Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2011 Feb 10;66B(2):249–259. doi: 10.1093/geronb/gbq106

Precarious Employment, Bad Jobs, Labor Unions, and Early Retirement

James M Raymo 1,✉, John R Warren 2, Megan M Sweeney 3, Robert M Hauser 1, Jeong-Hwa Ho 4
PMCID: PMC3041978  PMID: 21310772

Abstract

Objectives.

We examined the extent to which involuntary job loss, exposure to “bad jobs,” and labor union membership across the life course are associated with the risk of early retirement.

Methods.

Using data from the Wisconsin Longitudinal Study, a large (N = 8,609) sample of men and women who graduated from high school in 1957, we estimated discrete-time event history models for the transition to first retirement through age 65. We estimated models separately for men and women.

Results.

We found that experience of involuntary job loss and exposure to bad jobs are associated with a lower risk of retiring before age 65, whereas labor union membership is associated with a higher likelihood of early retirement. These relationships are stronger for men than for women and are mediated to some extent by pre-retirement differences in pension eligibility, wealth, job characteristics, and health.

Discussion.

Results provide some support for hypotheses derived from theories of cumulative stratification, suggesting that earlier employment experiences should influence retirement outcomes indirectly through later-life characteristics. However, midlife employment experiences remain associated with earlier retirement, net of more temporally proximate correlates, highlighting the need for further theorization and empirical evaluation of the mechanisms through which increasingly common employment experiences influence the age at which older Americans retire.

Keywords: Cumulative Stratification, Employment, Life Course, Retirement

CHANGES in the nature of employment in the United States since the mid-1970s are well documented, with particular attention paid to increasing employment insecurity, growth in jobs that do not provide pension or health insurance benefits, and decline in unionized jobs (e.g., Kalleberg, 2000, 2009; Kalleberg, Reskin, & Hudson, 2000; Loveman & Tilly, 1988). A number of studies have demonstrated that exposure to precarious employment and “bad jobs” has negative implications for various outcomes, including earnings, wealth, and health. This work has focused primarily on young adulthood and midlife, despite compelling theoretical reasons to believe that employment experiences across the life course may have important implications for understanding individual variation in experiences at older ages (O’Rand, 1996a, 1996b; O’Rand & Henretta, 1999).

Theories of cumulative stratification provide insights into the ways in which experiences and exposures across the life course influence outcomes later in life. In short, this perspective suggests that differentials in well-being and employment circumstances within a cohort will increase over time as initial levels of advantage (or disadvantage) experienced by individuals interact with later opportunities and experiences (Dannefer, 1987; O’Rand, 1996a, 1996b; O’Rand & Henretta, 1999). In this way, employment experiences such as involuntary job loss and lack of pension and health care coverage may play an important role in shaping the distribution of financial resources, health, and other dimensions of well-being at older ages. The fact that these characteristics are also well-established predictors of retirement timing suggests that earlier employment experiences may be an important indirect source of variation in age of retirement.

A better understanding of linkages between earlier employment experiences and retirement timing is particularly important in the context of growing tension between the long-term trend toward earlier retirement and recent efforts to promote extended labor force participation at older ages. Although the modal age of retirement in the United States is now 62 years (Warner, Hayward, & Hardy, 2010), public and private policy efforts have increasingly sought to encourage and facilitate extended labor force attachment at older ages in response to concerns about 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 & Quinn, 2002; Morton, Foster, & Sedlar, 2005; Munnell & Sass, 2008).

Theories of cumulative stratification suggest several, potentially offsetting, ways in which precarious employment and exposure to bad jobs across the life course may influence the likelihood of early retirement via temporally proximate correlates of retirement. On the one hand, the experience of job loss or exposure to bad jobs may reduce the likelihood of retiring early via lower wealth accumulation and limited pension benefits. On the other hand, these same employment experiences might contribute to early retirement via poor health and lower commitment to work in late midlife. The fact that labor force changes since the mid-1970s have disproportionately affected women (Kalleberg et al., 2000) highlights the importance of examining whether these posited indirect influences on retirement timing differ by gender.

In this article, we use detailed employment history data from the Wisconsin Longitudinal Study (WLS) to examine how exposure to precarious employment, bad jobs, and labor union membership across the life course are related to the risk of early retirement (prior to age 65). We also assess the extent to which these relationships are mediated by established correlates of retirement timing (such as private pension eligibility, wealth, and health) and examine the extent to which linkages between employment history and early retirement differ by gender.

BACKGROUND

The Changing Nature of Employment

A key feature of changes in the nature and quality of employment since the mid-1970s is declining job security (Hacker, 2008). Kalleberg (2009, p. 5) argues that layoffs have become a basic component of employers’ restructuring strategies over the past thirty years and that precarious employment has spread from unskilled, less educated segments of the labor force to all sectors of the economy. Involuntary job loss is a widely-studied indicator of employment insecurity, and recent research shows that the proportion of employees who experience involuntary job loss in a given three-year period has fluctuated between 9% and 13% since the early 1980s (Farber, 2005). Estimates from the Bureau of Labor Statistics indicate that more than 30 million involuntary job losses occurred between the early 1980s and 2004 (Kalleberg, 2009, p. 2).

Closely related to the increase in precarious employment is growth in bad jobs. The quality of jobs can be evaluated in a variety of ways, but bad jobs are frequently defined as those that do not offer private pension plans or health insurance benefits and are characterized by low wages (Hacker, 2008; Kalleberg et al., 2000; Mishel, Bernstein, & Allegretto, 2007). Recent studies show that 21% of employed Americans did not have private health insurance coverage in 2008 (Turner, Boudreaux, & Lynch, 2009) and that half did not have private pension coverage in 2004 (Munnell & Perun, 2006). In their tabulations of data from the 1995 Current Population Survey, Kalleberg and colleagues (2000) show that 60% of American workers had a job that was bad on at least one indicator (no pension, no health insurance, or low wages) and that one in seven had a job that was bad on all three indicators.

A third widely-studied component of changes in the nature of employment is decline in unionized jobs. The proportion of employees in labor unions fell from .28 in 1970 to .12 in 2007 (Statistical Abstract of the United States, 1980 and 2009 volumes). Because unionized jobs are more likely to provide employment security, benefits, and higher pay, declining union membership is closely related to the increase in precarious employment and bad jobs (Kalleberg et al., 2000).

