higher rates of disability as well as weaker attachment to work and a lower likelihood of engagement in rewarding work (Hayward, Friedman, and Chen 1998). These findings suggest that career stability should facilitate the realization of preferences for continued work among men and preferences for complete retirement among women. Those who have experienced stable careers may also be less likely to experience unforeseen events such as health decline, job loss, and changes in marital status that hinder one's ability to realize preferences for continued employment or retirement (Dwyer 2001; Williamson and McNamara 2002). We expect that orderly careers in occupational settings with relatively clear and standardized retirement provisions or norms, such as large companies, public bureaucracies, and unionized jobs, may also contribute to a higher likelihood of realizing preferences for complete retirement.

Employment circumstances

Relationships between retirement timing and employment circumstances are well-established. Of particular note is relatively late retirement among the self-employed and relatively early retirement among those employed in the public sector or in large, bureaucratic firms (Elder and Pavalko 1993; Han and Moen 1999; Hayward, Friedman, and Chen 1998). Other studies have shown that the likelihood of continued employment at older ages is relatively high among men who had careers in higher status occupations characterized by jobs with greater substantive complexity, whereas retirement tends to be earlier for those who had careers in lower status occupations characterized by jobs with low autonomy and repetitive, dangerous, physically demanding work (Hayward 1986; Hayward, Friedman, and Chen 1998; Hayward, Grady, Hardy, and Sommers 1989). These patterns presumably reflect differences in job flexibility, job rewards, economic resources, and institutional incentives to retire at specific ages.

Based on this earlier work, we expect that experience with self-employment across the life course will be associated with preferences for continued employment and a higher likelihood of realizing those preferences. At the same time, more limited financial resources, including pension wealth, may limit the likelihood of realizing preferences for complete retirement among those with greater career involvement in self-employment or in part-time employment. In contrast, we expect that public sector employment across the life course, labor union membership, and employment in jobs providing more extensive benefits will be associated with preferences for full retirement and a higher likelihood of realizing those preferences. Similar to those with stable, orderly careers, we expect that pension accumulation, better health, and exposure to well-established normative and institutional expectations regarding retirement will facilitate the realization of preferences for full retirement among these groups.

Achievement of earlier occupational aspirations

Although not often addressed in the literature on careers and retirement, occupational aspirations earlier in life and the extent to which those aspirations were achieved may have significant implications for the way in which older Americans approach retirement. Research using data from the Wisconsin Longitudinal Study has shown that failure to realize occupational aspirations expressed in early mid-life (age 35–36) is associated with lower levels of psychological well-being in late mid-life (age 53–54) (Carr 1997). Interestingly, this work also showed that psychological well-being in late mid-life is particularly low among those who did not express any specific occupational aspirations and among women with aspirations to be a homemaker.

Based on these findings, we posit two reasons to expect that failure to realize earlier occupational aspirations (or the absence of expressed aspirations) may be associated with a lower likelihood of realizing later-life employment preferences. First, it may be that those who failed to realize earlier aspirations have a lower propensity to plan and are thus less likely to plan for their retirement. This scenario rests on the assumption that retirement planning is positively associated with the likelihood of realizing later-life employment preferences. Second, it is possible that failure to realize earlier occupational aspirations influences the realization of employment preferences indirectly through its influence on psychological well-being and stress in late mid-life. The assumption in this case is that psychological well-being is positively related to the likelihood of realizing later-life employment preferences.

A central finding of previous research is that relationships between work experiences across the life course and retirement outcomes are largely mediated by individual characteristics in late mid-life (e.g., Hayward 1986; Hayward, Friedman, and Chen 1998; Hayward et al. 1989; O'Rand and Henretta 1982; O'Rand and Landerman 1984). Because these characteristics include several well-established correlates of retirement expectations and their realization, we posit that relationships between earlier work experiences and the realization of later-life employment preferences will be largely mediated by economic circumstances and benefits, health, employment circumstances, and unexpected events in late mid-life. Because these posited mediators have also been the focus of recent research on theories of cumulative advantage and cumulative disadvantage (e.g., Crystal and Waehrer 1996; Dannefer 2003; Mehdizadeh and Luzadis 1994; O'Rand 1996), evidence consistent with this scenario would provide important insights into the ways in which processes of cumulative stratification indirectly influence retirement outcomes. Evidence that linkages between mid-life work experiences and the realization of later-life employment preferences are not mediated by these characteristics would suggest cumulative work experiences may have a direct influence on retirement outcomes. We also recognize that increasing individualization of both the life course and the retirement process suggests that earlier work experiences may be related to the realization of later-life employment preferences in ways that have yet to be adequately theorized. For example, it may be that both the nature of work histories and the likelihood of realizing later-life employment preferences are associated with characteristics such as planfulness or competence (e.g., Ameriks, Caplin, and Leahy 2003; Lusardi and Mitchell 2007) that are typically not measured in surveys.

Data and methods

We examine relationships between mid-life work experiences, employment preferences, and subsequent employment status using data from the large sample of men and women in the Wisconsin Longitudinal Study (WLS). The WLS is a long-term study of a random sample of 10,317 men and women who graduated from Wisconsin high schools in 1957. In addition to the original 1957 survey and a brief mail survey of their parents in 1964, the graduates were interviewed in 1975, 1993, and 2004. We use data from the surveys conducted in 1993 and 2004, both of which included a telephone interview (response rates of 87% and 85%, respectively) and a mail questionnaire (response rates of 70% and 75%, respectively). Our base sample consists of the 5,465 surviving members of the original sample who completed both components of the 1993 and 2004 surveys.¹ Because we are interested in examining realization of preferences for employment or retirement at age 63–64, we exclude participants who considered themselves fully or partially retired (n=632) at the time of the 1993 survey when they were 53–54 years old. After excluding those who did not respond to questions about future employment preferences in the 1993 survey (n=36) or occupational history in the 2004 survey (n=111) and those with missing data on any of the other variables used in the analysis (n=282), we are left with an analytical sample of 4,404.