Explanations of these changes have emphasized increasing global competition, technological change, deregulation, the ability to outsource low-skill work to low-wage countries, and the shift from manufacturing to service and information industries, all of which have prompted employers to seek more flexible employment relations (Hacker, 2008; Hipple & Stewart, 1996; Tilly, 1996). This flexibility has been achieved by moving from long-term employment to an increased reliance on nonstandard work arrangements, including temporary/fixed-term employment and part-time employment, which allow for reductions in the cost of providing benefits and facilitate the expansion or contraction of employee numbers in response to shifting demand (Farber & Levy, 2000; Hipple & Stewart, 1996; Kalleberg, 2009; Tilly, 1996).

Relationships Between Midlife Employment Experiences and Early Retirement

Theories of cumulative stratification suggest several reasons to expect that the quality of employment across the life course has important implications for when older Americans retire. Of particular importance is evidence that precarious employment, job loss, and employment in bad jobs are negatively associated with various dimensions of well-being, including earnings (Farber, 2005; Stevens, 1997), wealth (Chan & Stevens, 2001), physical health (Strully, 2009), and psychological health (Burgard, Brand, & House, 2007; Grzywacz & Dooley, 2003). Involuntary job loss has also been linked to reductions in subsequent labor force participation and a higher likelihood of subsequent employment in part-time jobs (Farber, 2005). Job loss in late midlife, in particular, has been linked to earlier retirement (Chan & Stevens, 2004).

Because financial well-being, health, and employment circumstances are also well-established correlates of retirement age, processes of cumulative stratification provide an important theoretical link between earlier employment experiences and the timing of retirement. Developing an empirical understanding of this link is particularly important in light of growing efforts to reverse the long-term trend toward earlier retirement and prolong labor force attachment at older ages. Efforts to project the retirement behavior of the baby boom cohort, to respond to concerns about the loss of skilled older workers and pressures on social security, and to ensure the financial well-being of older Americans will benefit from an understanding of the extent to which exposure to increasingly common employment experiences across the life course is associated with the likelihood of early retirement.

Hypothesized relationships are mixed. On the one hand, earlier exposure to precarious employment and bad jobs may reduce the likelihood of early retirement by limiting the financial capacity to retire. We posit that involuntary job loss and employment in bad jobs will contribute to a lower likelihood of early retirement via more limited access to private pension coverage and lower wealth accumulation (Crystal & Shea, 1990; O’Rand & Shuey, 2007). We also expect the risk of early retirement to be lower among those who did not belong to a labor union and are thus less likely to rehired following layoff, to be eligible for employer-provided pension benefits, or to face the institutionalized incentives for early retirement present in many unionized jobs (Hardy, Hazelrigg, & Quadagno, 1996). This scenario is consistent with recent evidence that the long-term trend toward earlier retirement has slowed or reversed (Friedberg, 2007) and that the economic feasibility of early retirement is declining. Indeed, recent projections suggest that about one-third of the early baby boom cohort (born 1946–1954) will not be able to maintain their pre-retirement standard of living even if they work full-time until age 65 (Munnell, Golub-Sass, Perun, & Webb, 2007), and attitudinal survey data indicate that increasing proportions of Americans approaching retirement cite economic necessity as a reason for their plans to continue working beyond current peak ages of retirement (AARP, 2004; Merrill Lynch, 2006).

On the other hand, exposure to precarious employment and bad jobs is expected to promote early retirement via poor health and more marginal labor force attachment. Earlier studies have found that unstable careers and spells of unemployment across the life course are associated with earlier retirement for men via higher rates of disability as well as weaker attachment to work and a lower likelihood of engaging in rewarding work (Hayward, Friedman, & Chen, 1998). Exposure to involuntary job loss and bad jobs across the life course may also be associated with a higher likelihood of experiencing unforeseen events such as health decline or late-career job loss that result in relatively early, and perhaps involuntary, retirement (Chan & Stevens, 2004).

The rise in precarious employment and the spread of bad jobs has been particularly pronounced among women (Kalleberg et al., 2000), suggesting that there may be important gender differences in relationships between earlier work experiences and retirement timing. Again, theoretical expectations are ambiguous. The effect of earlier employment experiences may be particularly strong for women who are more likely to experience career interruptions during prime childbearing years that can result in shorter employment tenure and a higher likelihood of employment in part-time nonstandard jobs that do not provide pension and health insurance benefits (O’Rand, 1988). Alternatively, it is possible that earlier employment experiences are less relevant for women whose retirement decisions may be more sensitive than men’s to family circumstances such as caregiving obligations and to the characteristics and circumstances of their spouses (Szinovacz & DeViney, 2000).

Our objective in this article was to extend our understanding of linkages between work experiences across the life course and early retirement using data from the WLS. In particular, we examine how exposure to precarious employment, bad jobs, and labor union membership across the life course are related to the risk of early retirement (prior to age 65) and assess the extent to which these relationships are mediated by established correlates of retirement timing. In doing so, we make three important empirical and theoretical contributions. First, the WLS contains rich life history data for one of the first cohorts to approach retirement after facing increased exposure to precarious employment and bad jobs for a large proportion of their working lives, thus making it a uniquely valuable source of insight into the long-term implications of the changing nature of employment over the past three decades. Second, the WLS data allow for improvement on earlier studies on linkages between work experiences and variation in retirement outcomes (e.g., Hayward, 1986; Hayward, Grady, Hardy, & Sommers, 1989; Hayward et al., 1998; O’Rand & Henretta, 1982; O’Rand & Landerman, 1984), which are limited by their reliance on older surveys such as the National Longitudinal Surveys of the late 1960s, their focus on men, and their consideration of relatively limited aspects of work histories (e.g., characteristics of longest job). Finally, we extend recent life course theorization by evaluating hypotheses about the ways in which processes of cumulative stratification influence early retirement, one important aspect of the increasingly heterogeneous retirement process (e.g., O’Rand, 1996a, 1996b; O’Rand & Henretta, 1999).

DATA AND METHODS

Sample

The WLS is a long-term study of a random sample of 10,317 men and women who were first interviewed prior to graduating from Wisconsin high schools in 1957 and reinterviewed in 1975, 1993, and 2004. Several features of the WLS make it ideal for our purposes. First, it provides detailed information on almost all jobs that respondents have held between ages 35 and 65, allowing us to measure work histories in far greater detail than in previous studies. Second, the WLS respondents were 35–36 years old in 1975, making this survey a rich source of information on the employment histories of one of the first cohorts exposed to the rise in precarious employment and bad jobs and the decline in union membership from early career through peak retirement ages. Third, the WLS cohort is one of the first in which significant numbers of women have worked throughout the life course, allowing us to examine gender differences in relationships between earlier work experiences and retirement.

Our analyses are based on data collected in the 1993 survey (conducted in 1992–1994 with a response rate of 87%) and the 2004 survey (conducted in 2003–2005 with a response rate of 85%). We excluded 1,593 members (15%) of the original cohort who did not participate in either the 1993 phone interview (when they were 53–54 years old) or the 2004 phone interview (when they were 64–65 years old) and six respondents who either had missing information on age or were several years older than the rest of the Class of 1957 graduates. These restrictions leave us with a base sample of 8,718 individuals. Auxiliary analyses indicate that those with only a high school degree were more likely to be among the 1,593 who were excluded due to refusal to participate in either survey (n = 1,048) or death prior to the 1993 survey (n = 545). Our results may thus be biased to the extent that relationships between unobserved earlier work experiences and retirement outcomes are systematically different for these missing cases.

It is also clear that the likelihood of being censored in 1993 due either to refusal to participate in the 2004 survey (n = 879) or to death between 1993 and 2004 (n = 578) was somewhat higher among those with less education and longer exposure to bad jobs. Because we have no information about these individuals’ experiences after the 1993 interview (other than year and month of death for those who died), we censored these observations in the year of the 1993 survey. To assess the potential impact of this attrition, we estimated auxiliary models based on different assumptions about the unobserved experiences of this group subsequent to the 1993 survey. In particular, we assumed that (a) all retired in the year following the 1993 survey, (b) all retired in 2004 (for those lost to follow-up) or in the year of death (for those who died), (c) all retired halfway between the 1993 survey year and either the year of the 2004 survey or the year of death, and (d) none retired prior to 2004 or year of death. In all cases, we assume that values of independent variables remained constant from 1993 until the year of retirement or censoring. In all but one scenario, estimated coefficients for indicators of precarious employment, exposure to bad jobs, and belonging to a labor union were robust to these alternative treatments of data missing due to attrition following the 1993 survey. We note the one exception later when discussing the results.

Variables

Retirement timing.—

In both the 1993 and 2004 surveys, respondents were asked whether they considered themselves to be partly retired, completely retired, or not retired at all. Those who were partly or completely retired were asked the month and year in which they retired. For 178 respondents who did not report retirement status in either survey, or reported being retired but did not provide a date of retirement, we used other sources of information to determine retirement timing. Reports of retirement as a reason for leaving specific employment spells and information on age of initial receipt of private pension benefits and social security benefits enabled us to construct measures of retirement experience and timing for 159 of those with missing data on self-stated retirement. We excluded the remaining 19 (0.02% of the total sample) who provided no usable information on retirement. We also excluded 90 respondents whose reported retirement age was younger than their age in 1975, most (89%) of whom were women with very young reported ages of retirement (the mean age of retirement for this group was 26). Our final analytical sample is thus 8,609 (i.e., 8,718 − 19 − 90 = 8,609) or 88% of the 9,730 sample members who survived to the 1993 survey.

We began by constructing a data file consisting of person-year records in which each respondent contributed one observation per year from 1975 through the year of first self-stated retirement or the 2004 survey, whichever came first. As noted previously, those who did not participate in the 2004 survey were censored in the year of the 1993 survey. By constructing these person-year data, we were able to easily incorporate time-varying information on pension eligibility, job characteristics, family circumstances, and health status collected in the 1993 and 2004 surveys and to flexibly specify the baseline hazard of retirement in discrete-time event history models.

Midlife employment experiences.—

We constructed five measures of midlife work experience that reflect cumulative exposure to precarious employment, bad jobs, and unionized jobs. Except for the measure of labor union membership, these variables were constructed from the employment history data collected at ages 53–54 in the 1993 survey and at ages 64–65 in the 2004 survey. Employment histories in the WLS surveys are comprised of multiple employment spells—uninterrupted periods of time working for the same employer, including self-employment. These data provide detailed information on employment status and the characteristics of jobs held by respondents during each year between 1975 and 2004. This information includes the years that respondents started and stopped working for that employer, the reason for ending that employment spell, whether they worked full- or part-time, the industry and occupation when they began working for that employer, health insurance coverage, and pension coverage.

Our measure of exposure to precarious employment is a time-varying indicator of whether respondents ever left a job involuntarily prior to first retirement. Based on open-ended responses to questions asking the reason for ending each employment spell, we define involuntary job loss as cases in which the reported reason was business closing, downsizing, relocation, termination, or layoff. This measure is equal to zero in all person-years prior to the year in which involuntary job exit was reported (if any) and equal to one in that year and all subsequent years. For the small number of respondents who report multiple job losses, we focus on the most recent experience. Based on evidence that job loss in late midlife is a particularly import predictor of early retirement (Chan & Stevens, 2004), we define two categories of involuntary job loss to distinguish those that occurred at younger ages (before age 50) from those that occurred in late midlife (age 50 and older).

To operationalize exposure to bad jobs, we constructed time-varying indicators of the cumulative proportion of working years since 1975 spent in jobs that (a) did not offer private pension plans, (b) did not offer health insurance coverage, and (c) were characterized by low earnings. As discussed previously, the primary objective of this study was to advance our understanding of the implications of structural changes in the labor market by examining relationships between exposure to bad jobs across the life course and early retirement. We therefore use information on whether each reported job offered private pension and health care coverage (rather than whether respondents participated in such programs) and follow the measurement strategy employed by Kalleberg and colleagues (2000) in using information on job-specific average wages (rather than the wages of individual respondents). We define low-wage jobs as those that fall below the median value of occupational earnings—the percentage of people in a given occupation who reported hourly wages of at least $14.30 in the 1990 census (Hauser & Warren, 1997). For each of these three indicators of bad jobs, we calculated time-varying measures of exposure by dividing the cumulative number of years employed in a bad job (since age 35) by the cumulative number of years employed since age 35. At any given age, these measures thus indicate the cumulative proportion of working years since age 35 spent in bad jobs.