Several features of the WLS make these data particularly well-suited to our purposes. First, the WLS is one of the only surveys to ascertain preferences (in addition to expectations) for retirement or continued employment at prime retirement ages. Second, the WLS graduate cohort was born in the late 1930s, making their later-life employment experiences an important comparative reference with which to evaluate the experiences of the slightly younger, and much larger, baby boom cohorts who are now approaching retirement. Third, as the longest-running longitudinal study of older Americans, the WLS provides rich data on work experiences across the life course that will not be available in other surveys for many years to come. Because the WLS has collected detailed information on most jobs that participants have held, we are able to measure multiple aspects of individual work histories in greater detail than in previous studies of career influences on retirement.

Measures

Our measure of employment preferences comes from a question in the 1993 survey about preferred employment status ten years in the future. The precise wording of the question is “If you were free to choose, what would you like to be doing 10 years from now, in terms of your work? Would you like to be working full-time, working part-time, not working, retired, or something else?” Because nearly all respondents were 53–54 years old at the time of the survey, this measure reflects desired work status just beyond the modal age of retirement. As shown in the last two columns of Table 1, the majority of respondents preferred to be fully retired (not working) in ten years (58% of men and 64% of women) and about one-fourth of both men and women said that they would like to be working part-time. Preferences for full-time work were the least common, but were more prevalent among men (19%) than women (10%).

Table 1.

Distribution of work status ten years after 1993 survey, by employment preferences in 1993 and sex

MEN
	Observed work status ten years after 1993 survey
Employment Preferences	Full-time work	Part-time work	Not Working	Total Pct.	N
Full-time work (row pct.)	11 (56)	3 (18)	5 (26)	19 (100)	389
Part-time work (row pct.)	10 (42)	7 (28)	7 (30)	23 (100)	480
Not Working (row pct.)	12 (21)	12 (20)	34 (59)	58 (100)	1,208
Total pct.	32	22	46	100	2,077
N	671	448	958	2,077

WOMEN
	Observed work status ten years after 1993 survey
Employment Preferences	Full-time work	Part-time work	Not Working	Total Pct.	N
Full-time work (row pct.)	5 (45)	3 (25)	3 (30)	10 (100)	243
Part-time work (row pct.)	8 (29)	8 (31)	10 (40)	26 (100)	604
Not Working (row pct.)	11 (17)	12 (18)	41 (65)	64 (100)	1,480
Total pct.	23	22	55	100	2,327
N	534	516	1,277	2,327

We determined each respondent's labor force status 120 months after the 1993 interview date using data from the employment history module in the 2004 telephone survey. In this series of questions, respondents provided information on job characteristics, including the month and year in which they began and ended all employment spells subsequent to the 1993 survey. The main diagonal of Table 1 indicates that just over half of the sample was in their preferred work status ten years after the 1993 survey (52% for men and 54% for women). Row percentages (in italics) show that realization of preferences was most likely among those who preferred to work full-time (56% of men and 45% of women) or not at all (59% of men and 65% of women). Consistent with evidence that work arrangements are less flexible than older workers would like (Abraham and Houseman 2005; Gustman and Steinmeier 2004), those preferring part-time work were far less likely to realize that preference (28% for men and 31% for women).

We construct eight measures of mid-life work experiences based on the employment history data collected at age 53–54 as well as employment information collected in the 1975 survey when respondents were 35 years old. Similar to the employment histories collected in 2004, the 1993 survey collected detailed information about spells of employment between ages 35–36 and 53–54. These data provide information about the number and sequencing of jobs that graduates held during mid-life, information sufficient to determine the industry, occupation, and class of worker for the job they held when they began working for that employer, and information on whether they worked full- or part-time and whether they had health insurance and private pension coverage.

We measure employment stability using an indicator of discontinuous employment between ages 35–36 and 53–54 that distinguishes respondents who were employed for the entire 18-year inter-survey period from those who were out of the labor force for one or more years. A second indicator of stability indicates whether respondents ever left a job involuntarily between ages 35–36 and 53–54. This measure is constructed based on open-ended responses to questions asking the reasons for leaving each employer spell between 1975 and 1993 (see Brand 2006).

We measure mid-career occupational status as the mean value of occupational earnings for jobs held between ages 35–36 and 53–54. For a given combination of occupation, industry, and class-of-worker, occupational earnings refers to the percentage of persons in each occupation in the 1990 Census who earned at least $14.30 per hour in 1989 (Hauser and Warren 1997). To distinguish those with and without exposure to well-defined structural and normative incentives to retire, we create measures of the proportion of years between ages 35–36 and 53–54 that the respondent was (a) self-employed and (b) working for an employer that provided private pension coverage. For women, we also construct a corresponding measure of the proportion of years spent in part-time employment. We do not include a parallel measure for men as very few were employed in part-time jobs during mid-life. A final measure of employment circumstances associated with structural and normative incentives to retire is a dichotomous indicator of union membership. Because the occupational history questions did not ask about union membership, this variable indicates whether respondents belonged to a labor union at the time of either the 1993 survey or the 1975 survey.

Finally, we construct a measure of respondents' realization of career aspirations by calculating the difference between current occupational status and the status of the occupation to which respondents aspired at age 35–36. We construct this measure using occupational education – the percentage of persons in each occupation in the 1990 Census who completed one or more years of college (Hauser and Warren 1997). Because many respondents said they did not know what they wanted to do or that they wanted to be a homemaker, and because these responses are associated with lower levels of emotional well-being at age 53–54, we follow Carr (1997) in constructing a multiple category measure. The four categories are failed to achieve aspirations, met or exceeded aspirations, aspired to be a homemaker or did not indicate aspirations, and a residual category for those who were not asked this question because they did not participate in the 1975 survey.

The distributions of these work history measures and all other variables included in the models below are presented separately by sex in the first two columns of Table 2. These figures show that mid-life employment was more stable for men than for women. The large majority of men (88%) were continuously employed between ages 35–36 and 53–54 but nearly half of women (48%) were out of the labor force for at least one year. On average, women spent about one-third of mid-life in part-time employment, had substantially lower average levels of occupational status than men, and spent less time in jobs that provided private pension coverage. Other measures of mid-life work experience differed less by gender. Roughly one out of six men (17%) and women (16%) reported involuntary job loss between the two surveys. On average, respondents were self-employed for a relatively small proportion of mid-life (14% for men and 9% for women) and one-third of men and one-fifth of women reported labor union membership. About one-half of men (53%) and women (44%) met or exceeded their occupational aspirations expressed at age 35–36.

Table 2.