Due to the limited information on labor union membership available in the surveys, our measure of belonging to a union is not truly a time-varying variable. It is constant with a value of one for those who reported that they belonged to a labor union in 1975, constant with a value of zero for those who did not report union membership in either 1975 or 1993, and changes from a value of zero to a value of one in 1993 for those who reported union membership in 1993 but not in 1975. Because these questions referred only to the jobs held in 1975 and at the time of the 1993 survey, we cannot identify labor union membership for all jobs held since age 35.

Correlates of retirement timing.—

Drawing on theories of cumulative stratification, we attempt to evaluate pathways through which earlier exposure to precarious employment, bad jobs, and unionized jobs may be associated with retirement outcomes by including time-varying contemporaneous measures of financial well-being, job characteristics, employment status, and health. We lag these measures by one year so that respondents’ economic, employment, and health statuses at age t−1 are used to predict the likelihood of retirement at age t. In contrast, the measures of unexpected job loss, exposure to bad jobs, and union membership (which are also lagged by one year) summarize cumulative employment experiences between age 35 and age t−1.

We measured financial well-being using information on private pension eligibility and wealth. Pension eligibility is a time-varying trichotomous indicator that distinguishes those who are eligible for private pension benefits at a given age from those who are covered but not yet eligible and those who are not covered by an employer-sponsored pension plan. Wealth is a time-constant indicator of respondents’ percentile rank in the within-sample distribution of net worth at the time of the 1993 survey. Net worth in the WLS is constructed as the reported value of home equity, real estate, business or farm, motor vehicles, savings, and investments owned by the respondent and his/her spouse minus their reported debt. Because available information on net worth refers only to the time of the 1993 and 2004 surveys, it is impossible to construct time-varying measures without making untestable assumptions about how wealth changes over time. We therefore make the simplest assumption of no change over time in respondents’ relative position in the within-sample distribution of net worth.

Job characteristics include time-varying indicators of occupational sector of the current or last job (private sector, public sector, self-employed), employment status (full-time, part-time, not employed), occupational socioeconomic standing, and duration of the current (non-) employment spell. Occupational socioeconomic standing is measured as the value of occupational education—the percent of persons in each occupation in the 1990 Census who completed one year of college or more (see Hauser & Warren, 1997, for details).

Health status is measured using time-varying indicators of four serious illnesses or health events after age 35. These measures are constructed from information in the 2004 survey on the diagnosis of diabetes, cancer, heart problems, and stroke and are equal to zero in person-years prior to the experience of these health problems and equal to one in the year of diagnosis and all subsequent years. Because the 1993 survey did not ascertain the year of diagnosis, we assume that these health events occurred midway between 1975 and the 1993 survey for the small number of respondents who reported a diagnosis in 1993 but did not respond to the 2004 survey. Among the 1,457 respondents censored at the time of the 1993 survey, 75 reported a diagnosis of diabetes, 77 reported heart trouble, and 33 reported cancer. The 1993 survey did not ask about experience of stroke so we set this variable equal to zero for those who did not respond to the 2004 survey. We also set all four health indicators equal to zero for 558 respondents who did not respond to the part of the 1993 survey that was conducted by mail (which contained the questions on these health events).

Models

We examine relationships between earlier work experiences and retirement through age 65 by estimating discrete-time event history models using logistic regression. These models can be expressed generally as:

graphic file with name geronbgbq106fx1_ht.jpg (1)
graphic file with name geronbgbq106fx2_ht.jpg (2)

where pit is the probability that the ith respondent retires at age t, conditional on being not retired at age t−1. Xit−1 includes age, a time-invariant measure of educational attainment (high school, some college, bachelor’s degree, and graduate or professional degree), and a time-varying indicator of marital status (married, divorced/separated, widowed, and never married). Yit−1 includes the five employment experience measures of central interest. The subscript t−1 indicates that independent variables are lagged one year. Two exceptions are age and pension eligibility, both of which are measured at age t. By including the correlates of retirement timing described previously (Zit−1 in Model 2), we can evaluate the extent to which relationships between earlier employment experiences and retirement timing change after controlling for factors posited to delay retirement (lower wealth, lack of pension eligibility, shorter employment tenure, and self-employment) and factors posited to promote early retirement (lower occupational status, limited labor force attachment, and health problems). We estimate the two models separately for men and women in light of the posited gender differences described previously.

Based on the results of preliminary descriptive analyses, we specify the baseline hazard of retirement using a linear term in combination with age-specific dummy variables between ages 60 and 65. As in other studies using different data and different methods, we find that the risk of retirement rises through age 60, increases rapidly to a peak at age 62, and declines subsequently. We do not see a second peak at age 65 because age in a given person-year refers to respondents’ birthday in that calendar year and many of those coded as age 65 had not yet reached their 65th birthday by the survey date.

RESULTS

Table 1 describes the analytic sample by sex in two ways. Columns 1 and 2 describe the characteristics of individuals (observed in the year prior to retirement or censoring) and columns 3 and 4 present summary statistics for all person-year observations. Individual data show that 64% of both men and women had retired by the time of the 2004 survey (when they were 63–65 years old). Among those who retired, 95% retired before age 65 (not shown). The measures of midlife work experiences indicate that the proportion of working years that women spent in jobs without private pension coverage (.43), without health insurance coverage (.35), and with low earnings (.64) is substantially higher than that for men. The proportion with experience of involuntary job loss is similar for men (.20) and women (.17), but men were somewhat more likely to report an involuntary job loss in late midlife. Women were less likely to report belonging to a labor union (.18 vs. .33 for men). Other measures indicate that men are more highly educated; less likely to be divorced or widowed; more likely to be eligible for private pension benefits; more likely to be self-employed; much less likely to be employed part-time or not working; and have higher net worth, longer employment tenure, and somewhat higher occupational socioeconomic standing. The prevalence of major health problems is low and gender differences are small.

Table 1.

Sample Characteristics, by Sex

Individualsa
 Person-years
Variable Men Women Men Women
Retired
    No 0.36 0.36 0.97 0.97
    Yes 0.64 0.64 0.03 0.03
Age 58.36 (5.17) 57.88 (5.69) 48.68 (7.54) 48.51 (7.58)
Educational attainment
    High school 0.55 0.69 0.54 0.69
    Some college 0.14 0.13 0.15 0.13
    Bachelor's degree 0.14 0.12 0.14 0.12
    Graduate/professional degree 0.17 0.06 0.17 0.06
Marital status
    Married 0.85 0.77 0.87 0.82
    Divorced/separated 0.09 0.13 0.08 0.11
    Widowed 0.01 0.06 0.01 0.03
    Never married 0.05 0.04 0.05 0.05
Midlife work experiences
    Involuntary job exit
        None 0.80 0.83 0.89 0.90
        Before age 50 0.08 0.08 0.06 0.05
        After age 50 0.12 0.09 0.06 0.04
    Proportion of years w/o access to private pension 0.26 (0.37) 0.43 (0.42) 0.27 (0.40) 0.42 (0.45)
    Proportion of years w/o access to health insurance 0.15 (0.30) 0.35 (0.41) 0.14 (0.31) 0.34 (0.43)
    Proportion of years in low earning occupation 0.34 (0.42) 0.64 (0.43) 0.34 (0.44) 0.59 (0.46)
    Belonged to a labor unionb 0.33 0.18 0.31 0.16
Net worth in 1993 ($1000)c 249.75 (265) 190.12 (227) 250.22 (264) 185.92 (222)
Eligible for private pension benefits
    Not yet eligible 0.30 0.23 0.60 0.42
    Eligible 0.41 0.24 0.11 0.06
    Not covered 0.29 0.54 0.29 0.51
Occupational sector of current/last job
    Private 0.64 0.68 0.66 0.71
    Public 0.18 0.21 0.17 0.19
    Self-employed 0.18 0.11 0.17 0.10
Occupational educationd 56.37 (27.90) 51.67 (29.59) 57.69 (27.94) 48.55 (31.30)
Duration of current employment spell 15.21 (12.55) 9.94 (9.90) 11.67 (9.78) 6.63. (7.78)
Employment status
    Working full-time 0.88 0.59 0.93 0.59
    Working part-time 0.08 0.14 0.04 0.20
    Not working 0.04 0.27 0.03 0.22
Diagnosed with diabetesb 0.09 0.06 0.04 0.03
Diagnosed with heart diseaseb 0.06 0.02 0.02 0.01
Diagnosed with cancerb 0.05 0.07 0.02 0.03
Diagnosed with strokeb 0.01 0.01 0.00 0.01
N 4,087 4,522 95,094 103,630
Open in a new tab

Notes: Standard deviations of continuous variables are shown in parentheses.

a

Observations for individuals are from the year prior to retirement or censoring.

b

Proportion for whom the value is 1 = yes.

c

In regression analyses, this measure is transformed to percentile rank. The mean value presented in this table is based on a variable that is topcoded at $1,000,000.

d

Occupational education is a measure of occupational status equal to the percentage of persons in each occupation in the 1990 Census who completed one year of college or more (see Hauser & Warren, 1997, for details).

Table 2 presents exponentiated coefficients (odds ratios) from two discrete-time event history models for the transition to retirement through age 65. Looking at the results for Model 1, it is clear that our measures of exposure to increasingly common employment experiences are associated with a significantly lower risk of retirement (or equivalently, a lower probability of retiring by age 65). This is especially true for men, with four of the five employment history measures negatively associated with the risk of retirement. Specifically, the odds of retirement at a given age are about one-third lower for men with earlier experience of involuntary job loss (regardless of when it occurred), 51% lower for those who were consistently employed in jobs that did not provide private pension plans (relative to those who always had access to pension coverage), 17% lower for those who were always employed in jobs characterized by low earnings, and 47% higher for those who belonged to a labor union (or, equivalently, 32% lower for those who did not belong to a labor union). The proportion of working years spent in jobs that do not provide health insurance coverage is not significantly related to the risk of retirement.

Table 2.

Odds Ratios from Models of Retirement Timing, by Sex

Men
 Women
Variable Model 1 Model 2 Model 1 Model 2
Age
    Linear 1.25** 1.18** 1.20** 1.18**
    60 1.36** 1.29** 1.33** 1.25**
    61 1.43** 1.46** 1.24** 1.20*
    62 2.86** 2.59** 2.45** 2.15**
    63 1.53** 1.46** 1.35** 1.20*
    64 1.13 1.15 0.89 0.81*
    65 0.93 0.94 0.69** 0.61**
Educational attainment (a)
    Some college 0.94 1.02 0.92 0.92
    Bachelor's degree 0.88* 0.96 0.88 0.89
    Graduate/professional degree 0.72** 0.78** 0.81* 0.81*
Marital status (b)
    Divorced/separated 1.15 1.31** 0.60** 0.69**
    Widowed 0.97 1.00 0.81* 0.85
    Never married 1.33** 1.42** 0.91 1.03
Midlife work experiences
    Involuntary job exit (c)
        Before age 50 0.66** 0.84 0.89 0.95
        After age 50 0.65** 0.80** 0.79** 0.86
    Proportion of years w/o access to private pension 0.49** 0.64** 0.94 0.94
    Proportion of years w/o access to health insurance 1.08 1.20 1.12 1.09
    Proportion of years in low earning occupation 0.83** 0.74** 0.91 0.86*
    Belonged to a labor union (d) 1.47** 1.14** 1.35** 1.28**
Net worth in 1993 (percentile rank) 1.68** 1.96**
Eligible for private pension benefits (e)
    Eligible 3.44** 2.33**
    Not covered 1.93** 1.68**
Occupational sector (f)
    Public 1.59** 1.11
    Self-employed 0.63** 0.82**
Duration of current employment spell 1.01** 1.01**
Occupational education 0.99** 0.99**
Employment status (g)
    Not working 1.57** 1.17**
    Working part-time 1.29* 0.79*
Diagnosed with diabetes (d) 1.48** 1.23**
Diagnosed with heart disease (d) 1.13 1.72**
Diagnosed with cancer (d) 1.40** 1.27**
Diagnosed with stroke (d) 2.76** 1.34
N 95,094 95,094 103,630 103,630
df 19 32 19 32
Log-likelihood −8,921 −8,529 −10,595 −10,410
p value for likelihood ratio test—Model 2 vs. Model 1 0.00 0.00
Open in a new tab

Notes: Omitted categories are (a) high school, (b) married, (c) none, (d) no, (e) not yet eligible, (f) private sector, (g) working full-time.

*p < .05; ** p < .01.