Descriptive statistics, by sex and employment preferences

	Total			By employment preferences
	Men	Women	Men			Women
	Pct., Mean	Pct., Mean	Full-time	Part-time	Not working	Full-time	Part-time	Not working
Mid-life work experiences (ages 35–53)
Discontinous employment
Yes	0.12	0.48	0.120	0.12	0.09	0.53	0.51	0.53
No	0.88	0.52	0.880	0.88	0.91	0.47	0.49	0.47
Involuntary job exit
Yes	0.17	0.16	0.22	0.19	0.15	0.21	0.17	0.15
No	0.83	0.84	0.78	0.81	0.85	0.79	0.83	0.85
Mean occupational earnings	40.48	19.39	39.81	40.86	40.55	21.76	20.80	18.43
Proportion years self-employed	0.14	0.09	0.19	0.25	0.08	0.12	0.13	0.07
Proportion years covered by private pension	0.78	0.53	0.71	0.68	0.83	0.56	0.50	0.54
Proportion years in part-time work	-	0.29	-	-	-	0.19	0.33	0.30
Union membership
Yes	0.33	0.22	0.19	0.25	0.41	0.20	0.20	0.23
No	0.67	0.78	0.81	0.75	0.59	0.80	0.80	0.77
Occupational aspirations and outcomes
Did not achieve aspirations	0.31	0.29	0.28	0.31	0.32	0.28	0.28	0.30
Achieved aspirations	0.53	0.44	0.61	0.53	0.51	0.51	0.48	0.41
Did not state aspirations/homemaker	0.09	0.20	0.06	0.07	0.11	0.15	0.17	0.22
Not interviewed in 1975	0.07	0.07	0.05	0.09	0.07	0.06	0.08	0.07
Background characteristics
Age	53.25	53.11	53.105	53.2	53.29	53.09	53.10	53.12
Marital Status
Married, spouse no pension	0.47	0.20	0.47	0.48	0.47	0.18	0.22	0.20
Married, spouse has pension	0.41	0.60	0.40	0.43	0.42	0.42	0.57	0.64
Not married	0.11	0.20	0.13	0.09	0.11	0.41	0.21	0.16
Educational Attainment
High school	0.48	0.60	0.32	0.45	0.55	0.46	0.53	0.65
Some college	0.16	0.15	0.18	0.14	0.16	0.16	0.17	0.15
BA+	0.36	0.25	0.50	0.41	0.29	0.38	0.30	0.20
Economic circumstances and benefits
Wages ($/hour)	23.74	11.22	24.03	28.76	21.66	11.81	11.75	10.90
Net worth (log $1,000)	12.01	11.50	11.70	12.32	11.99	10.77	11.52	11.62
Covered by health insurance
Yes	0.84	0.54	0.78	0.78	0.89	0.65	0.52	0.54
No	0.16	0.46	0.22	0.22	0.11	0.35	0.48	0.46
Pension eligibility 10 years after 1993 survey
Yes	0.73	0.49	0.56	0.65	0.81	0.47	0.46	0.51
No	0.27	0.51	0.44	0.35	0.19	0.53	0.54	0.49
Estimated value of monthly pension benefits (log)	6.55	4.73	6.18	6.14	6.83	4.76	4.93	4.65
Missing monthly pension benefits
Yes	0.16	0.18	0.18	0.19	0.13	0.18	0.20	0.18
No	0.84	0.82	0.82	0.81	0.87	0.82	0.80	0.82
Health
Self-rated health
Fair/poor/very poor	0.10	0.10	0.06	0.09	0.11	0.07	0.07	0.11
Good/excellent	0.90	0.91	0.94	0.91	0.89	0.93	0.93	0.89
CES-D (log)	2.36	2.572	2.23	2.37	2.40	2.40	2.57	2.51
Illnesses in midlife
Yes	0.47	0.48	0.43	0.48	0.48	0.41	0.42	0.51
No	0.53	0.52	0.57	0.52	0.52	0.59	0.58	0.49
Serious illness after age 53 – 54
Yes	0.29	0.23	0.27	0.26	0.31	0.23	0.20	0.24
No	0.71	0.77	0.73	0.74	0.69	0.77	0.80	0.76
Employment circumstances
Occupational sector
Private	0.64	0.68	0.61	0.54	0.68	0.63	0.65	0.69
Public	0.21	0.24	0.18	0.18	0.23	0.25	0.23	0.25
Self-employed	0.15	0.08	0.20	0.28	0.08	0.12	0.12	0.06
Weekly work hours	47.33	38.21	49.64	48.69	46.04	43.11	37.09	37.78
0 hours (Women)	-	0.16	-	-	-	0.06	0.11	0.20
0–34 hours (Men) 1–34 hours (Women)	0.04	0.19	0.06	0.04	0.03	0.08	0.25	0.18
35–49 hours	0.57	0.53	0.42	0.49	0.65	0.61	0.50	0.53
50+ hours	0.39	0.12	0.51	0.46	0.32	0.25	0.14	0.09
Very satisfied with job
Yes	0.51	0.48	0.61	0.52	0.47	0.61	0.51	0.44
No	0.49	0.52	0.39	0.48	0.53	0.39	0.49	0.56
Perceived liklihood of losing job
No chance	0.45	0.58	0.42	0.48	0.44	0.58	0.54	0.60
1–50%	0.37	0.27	0.41	0.36	0.37	0.32	0.30	0.26
Greater than 50%	0.18	0.14	0.17	0.16	0.19	0.10	0.16	0.14
Flexible hours
Yes	0.59	0.36	0.68	0.68	0.53	0.51	0.42	0.32
No	0.41	0.64	0.32	0.32	0.47	0.49	0.58	0.68
Job requires physical effort
Yes	0.11	0.13	0.09	0.10	0.12	0.16	0.12	0.13
No	0.89	0.87	0.91	0.90	0.88	0.84	0.88	0.87
Exposed to dangerous conditions at work
Yes	0.42	0.22	0.31	0.43	0.45	0.23	0.24	0.21
No	0.58	0.78	0.69	0.57	0.55	0.77	0.76	0.79
Involuntary job exit after age 53 – 54
Yes	0.16	0.14	0.19	0.18	0.14	0.14	0.15	0.13
No	0.84	0.86	0.81	0.82	0.86	0.86	0.85	0.87
Unexpected life events
Lost spouse after age 53 – 54
Yes	0.06	0.110	0.08	0.05	0.06	0.11	0.09	0.11
No	0.94	0.940	0.92	0.95	0.94	0.89	0.91	0.89
N	2,077	2,327	389	480	1,208	243	604	1,480

Table 2 also presents tabulations of all variables by retirement preferences articulated at age 53–54. These figures indicate that men and women who preferred full- or part-time employment were somewhat more likely to have had discontinuous work histories or experience of involuntary job loss. Preference for full- or part-time employment is also positively associated with time spent in self-employment and, among men, is negatively related to union membership and time covered by private pension plans. Those who expressed a preference for part-time work were characterized by relatively high levels of mid-life self-employment (for men) and part-time employment (for women). Those preferring not to work were more likely to have belonged to a labor union and either not reported occupational aspirations or aspired be a homemaker when they were 35–36 years old.