In Model 2, measures of net worth, pension eligibility, occupational sector, spell duration, occupational socioeconomic standing, employment status, and health problems are all significantly related to men’s risk of retirement in expected ways. More importantly, the inclusion of these temporally proximate correlates of retirement attenuates, but does not eliminate, the statistically significant relationships between earlier employment experiences and the risk of retirement. The only change is that the coefficient for involuntary job loss prior to age 50 is no longer statistically significant at p < .05 (p = .052). Furthermore, the positive coefficient for longer exposure to jobs without health insurance benefits now approaches statistical significance (p = .09), a relationship that emerged after controlling for self-employment, a strong predictor of both limited access to health care across the life course and later retirement. One plausible explanation for this finding is that those with limited health insurance coverage across the life course may need to take another (postretirement) job that provides health insurance until becoming eligible for Medicare. However, supplementary analyses reveal that this is not the case. In most cases, early retirement for this group is followed by self-employment.

In only one case were the results presented in Table 2 sensitive to the inclusion of imputed observations for men who did not respond to the 2004 survey. When we assumed that none of these men retired and were censored at the year of the 2004 survey or the year of their death, coefficients for involuntary job exit and not belonging to a union were no longer significantly different from zero. When we make less extreme assumptions about the unobserved behavior of this group, the results are very similar to those presented in Table 2. We are thus inclined to believe that our results are reasonably robust to nonrandom loss to follow-up and death.

For women, like men, results of Model 1 indicate that experience of involuntary job loss (after age 50) is associated with a lower likelihood of early retirement and that belonging to a labor union is positively associated with the risk of retirement. Unlike men, however, women’s exposure to bad jobs across the life course is unrelated to their risk of retirement. After controlling for established correlates of retirement timing in Model 2, the relationship between employment in unionized jobs and earlier retirement is unchanged, but the coefficient for experience of involuntary job loss after age 50 is no longer significantly different from zero. Results of auxiliary models show that later retirement among women who left a job involuntarily in late midlife primarily reflects their relatively short employment duration at the time of retirement. One exception to the general pattern of attenuation in the coefficients for midlife work experiences is exposure to low-paying jobs, which is now associated with a significantly lower risk of early retirement. This relationship emerges after controlling for current occupational socioeconomic standing which is negatively associated with exposure to low-paying jobs across the life course as well as the risk of early retirement. In no case were these results sensitive to the inclusion of imputed observations for women who did not respond to the 2004 survey.

To provide a better sense of the substantive magnitude of the estimated differences with respect to earlier employment experiences, we used predicted probabilities from Model 2 to construct life table measures of the age-specific cumulative proportions ever retired. For illustrative purposes, we compare two extreme groups of hypothetical men and women. One group (A) did not belong to a labor union, experienced involuntary job loss at age 50, and spent all of their working years in jobs without private pension or health insurance coverage and characterized by low wages, whereas the other group (B) belonged to a labor union, never lost a job involuntarily, and never worked in bad jobs. Figure 1 presents the cumulative proportions ever retired by age for these two groups of men and women, assuming mean or modal values for other variables in Model 2 (i.e., high school graduates; continuously married; employed in the private sector; working full-time; not yet eligible for private pension benefits; no health problems; and mean values of employment spell duration, net worth, and occupational education). The cumulative probability of retirement by age 65 for men who experienced involuntary job loss, were exposed to bad jobs across the working life course, and did not belong to a labor union was 23 percentage points lower than that of men without such experiences. The corresponding difference for women was 9 percentage points. Differences between the two groups increase with age as the age-constant differences in the risk of retirement compound over time. This divergence is accelerated by the experience of involuntary job loss at the arbitrarily chosen age of 50 for those in group A. The larger differences for men reflect the stronger links between involuntary job loss and exposure to bad jobs across the life course and earlier retirement for men which more than offset the slightly stronger relationship between labor union membership and earlier retirement for women. Taken as a whole, these results highlight the significance of increasingly common employment experiences across the life course for understanding variation in the likelihood of early retirement, especially among men.

Figure 1.

Figure 1.

Open in a new tab

Predicted cumulative percent retired, by age, sex, and employment experiences. Note: A refers to those with all five characteristics of interest (i.e., experienced involuntary job loss at age 50, did not belong to a labor union, and spent all working years employed in jobs without private pension coverage, without health insurance, and characterized by low wages). B refers to those who belonged to a union, did not experience involuntary job loss, and never worked at “bad jobs.”

DISCUSSION

Despite the well-documented increase in exposure to precarious employment and bad jobs and the decline in unionized jobs, research on linkages between these employment experiences and retirement outcomes remains limited. In this article, we used uniquely rich data collected across the lives of a large cohort of older American men and women to evaluate these relationships. We found that experience of involuntary job exit, exposure to jobs that do not provide pension benefits, and exposure to low-paying jobs were all associated with a lower likelihood of early retirement. These relationships are stronger for men than for women. We also found a strong relationship between labor union membership and early retirement for both men and women.

Consistent with hypotheses based on theories of cumulative stratification, these relationships were attenuated when we controlled for established temporally proximate correlates of retirement timing. Private pension eligibility and occupational sector, in particular, played an important role in explaining linkages between earlier employment experiences and the risk of retirement through age 65. However, employment experiences across the life course remained significantly related to men’s risk of early retirement net of pension eligibility, net worth, employment circumstances, and health. Involuntary job loss after age 50, exposure to low-paying jobs, and jobs without pension coverage remained significantly associated with a lower risk of early retirement, whereas labor union membership continued to be positively associated with early retirement. For women, exposure to low-paying jobs emerged as a significant correlate of later retirement, whereas union membership remained a strong predictor of early retirement. In sum, we find that exposure to precarious employment and bad jobs are associated with a lower likelihood of early retirement, especially for men. Part, but not all, of these relationships can be attributed to the fact that those with more exposure to precarious employment and bad jobs are less likely to be eligible for private pension benefits, more likely to be self-employed, and have shorter employment duration as they approach typical retirement ages.

Gender differences in the meaning of work and career may account for the relatively weaker linkages between women’s earlier work experiences and their retirement timing. The WLS respondents belong to a transitional cohort between earlier cohorts of women who derived their identity primarily from family roles and later cohorts who have derived their identity from both work and family roles (Goldin, 2006). We suspect that women’s retirement timing may be more sensitive to their husbands’ characteristics than to their own employment history, but the absence of time-varying information on spouses’ characteristics prevents us from examining this possibility. It will be important to reexamine gender differences in younger cohorts characterized by greater symmetry in men’s and women’s employment experiences across the life course when appropriate data are available. It will also be important to pay attention to other characteristics that may moderate relationships between employment experiences across the life course and retirement timing. For example, labor union membership and self-employment are associated with employment stability and job quality in ways that may shape relationships between employment experiences across the life course and retirement outcomes.

Our results are particularly interesting in the context of increasing interest in promoting extended labor force attachment at older ages, suggesting that exposure to less favorable employment circumstances across the life course may contribute to a lower likelihood of early retirement. Results are consistent with the hypothesis that changing employment patterns across the life course contribute to increasing heterogeneity in the timing of retirement via the rising proportion of older Americans who approach older ages without sufficient financial resources and benefits to retire early (Moore & Mitchell, 2000). However, our ability to do more than speculate about this possibility is constrained by two important limitations of our study.

The first limitation is that our ability to adequately measure established temporally proximate correlates of retirement timing is constrained by the information available in the survey. Although the WLS does contain information about savings, financial assets, and pension benefit levels, these measures refer to the time of the survey and are thus not adequate for constructing the truly time-varying indicators of economic well-being needed for our analyses. This is particularly true of pension benefit levels, which were assessed only in the 2004 survey and thus cannot be meaningfully incorporated in analyses of retirement across the full range of ages that we examine. Careful evaluation of the mechanisms underlying the relationships we have documented is an important task for subsequent research when the necessary detailed information on individual employment, economic, and health characteristics across the life course become available. The first large-scale study to provide such information will be the NLSY79, whose participants will enter prime retirement ages 10–15 years from now.

A second limitation is the fact that the WLS contains only limited information about the life histories of spouses. Spouses’ employment experiences across the life course and their economic, employment, and health circumstances at older ages may be related to respondents’ midlife employment experiences and retirement timing in ways that would shed light on our findings. For example, the relatively low likelihood of early retirement among men with less favorable employment circumstances in midlife may reflect their marriage to women with similar employment experiences or perhaps more limited labor force attachment across the life course. We also recognize that exposure to precarious employment and bad jobs across the life course may affect retirement timing in ways that have yet to be adequately theorized. For example, it may be that exposure to precarious employment and bad jobs across the life course influences preferences for work and leisure, life goals, risk aversion, and other characteristics that may affect retirement timing but are not included in the WLS or in most other surveys.

Refining analyses in these ways to further our understanding of linkages between employment experiences across the life course and retirement timing is of potentially great value for understanding variation in the retirement process of the large baby boom cohort. Relative to the WLS cohort and preceding cohorts, the baby boomers were at an earlier career stage when marked changes in the nature of employment emerged in the mid-1970s. A better understanding of the ways in which earlier employment experiences are associated with retirement timing 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 retirement policies. Social scientists have discussed and documented the deinstitutionalization or individualization of retirement, but they have yet to fully explore the ways in which employment experiences across the life course influence retirement outcomes in this new context. As more individuals approach retirement with plans for extended employment, including part-time work or phased retirement, effective planning and preparation may be enhanced by a fuller understanding of the ways in which earlier employment experiences are associated with the likelihood of early retirement. Similarly, the efforts of individual firms and policy makers to implement policies designed 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 variation in earlier employment experiences contribute to variation in the risk of early retirement.

FUNDING

The research reported herein was supported by the National Institute on Aging (R01 AG-9775 and P01-AG21079), by the National Science Foundation (NSF-0550752), by the William Vilas Estate Trust, and by the College of Letters and Science and the Graduate School of the University of Wisconsin–Madison. Additional support was provided by the Institute of Social and Economic Research at Osaka University. Research was conducted at the Center for Demography and Ecology and the Center for Demography of Health and Aging at the University of Wisconsin–Madison, which are supported by Center Grants from the National Institute of Child Health and Human Development (R24 HD047873) and the National Institute on Aging (P30 AG17266). The opinions expressed herein are those of the authors. Data from the Wisconsin Longitudinal Study used in this analysis are publicly available at http://www.ssc.wisc.edu/wlsresearch/.