As described above, one of our primary objectives is to assess the extent to which relationships between mid-life work experiences and the realization of later-life employment preferences are mediated by temporally proximate characteristics identified in research on the realization of retirement expectations and studies of career influences on retirement outcomes. To this end, we construct a wide array of established correlates of both employment/retirement preferences (see Raymo and Sweeney 2006) and retirement outcomes including (a) age, educational attainment, marital status, loss of spouse, (b) economic circumstances and benefits, (c) health, and (d) employment circumstances. With a few exceptions noted below, these characteristics are all measured at the time of the 1993 survey when respondents were 53–54 years old.

Economic circumstances include hourly wages and the logged value of net worth measured in increments of $1,000. Benefits include dichotomous indicators of health insurance coverage at age 53–54 and eligibility for private pension benefits 10 years after the 1993 survey. We also construct an estimated value of private pension benefits by adding the monthly benefits that respondents reported receiving at the time of the 2004 survey to the monthly benefits that they reported being eligible for but not yet receiving. We use the natural log of this value and replace missing values with gender-specific means and a dichotomous indicator of missingness.

Health includes measures of both physical and emotional well-being. Self-rated health is a two-category measure (excellent/good and fair/poor/very poor) and emotional health is the natural log of the CES-D score. The CES-D is the sum of responses to 20 questions asking respondents how many days during the past week they felt a particular way (e.g., lonely, sad, depressed) and ranges in value from 0–140. A third dichotomous measure of health identifies respondents who had been diagnosed with at least one of the following chronic health conditions prior to the 1993 survey: asthma, bronchitis, sinus problems, circulation problems, fibromyalgia, high cholesterol, kidney or bladder problems, osteoporosis, back problems, ulcers, or prostate problems. We also use respondents' reports of serious health conditions and the timing of their onset collected in 2004 to define an indicator of whether the respondent had a stroke or heart attack or was diagnosed with cancer, diabetes, or high blood sugar subsequent to reporting future employment preferences at age 53–54.

Employment circumstances include seven indicators of the jobs respondents held when they stated later-life employment preferences at age 53–54 and an indicator of involuntary job loss after age 53–54. Occupational sector distinguishes those in private sector employment from those working in the public sector and those who were self-employed. Weekly hours of work is a categorical measure ranging from distinguishing those working less than full-time (0–34 hours per week), full-time (35–49 hours per week), and more than full-time (50+ hours per week). Because a significant proportion of women (16%) were not working in 1993, we include a separate category 0 work hours per week. Job satisfaction is a dichotomous measure distinguishing those who report being very satisfied with their job from those who report lower levels of satisfaction. A measure of job security is based on responses to a question that asked respondents “on a scale from zero to ten, what chance do you think there is that you will lose your job completely in the next two years?” Based on the results of preliminary analyses, we collapse responses into three categories, 0 (no chance), 1–5, and 6–10. Flexible work hours is a dichotomous indicator distinguishing respondents who can decide what time to come to work and leave from those who cannot. Physically demanding work is equal to one for respondents who reported that their jobs always require a lot of physical effort and zero otherwise. Exposure to dangerous conditions is also a 0–1 measure distinguishing those exposed to dangerous conditions at work from those who are not. Using occupational history data collected in 2004, we construct a dichotomous indicator of whether the respondent left a job involuntarily due to layoff, downsizing, or company closing. This is similar to the measure of employment stability described above but refers to the period subsequent to reporting of employment preferences.

Methods

We examine relationships between mid-life work experiences and the realization of later-life employment preferences by estimating multinomial logistic regression models for employment status ten years after the 1993 survey. The decision to use multinomial models rather than binary models for a dichotomous indicator of whether or not preferences were realized is motivated by our interest in describing and understanding how preferences differ from outcomes. For example, evidence that preferences for part-time employment at older ages have increased while opportunities for this kind of work remain limited highlights the importance of understanding the ways in which mid-life work experiences are associated with the relative likelihood of both full-time employment and not working among those who preferred to work part-time. Because the distribution of employment outcomes differs markedly by preferences (Table 1), we estimate models separately by employment preferences. For each model, we estimate the log-odds of the preferred outcome relative to the log-odds of each of the other two outcomes. To facilitate interpretation, we present results as odds ratios with values greater (less) than 1.0 indicating a higher (lower) likelihood of realizing preferences.

We estimate models separately for men and women given substantial gender differences in work experiences across the life course (as shown in Table 2) and evidence that fully-interactive models fit the data better than models based on pooled data in some cases (results not shown). For each of the six combinations of employment preferences and gender, we estimate two models. The first model estimates average associations between mid-life work experiences and employment status 10 years after the 1993 survey, net of age, educational attainment, and marital status in 1993. In Model 2, we control for the posited mediators described above: economic circumstances and benefits; health; employment circumstances; and loss of spouse. By observing change across models in the magnitude and statistical significance of estimated coefficients for the work history measures, we assess the extent to which relationships between mid-life work experiences and the realization of later-life employment preferences are mediated by the individual characteristics emphasized in research on the realization of retirement expectations and in studies of career influences on the timing and pathways to retirement. Examining changes across intermediate, stepwise models in which late mid-life characteristics are entered sequentially (results not shown) allows us to identify which of the temporally proximate characteristics mediate relationships between earlier work experiences and the realization of later-life employment preferences.

Results

Tables 3, 4, and 5 present the findings for models of realization of employment preferences for each of the six combinations of preferences and gender. Because sample size is small for many of these models, we pay attention to all relationships that are statistically significant at p <.10. Hausman tests conducted after each model give no reason to believe that the assumption of independence of irrelevant alternatives is violated.

Table 3.

Odds ratios from multinomial logistic regression models for employment status among those who preferred to work full-time at age 63–64

	Men				Women
	Model 1		Model 2		Model 1		Model 2
Variable	PT	NW	PT	NW	PT	NW	PT	NW
Mid-life work experiences (ages 35–53)
Discontinous employment (ref=no)	1.41	1.06	1.68	1.23	1.34	0.75	1.22	0.88
Involuntary job exit (ref=no)	1.04	1.23	1.54	1.44	2.17 #	1.94	2.07	1.72
Mean occupational earnings	1.00	1.00	1.00	1.00	1.00	0.99	1.00	0.99
Proportion years self-employed	2.81 *	1.76	5.18 *	2.04	0.67	0.93	2.49	0.46
Proportion years covered by private pension	1.40	0.39 *	1.19	0.34 *	1.24	1.12	0.75	0.69
Proportion years in part-time employment	-	-	-	-	0.23 **	0.44	0.25 #	0.41
Union membership (ref=no)	0.92	0.61	1.04	0.97	1.44	1.12	1.60	0.89
Occupational aspirations and outcomes (ref=achieved aspirations)
Did not achieve aspirations	1.57	1.08	1.88	1.19	0.93	0.89	0.95	1.19
Did not state aspirations/homemaker	1.14	1.46	1.06	1.13	0.59	0.45	0.50	0.34 #
Not interviewed in 1975	0.94	3.29	0.90	2.96	1.33	0.36	2.44	0.51
Background characteristics
Age	0.92	1.15	0.96	1.25	1.03	0.68	1.18	0.64
Marital Status (ref=married, spouse no pension)
Married, spouse has pension	0.79	0.72	0.81	0.69	1.43	0.73	2.42	0.81
Not married	0.94	0.65	1.42	0.76	1.06	1.64	1.79	2.46
Educational Attainment (ref=high school)
Some college	1.03	0.99	1.08	1.24	0.79	1.02	0.52	0.96
BA+	1.94 #	1.46	1.95	1.94 #	1.00	0.83	0.56	0.76
Economic circumstances and benefits
Wages ($/hour)			1.00	1.00			1.00	1.03
Net worth (log $1,000)			1.10	0.89			0.99	0.91
Covered by health insurance (ref=no)			1.72	1.96			1.47	1.20
Pension eligibility 10 years after 1993 survey (ref=no)			0.70	0.88			0.30 *	0.90
Estimated value of monthly pension benefits (log)			1.04	0.97			1.10	1.05
Missing monthly pension benefits (ref=no)			2.59 #	0.75			1.90	1.82
Health
Self-rated health Fair/poor/very poor (ref=good/excellent)			0.93	0.92			0.37	0.39
CES-D (log)			0.79	0.74 #			0.94	0.93
Illnesses in midlife (ref=no)			0.79	0.84			0.94	0.69
Serious illness after age 53 – 54 (ref=no)			0.86	0.61			0.54	0.58
Employment circumstances
Occupational sector (ref=private)
Public			0.59	0.53 #			0.90	0.63
Self-employed			0.38	1.17			0.11 *	1.34
Weekly work hours (ref=35 – 49)
0 (Women), 0–34 (Men)			0.29 #	0.45			0.37	0.16 #
1–34 (Women)			-	-			0.24 #	0.70
50+			2.04 *	1.39			1.17	1.32
Very satisfied with job			1.09	0.76			1.56	0.86
Perceived liklihood of losing job (ref=no chance)
50% or less			0.74	1.58			1.21	0.93
Greater than 50%			0.72	1.18			1.85	1.65
Flexible hours (ref=no)			0.94	1.82 #			1.07	0.83
Job requires physical effort (ref=no)			0.81	0.28 *			1.09	2.35
Exposed to dangerous conditions (ref=no)at work			0.89	1.54			0.61	0.92
Involuntary job exit after age 53 – 54 (ref=no)			0.56	0.64			0.58	0.36 #
Unexpected life events
Lost spouse after age 53 – 54 (ref=no)			1.00	1.09			2.68	5.14 *
N	389		389		243		243
df	28		72		30		76
log-likelihood	−364		−330		−241		−213
p value for LR test comparing Model 2 to Model 1			0.01				0.15
Note: PT refers to part-time employment and NW refers to not working

Table 4.

Odds ratios from multinomial logistic regression models for employment status among those who preferred to work part-time at age 63–64

	Men				Women
	Model 1		Model 2		Model 1		Model 2
Variable	FT	NW	FT	NW	FT	NW	FT	NW
Mid-life work experiences (ages 35–53)
Discontinous employment (ref=no)	0.81	0.47 #	0.65	0.42 #	0.92	0.52 **	0.86	0.52 **
Involuntary job exit (ref=no)	0.50 *	0.81	0.38 **	0.57	0.80	1.10	0.81	1.18
Mean occupational earnings	1.02 *	1.00	1.01 #	1.00	1.01 #	1.00	1.02 **	1.00
Proportion years self-employed	0.76	2.67 *	0.55	1.17	1.76	1.66	0.37	0.66
Proportion years covered by private pension	1.07	0.70	1.24	0.59	0.78	0.66	0.95	0.52 *
Proportion years in part-time employment	-	-	-	-	2.34 **	2.12 **	0.90	1.29
Union membership (ref=no)	1.24	0.83	1.12	1.01	1.01	0.73	1.13	0.90
Occupational aspirations and outcomes (ref=achieved aspiration
Did not achieve aspirations	1.26	0.74	1.42	0.81	1.56	1.36	1.27	1.39
Did not state aspirations/homemaker	0.81	0.84	0.84	0.98	2.25 *	1.30	1.63	1.30
Not interviewed in 1975	0.63	0.33 *	0.77	0.36 #	2.93 *	2.37 *	2.99 *	3.86 **
Background characteristics
Age	1.13	1.11	1.13	1.18	0.87	0.78	0.90	0.71 #
Marital Status (ref=married, spouse no pension)
Married, spouse has pension	1.30	0.71	1.42	0.86	1.44	0.49 *	1.24	0.39 **
Not married	0.83	0.42 #	0.66	0.40 #	0.47 *	0.49 *	0.47 *	0.41 *
Educational Attainment (ref=high school)
Some college	1.94 #	1.79	1.90 #	1.74	1.30	1.06	1.40	1.21
BA+	1.62 #	2.02 *	1.64	1.89 #	1.24	1.49	1.36	1.79 *
Economic circumstances and benefits
Wages ($/hour)			1.00	1.00			0.98	0.98
Net worth (log $1,000)			1.01	0.87			1.07	0.98
Covered by health insurance (ref=no)			1.61	2.10 #			0.75	0.98
Pension eligibility 10 years after 1993 survey (ref=no)			0.71	0.82			0.93	1.08
Estimated value of monthly pension benefits (log)			1.02	1.02			0.98	1.07
Missing monthly pension benefits (ref=no)			0.56 #	0.45 *			0.77	0.71
Health
Self-rated health Fair/poor/very poor (ref=good/excellent)			1.24	0.54			1.08	0.68
CES-D (log)			1.17	1.19			1.03	1.16
Illnesses in midlife (ref=no)			1.19	1.36			1.03	1.01
Serious illness after age 53–54 (ref=no)			1.31	0.94			1.28	0.63 #
Employment circumstances
Occupational sector (ref=private)
Public			0.93	0.68			0.97	0.67
Self-employed			1.43	2.93 #			5.68 **	2.72 #
Weekly work hours (ref=35–49)
0 (Women), 0–34 (Men)			5.46 **	4.16 #			8.95 #	0.48
1–34 (Women)			-	-			4.93 **	2.07*
50+			0.62 #	0.61 #			0.58	0.89
Very satisfied with job			1.41	1.28			0.97	1.86 *
Perceived liklihood of losing job (ref=no chance)
50% or less			0.90	1.03			0.69	1.37
Greater than 50%			0.98	0.76			0.95	1.38
Flexible hours (ref=no)			1.09	1.48			0.60 #	0.86
Job requires physical effort (ref=no)			1.06	0.50			0.79	1.05
Exposed to dangerous conditions (ref=no)at work			1.00	0.77			0.88	1.02
Involuntary job exit after age 53–54 (ref=no)			1.62	1.13			1.61	0.73
Unexpected life events
Lost spouse after age 53–54 (ref=no)			2.70 #	1.56			0.51	0.66
N	480		480		604		604
df	28		72		30		76
log-likelihood	−482		−456		−609		−541
p value for LR test comparing Model 2 to Model 1			0.19				0.00

Note: FT refers to full-time employment and NW refers to not working

Table 5.

Odds ratios from multinomial logistic regression models for employment status among those who preferred not to work at age 63–64

	Men				Women
	Model 1		Model 2		Model 1		Model 2
Variable	FT	PT	FT	PT	FT	PT	FT	PT
Mid-life work experiences (ages 35–53)
Discontinous employment (ref=no)	1.06	1.98 *	1.11	2.14 *	1.21	1.47 *	0.91	1.36 #
Involuntary job exit (ref=no)	0.58 **	1.13	0.69 #	1.46	0.60 **	0.90	0.63 *	0.99
Mean occupational earnings	1.01	1.00	1.01 #	1.00	1.01	1.00	1.01	1.00
Proportion years self-employed	0.47 *	0.69	0.67	0.69	0.61	0.53 *	1.66	1.38
Proportion years covered by private pension	2.29 **	1.78 *	1.85 *	1.32	0.77	1.31	1.35	1.46 #
Proportion years in part-time employment	-	-	-	-	1.38	0.67 *	1.12	0.80
Union membership (ref=no)	2.04 **	1.41 *	1.61 *	1.13	1.65 *	1.47 *	1.94 **	1.18
Occupational aspirations and outcomes (ref=achieved aspiration)
Did not achieve aspirations	1.13	1.57 *	1.19	1.61 *	1.53 *	1.30	1.30	1.21
Did not state aspirations/homemaker	1.66 #	1.45	1.70 #	1.56	1.23	1.31	1.26	1.28
Not interviewed in 1975	2.27 *	1.99 *	2.27 *	2.05 *	0.97	0.87	0.93	0.82
Background characteristics
Age	1.27 *	0.90	1.25 #	0.88	1.30 *	1.04	1.30 #	1.02
Marital Status (ref=married, spouse no pension)
Married, spouse has pension	1.41 *	0.95	1.36 #	1.00	1.79 *	1.67 *	1.76 *	1.60 *
Not married	1.29	1.29	1.47	1.27	0.58 *	1.11	0.77	1.15
Educational Attainment (ref=high school)
Some college	0.70	1.13	0.72	1.23	0.71 #	0.74	0.67 #	0.71 #
BA+	0.89	0.77	0.91	0.78	1.15	0.90	1.10	0.71
Economic circumstances and benefits
Wages ($/hour)			1.00	1.01			1.03 *	1.00
Net worth (log $1,000)			1.22 **	0.99			1.09 *	1.03
Covered by health insurance (ref=no)			1.16	1.84 #			0.61 *	1.00
Pension eligibility 10 years after 1993 survey (ref=no)			1.67 *	1.48			1.45 #	1.47 #
Estimated value of monthly pension benefits (log)			1.00	1.02			0.92 *	0.95 #
Missing monthly pension benefits (ref=no)			0.55 **	0.77			1.15	1.26
Health
Self-rated health Fair/poor/very poor (ref=good/excellent)			1.36	3.55 **			1.34	1.55
CES-D (log)			0.87	1.08			1.08	0.99
Illnesses in midlife (ref=no)			1.32 #	1.07			1.52 **	1.01
Serious illness after age 53–54 (ref=no)			1.05	1.13			1.33	1.01
Employment circumstances
Occupational sector (ref=private)
Public			1.21	1.10			0.96	2.19 **
Self-employed			0.78	1.48			0.44 #	0.41 *
Weekly work hours (ref=35–49)
0 (Women), 0–34 (Men)			1.31	2.06			6.73 **	1.69
1–34 (Women)			-	-			2.69 **	0.74
50+			0.72 #	1.05			0.59 *	0.92
Very satisfied with job			0.80	1.13			1.09	0.80
Perceived liklihood of losing job (ref=no chance)
50% or less			0.65 *	0.97			0.78	0.74
Greater than 50%			0.64 *	1.39			0.86	0.73
Flexible hours (ref=no)			0.85	0.78			1.01	1.02
Job requires physical effort (ref=no)			1.03	1.11			1.18	0.79
Exposed to dangerous conditions (ref=no)at work			1.06	1.07			0.80	0.86
Involuntary job exit after age 53–54 (ref=no)			0.66 #	0.44 **			1.11	0.97
Unexpected life events
Lost spouse after age 53–54 (ref=no)			1.59	2.06 #			0.83	0.99
N	1,208		1,208		1,480		1,480
df	28		72		30		76
log-likelihood	−1,099		−1,048		−1,254		−1,168
p value for LR test comparing Model 2 to Model 1			0.00				0.00

Note: FT refers to full-time employment and PT refers to part-time employment

Preferences for full-time employment

We begin by describing results for those who stated a preference for full-time work. Table 3 shows that mid-life employment stability is unrelated to the likelihood that men realize preferences for full-time work at ages 63–64. The results of Model 1 indicate that those who spent a greater proportion of mid-life years in self-employment were more 2.8 times more likely realize preferences for full-time work rather than working part-time whereas the odds of full-time work were substantially lower than the odds of not working for those who spent more time with private pension coverage. These differences remained statistically significant in Model 2 after controlling for late mid-life measures of economic circumstances and benefits, health, employment circumstances, and spousal loss.

Among women, those who experienced earlier involuntary job loss were twice as likely to realize preferences for full-time work rather than working part-time at age 63–64. Realization of preferences for full-time employment was negatively associated with the proportion of mid-life years in part-time employment in the contrast with part-time work. Controlling for posited mediators in Model 2 did not alter the magnitude of these relationships between earlier work experiences and the likelihood of realizing preferences for full-time work. However, only the coefficient for time spent in part-time work remains significant at p <.10, reflecting the difficulty of precisely estimating coefficients for this small sample.

Preferences for part-time employment

Table 4 describes the likelihood of realizing preferences for part-time employment, the least likely preference to be realized. For men, several measures of mid-life work experiences were related to employment status at age 63–64 and, in most cases, these relationships remained statistically significant after controlling more temporally proximate correlates. For example, earlier experience of involuntary job loss was associated with a 50% lower likelihood of realizing preferences for part-time employment (relative to full-time employment) and discontinuous employment was associated with a similar reduction in the odds of part-time employment relative to not working. The magnitude and significance of these relationships was largely unchanged in Model 2. In contrast, employment in higher status jobs across midlife was associated with a higher likelihood of part-time work (relative to full-time work) and men who spent more time self-employed in mid-life were substantially more likely to realize their preferences for part-time employment relative to not working. The latter relationship was no longer significant in Model 2, reflecting the fact that men who spent more time self-employed during mid-career were also more likely to be self-employed at age 53–54, a characteristic associated with realization of preferences for part-time work at ages 63–64.

Like men, women who did not work continuously across mid-life were likely to be not working at age 63–64 rather than to realize their preferences for part-time employment. This relationship was not explained by the inclusion of economic circumstances, health, employment circumstances, and loss of spouse in Model 2. Also similar to the results for men, women who worked in higher status occupations across mid-life were more likely to realize preferences for part-time employment at age 63–64 (relative to full-time employment) and this relationship is not attenuated in Model 2. Model 1 also indicates a relatively high likelihood of realizing preferences for part-time employment among women who spent a greater proportion of mid-life in part-time employment and women who either did not state occupational aspirations at age 35–36 or indicated a desire to be a homemaker. Both of these relationships were explained by the fact that these women were more likely to work part-time or not at all at age 53–54, a strong predictor of part-time employment at age 63–64. Women who spent more of their mid-life years covered by private pension plans had a relatively low likelihood of realizing preferences for part-time employment (relative to not working), a relationship that become statistically meaningful in Model 2.

Preferences for not working

Table 5 describes the relative likelihood of different employment statuses at age 63–64 for those who stated a preference for not working, the most common preference and the most likely to be realized. For these men and women, several mid-life work experiences were related to the likelihood of realizing preferences. For men, discontinuous employment across mid-life was associated a higher likelihood of realizing preferences for not working (relative to working part-time) whereas involuntary job loss was associated with a lower likelihood of not working relative to working full-time. Both of these relationships remained statistically significant after controlling for temporally proximate characteristics in Model 2. Both the proportion of mid-life years spent in jobs that provide private pension coverage and union membership were strongly associated with realization of preferences for not working. In the contrast between not working and part-time work, these relationships disappeared in Model 2, reflecting the fact that these mid-life work experiences were also associated with higher levels of health insurance coverage at age 53–54 and pension eligibility at age 63–64. The proportion of years spent in self-employment is associated with a relatively low likelihood of not working (relative to working full-time) and this is explained by the fact that self-employment at age 53–54 is negatively associated with eligibility for private pension benefits at age 63–64,a strong predictor of not working. Finally, earlier occupational aspirations were also related to the likelihood of realizing preferences for not working, with those who did not achieve their earlier occupational aspirations more likely to be not working relative to working part-time and those who did not state aspirations more likely to be not working relative to working full-time. These relationships are observed in both the baseline and full models.

For women, like men, discontinuous employment across mid-life was associated with a higher likelihood of not working relative to working part-time while involuntary job loss in mid life was associated with a lower likelihood of not working relative to full-time work. These relationships are observed in both Model 1 and Model 2. Results for Model 1 also show that the likelihood of realizing preferences for not working was higher for women who belonged to a labor union and lower (relative to part-time work) among those who spent more mid-life years self-employed and in part-time employment. Finally, women who did not achieve their occupational aspirations were more likely to realize preferences for not working (relative to full-time work). Some, but not all, of these relationships remained statistically significant in Model 2. The role of part-time employment in mid-life was explained by continued part-time work at age 53–54 and the positive (although not statistically significant) relationship between limited work hours in late mid-life and the odds of working part-time at age 63–64. Self-employment at age 53–54, a strong predictor of part-time employment at age 63–64, mediated the relationship between self-employment in mid-life and the high odds of part-time employment relative to not working. Similarly, public sector employment at age 53–54, a strong predictor of not working at age 63–64 mediated the relationship between union membership in mid-life and the high odds of not working relative to part-time employment. The fact that women who did not express occupational aspirations at age 35–36 were less likely to be working at age 53–54 explains their relatively high odds of realizing preferences for not working at age 63–64. The high odds of not working relative to full-time work among those who belonged to a union and the high odds of not working relative to part-time work among those who spent more of their mid-life years in jobs providing private pension coverage remained unchanged in Model 2.

Discussion

The diminished roles of established institutional incentives and normative guidelines in shaping the retirement process suggest that individual success or failure in realizing preferences for continued work or full retirement may be an increasingly important source of variation in economic and psychological well-being at older ages. At the same time, individualization of the life course provides strong theoretical reason to expect that realization of later-life employment preferences may be influenced by earlier life experiences and processes of cumulative stratification. In this paper, we have used uniquely rich data on the work lives of the large sample of older Americans in the Wisconsin Longitudinal Study to examine linkages between mid-life work experiences and the realization of preferences for work and retirement at older ages.

Our results show that several aspects of work experiences earlier in the life course are associated with the likelihood of realizing one's preferred work status at prime retirement ages. These relationships are complex and depend upon later-life employment preferences. For example, private pension coverage across mid-life is associated with a lower likelihood of realizing preferences for full-time employment but a higher likelihood of realizing preferences for not working. Measures of employment instability during mid-life are associated with a lower likelihood of realizing preferences for part-time employment at older ages. In contrast to our hypotheses, we find little evidence of gender differences in relationships between employment experiences across the life course and realization of later-life employment preferences. Indeed, our results indicate that, with two exceptions, gender differences in relationships between mid-life work experiences and realization of later life employment preferences are small. One exception is the role of private pension coverage across mid-life which is a strong predictor of men's realization of preferences for not working and their failure to realize preferences for full-time work but is largely unrelated to women's employment outcomes at age 63–64. The other exception is part-time employment in mid-life, a strong correlate of women's later-life employment status but a very uncommon experience for men.

The strong relationships between mid-life work experiences and the likelihood of realizing later-life employment preferences is consistent with ideas of cumulative stratification of the life course. Evidence that some of these relationships are mediated by established temporally-proximate correlates of retirement outcomes provides insights into the cumulative pathways through which earlier events are related to retirement outcomes. However, the fact that several of these relationships remained statistically significant in our full model suggests the need for further evaluation of the cumulative pathways linking mid-life work experiences to the realization of later-life employment preferences. One possible explanation is that we have not adequately measured the temporally proximate characteristics that mediate relationships between earlier work experiences and later-life employment outcomes conditional on preferences. Another possible interpretation is that, in the changing context of retirement, later-life employment outcomes may be linked to earlier work experiences via other mechanisms that have yet to be identified. Potentially fruitful areas for further investigation along these lines include individual variation in planning and preparation for retirement, attitudes toward work and leisure, and access to attractive work opportunities at older ages.

Our findings also highlight several key relationships between earlier work experiences and the realization of later-life employment preferences that should be reevaluated when data on the retirement experiences of the large baby boom cohorts become available. For example, does the relationship between unstable employment across mid-life and the lower likelihood of realizing preferences for part-time employment continue to hold for cohorts characterized by greater employment instability across the life course as well as increasing preferences for part-time employment at older ages? Will private pension coverage across the life course continue to be negatively associated with the likelihood that older men realize preferences for full-time employment among cohorts more directly influenced by emerging public and private efforts to promote extended labor force attachment at older ages? Will part-time employment across the life course continue to be associated with a lower likelihood that women realize preferences for full-time work at older ages? What are the implications of changes in the nature of private pension plans and declining union membership for the likelihood realizing preferences for full retirement at older ages?

There are many ways in which the work presented here may be improved and extended. We have used only a small portion of the rich information in the WLS on work experiences across the life course, and subsequent work should explore other potentially relevant measures of individual work histories. It should also build on the life course concepts of overlapping spheres and linked lives to incorporate information on earlier experiences in other life domains such as family and health as well as information on spouses' employment experiences across the life course. This may be particularly important for understanding relationships among later-life employment preferences and outcomes and employment experiences across the life course for women. Because the WLS sample is comprised largely of white high school graduates who live in or near Wisconsin, it would also be useful to calibrate our findings, to the extent possible, with results from nationally representative surveys such as the Health and Retirement Study.

It will also be important to focus on changes in preferences in response to new information or change in life circumstances. Given data constraints, we have been forced to assume that employment preferences are constant following their articulation in the 1993 survey. Research on retirement expectations has demonstrated change across the preretirement years (Ekerdt, DeViney, and Kosloski 1996) and it is reasonable to expect similar variability in preferences regarding employment and retirement. Given that a substantial proportion of the large baby boom cohorts now approaching retirement age prefer to continue working part-time at older ages and given evidence that realization of plans for continued part-time work are the least likely to be realized, understanding how people adjust preferences to existing opportunities and the extent to which realignment of preferences or failure to realize preferences is related to well-being will be an important component of research on the retirement experiences of Americans in the years to come.

Refining analyses in these ways to further our understanding of linkages between experiences earlier in the life course and the realization of later-life employment preferences is of potentially great value for understanding variation in the retirement process of the baby boom cohorts. These cohorts are now approaching an increasingly unscripted retirement process with life experiences and retirement preferences that are far more heterogeneous than those of the WLS cohort and preceding cohorts. At the same time, the recent economic downturn and associated deterioration in job security and pension wealth has presumably had a profound impact on the retirement plans of these cohorts. A better understanding of life course influences on retirement outcomes is valuable not only for social scientists, but also for individuals contemplating and planning for their own retirement and for those involved in the formulation of later-life employment policies. Social scientists have discussed and documented the deinstitutionalization or individualization of retirement, but they have yet to fully explore the ways in which experiences across the life course influence later-life patterns of employment in this new context. As more individuals approach retirement with plans for extended employment including part-time work or phased retirement (AARP 2004; Moen and Roehling 2005), effective planning and preparation may be enhanced by a fuller understanding of the ways in which earlier work experiences are associated with the likelihood of experiencing alternative employment trajectories at older ages. Similarly, individual firms and policy makers designing policies to facilitate extended labor force participation will benefit from an understanding not only of the prevalence and nature of plans for work at older ages, but also of the ways in which individual life experiences contribute to variation in when and how older workers leave the labor force.