References

  1. AARP. Baby boomers envision retirement II: Survey of baby boomers’ expectations for retirement. Washington, DC: Author; 2004. [Google Scholar]
  2. Burgard SA, Brand JE, House JS. Toward a better estimation of the effect of job loss on health. Journal of Health and Social Behavior. 2007;48:369–384. doi: 10.1177/002214650704800403. [DOI] [PubMed] [Google Scholar]
  3. Burtless G, Quinn JF. Is working longer the answer for an aging workforce? 2002. CRR Issues in Brief No. 11. Chestnut Hill, MA: Center for Retirement Research at Boston College. [Google Scholar]
  4. Chan S, Stevens AH. Job loss and employment patterns of older workers. Journal of Labor Economics. 2001;19:484–521. [Google Scholar]
  5. Chan S, Stevens AH. How does job loss affect the timing of retirement? Contributions to Economic Analysis and Policy. 2004;3:1–24. [Google Scholar]
  6. Crystal S, Shea D. Cumulative advantage, cumulative disadvantage, and inequality among elderly people. The Gerontologist. 1990;30:437–443. doi: 10.1093/geront/30.4.437. [DOI] [PubMed] [Google Scholar]
  7. Dannefer D. Aging as intracohort differentiation: Accentuation, the Matthew effect, and the life course. Sociological Forum. 1987;2:211–236. [Google Scholar]
  8. Farber HS. What do we know about job loss in the United States? Evidence from the Displaced Workers Survey, 1984-2004. 2005. Working Paper No. 498. Princeton, NJ: Princeton University, Industrial Relations Section. [Google Scholar]
  9. Farber HS, Levy H. Recent trends in employer-sponsored health insurance coverage: Are bad jobs getting worse? Journal of Health Economics. 2000;19:93–119. doi: 10.1016/s0167-6296(99)00027-2. [DOI] [PubMed] [Google Scholar]
  10. Friedberg L. The recent trend toward later retirement. 2007. CRR Issues in Brief No. 9. Chestnut Hill, MA: Center for Retirement Research at Boston College. [Google Scholar]
  11. Goldin C. The quiet revolution that transformed women’s employment, education, and family. American Economic Review. 2006;96:1–21. [Google Scholar]
  12. Grzywacz JG, Dooley D. Good jobs to bad jobs: Replicated evidence of an employment continuum from two large surveys. Social Science & Medicine. 2003;56:1749–1760. doi: 10.1016/s0277-9536(02)00170-3. [DOI] [PubMed] [Google Scholar]
  13. Hacker JS. The great risk shift: The new economic insecurity and the decline of the American dream. New York, NY: Oxford University Press; 2008. [Google Scholar]
  14. Hardy MA, Hazelrigg LE, Quadagno J. Ending a career in the auto industry: Thirty and out. New York, NY: Plenum Publishing; 1996. [Google Scholar]
  15. Hauser RM, Warren JR. Socioeconomic indexes for occupations: A review, update, and critique. Sociological Methodology. 1997;27:177–298. [Google Scholar]
  16. Hayward MD. The influence of occupational characteristics on men's early retirement. Social Forces. 1986;64:1032–1045. [Google Scholar]
  17. Hayward MD, Friedman S, Chen H. Career trajectories and older men's retirement. Journal of Gerontology: Social Sciences. 1998;53B:S91–S103. doi: 10.1093/geronb/53b.2.s91. [DOI] [PubMed] [Google Scholar]
  18. Hayward MD, Grady WR, Hardy MA, Sommers D. Occupational influences on retirement, disability, and death. Demography. 1989;26:393–409. [PubMed] [Google Scholar]
  19. Hipple S, Stewart J. Earnings and benefits of workers in alternative work arrangements. Monthly Labor Review. 1996;119:46–54. [Google Scholar]
  20. Kalleberg AL. Nonstandard employment relations: Part-time, temporary and contract work. Annual Review of Sociology. 2000;26:341–365. [Google Scholar]
  21. Kalleberg AL. Precarious work, insecure workers: Employment relations in transition. American Sociological Review. 2009;74:1–22. [Google Scholar]
  22. Kalleberg AL, Reskin BF, Hudson K. Bad jobs in America: Standard and nonstandard employment relations and job quality in the United States. American Sociological Review. 2000;65:256–278. [Google Scholar]
  23. Loveman GW, Tilly C. Good jobs or bad jobs? Evaluating the American job creation experience. International Labour Review. 1988;127:593–611. [Google Scholar]
  24. Merrill Lynch. The 2006 Merrill Lynch new retirement study: A perspective from individuals and employers. 2006. Retrieved from http://www.ml.com/media/66482.pdf. [Google Scholar]
  25. Mishel L, Bernstein J, Allegretto S. The state of working America 2006/2007. Ithaca, NY: Cornell University Press; 2007. [Google Scholar]
  26. Moore JF, Mitchell OS. Projected retirement wealth and saving adequacy. In: Mitchell OS, Hammond PB, Rappaport AM, editors. Forecasting retirement needs and retirement wealth. Philadelphia, PA: University of Pennsylvania Press; 2000. pp. 68–94. [Google Scholar]
  27. Morton L, Foster L, Sedlar J. Managing the mature workforce: Implications and best practices. New York, NY: The Conference Board; 2005. [Google Scholar]
  28. Munnell AH, Golub-Sass F, Perun P, Webb A. Households ‘at risk’: A closer look at the bottom third. 2007 CRR Issues in Brief No. 2007-2. Chestnut Hill, MA: Center for Retirement Research at Boston College. [Google Scholar]
  29. Munnell AH, Perun P. An update on private pensions. 2006 CRR Issue in Brief No. 50. Chestnut Hill, MA: Center for Retirement Research at Boston College. [Google Scholar]
  30. Munnell AH, Sass SA. Working longer: The solution to the retirement income challenge. Washington, DC: Brookings Institution Press; 2008. [Google Scholar]
  31. O’Rand AM. Convergence, institutionalization, and bifurcation: Gender and the pension acquisition process. Annual Review of Gerontology and Geriatrics. 1988;8:132–155. [PubMed] [Google Scholar]
  32. O’Rand AM. The cumulative stratification of the life course. In: Binstock RH, George LK, editors. Handbook of aging and the social sciences. 4th ed. San Diego, CA: Academic Press; 1996a. pp. 188–207. [Google Scholar]
  33. O’Rand AM. The precious and the precocious: Understanding cumulative disadvantage and cumulative advantage over the life course. The Gerontologist. 1996b;36:230–238. doi: 10.1093/geront/36.2.230. [DOI] [PubMed] [Google Scholar]
  34. O’Rand AM, Henretta JC. Delayed career entry, industrial pension structure, and early retirement in a cohort of unmarried women. American Sociological Review. 1982;47:365–373. [Google Scholar]
  35. O’Rand AM, Henretta JC. Age and inequality: Diverse pathways through later life. Boulder, CO: Westview Press; 1999. [Google Scholar]
  36. O’Rand AM, Landerman R. Women's and men's retirement income status: Early family role effects. Research on Aging. 1984;6:25–44. doi: 10.1177/0164027584006001002. [DOI] [PubMed] [Google Scholar]
  37. O’Rand AM, Shuey KM. Gender and the devolution of pension risks in the US. Current Sociology. 2007;55:287–304. [Google Scholar]
  38. Stevens AH. Persistent effects of job displacement: The importance of multiple job losses. Journal of Labor Economics. 1997;15:165–188. [Google Scholar]
  39. Strully KW. Job loss and health in the U.S. labor market. Demography. 2009;46:221–246. doi: 10.1353/dem.0.0050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Szinovacz ME, DeViney S. Marital characteristics and retirement decisions. Research on Aging. 2000;22:470–498. [Google Scholar]
  41. Tilly C. Half a job: Bad and good part-time jobs in a changing labor market. Philadelphia, PA: Temple University Press; 1996. [Google Scholar]
  42. Turner J, Boudreaux M, Lynch V. A preliminary evaluation of health insurance coverage in the 2008 American Community Survey. 2009. Census Bureau Working Papers. Washington, DC: U.S. Census Bureau. Retrieved from http://www.census.gov/hhes/www/hlthins/data/acs/2008/2008ACS_healthins.pdf. [Google Scholar]
  43. Warner DF, Hayward MD, Hardy MA. The retirement life course in America at the dawn of the twenty-first century. Population Research and Policy Review. 2010;29:893–919. doi: 10.1007/s11113-009-9173-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES