Journal of Gerontology: SOCIAL SCIENCES

2002, Vol. 57B, No. 4, S199S208

Copyright 2002 by The Gerontological Society of America

Who Expects to Continue Working After Age 62?

The Retirement Plans of Couples
Amy Mehraban Pienta1 and Mark D. Hayward2
1Institute
2Population

on Aging and Department of Health Policy and Epidemiology, University of Florida, Gainesville.
Research Institute and the Department of Sociology, The Pennsylvania State University, University Park.

Objectives. This study examines the individual, spousal, and household characteristics associated with the retirement
expectations of husbands and wives.

Results. Within a marriage, retirement expectations are shaped by individual, spousal, and household characteristics.
We observe some gender differences in cross-spousal influence with wives retirement expectations being more influenced by husbands resources and constraints than vice versa. Nonetheless, individual and household factors associated
with retirement expectations are widely shared by husbands and wives.
Discussion. Husbands and wives both respond to individual and joint constraints and opportunities when planning
for retirement. Findings support that there is considerable overlap in retirement planning of husbands and wives during
early parts of the retirement decision-making process. However, inequity in cross-spousal influences is a defining characteristic of retirement decision making. Implications for both policy makers and practitioners are briefly discussed.

ETIREMENT is a policy issue that will pervade public

debate for many years to come as the retirement of the
Baby Boom cohort threatens the security of the Social Security trust fund. Traditional policy solutions to preserve fiscal
solvency are most often in the form of financial penalties for
early labor force withdrawal. These penalties have long
been part of the Social Security pension system (Quinn &
Burkhauser, 1994), and they are likely to play a central role
in ensuring solvency in the next several decades. Future cohorts of older workers will soon face ever greater age eligibility criteria for full pension benefits as eligibility for full Social Security benefits gradually increases to age 67 between
2003 and 2007. Additional policy changes are likely to occur,
and issues such as privatizing Social Security and reducing
benefits now occupy center stage in national policy debates.
Projecting the fiscal implications of policies such as these
is typically based on research in which retirement behavior
is construed as an individuals decision. As Baby Boomers
approach the prime retirement ages, however, the retirement
decision is increasingly likely to be a family decisiona decision based on two earnings histories, two sources of pension wealth, and two persons health considerations. The
coupling of retirement decisions within a marriage is reinforced by Social Securitys structuring of husbands and
wives benefits around the earnings history of both persons.
In this study, we gain new insights into married couples retirement decisions by examining how retirement expectations are shaped not only by ones own characteristics, but
also by those of their spouse, and household characteristics
held in common. We also assess how factors influencing retirement expectations differ for husbands and wives, offering new insights into the gendering of retirement decisions

within families. Our results contribute to an emerging line of

inquiry that is mindful of the pairing of retirement through
marriage (Blau, 1998; Gustman & Steinmeier, 1994; Henretta & ORand, 1983; Henretta, ORand, & Chan, 1993;
Hurd, 1990; ORand, Henretta, & Krecker, 1992; Pozzenbon & Mitchell, 1989; Smith & Moen, 1998; Szinovacz,
Ekerdt, & Vinick, 1994).
Retirement Plans of Couples
Our analysis of retirement expectations allows us to focus
on the early part of the retirement process when individuals
are formulating expectations about eventual behavior. This
is important, we believe, because this is a point in the life
course when policy initiatives may have the greatest consequences, and when changes in retirement expectations may
foretell future changes in retirement behavior. Nonetheless,
much of the value of our approach rests on the idea that retirement expectations are reasonable predictors of retirement behavior. Fortunately, the evidence for accurate forecasting is relatively consistent and comes from an array of
studies (see Honig, 1996a, for a review). Where expectations and behavior do not coincide, unanticipated events,
such as changes in pension wealth and health, frequently
prompt retirement earlier than originally planned (Anderson, Burkhauser, & Quinn, 1986), making adaptation to retirement more difficult (Szinovacz, 1989).
A related issue is the extent to which factors influencing
retirement expectations parallel those associated with retirement behavior. Although the body of empirical work is not
dense, available evidence suggests overall congruence in the
structure of factors influencing expectations and behavior
(Honig, 1996a). Like retirement behavior, individuals exS199

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

Methods. Using data from the 1992 Health and Retirement Study, subjective probabilities of working full-time after
reaching age 62 and age 65 are used to measure retirement expectations. The retirement expectations of husbands and
wives are modeled simultaneously using a joint-generalized least-squares approach.

S200

PIENTA AND HAYWARD

The Gendering of Retirement Expectations

Within Marriage
Henretta and colleagues (1993) make a fundamental point
that guides our analysis of retirement expectations within
marriage. Drawing on the work of Treas (1991) and England and Farkas (1986) and others, Henretta and his colleagues emphasize differentiated work and family pathways
as key factors in understanding the retirement behavior of
couples. Building on this literature, we argue that a fundamental part of mens identity is that of family provider (Gradman, 1994), with success benchmarked to the work career
(Gradman, 1994). Furthermore, we expect that the salience
of the work role to mens identity increases with job status
(Gradman, 1994). Under these circumstances, we expect
that mens retirement decisions are closely tied to indicators
of success within the work career and the economic demands of their families. Prior research supports this argument, showing, for example, that men delay retirement
when working in jobs offering significant intrinsic rewards
or when faced with low levels of retirement income and dependent children (Anderson, Clark, & Johnson, 1982; Hayward, Friedman, & Chen, 1998; Henretta & ORand, 1983).

Many women retiring in recent years have balanced both

work and family demands over much of their adult lives.
Substantial numbers of retirement-aged women have interrupted careers because of childrearing demands (ORand et
al., 1992) and the caring of older family members (Hatch &
Thompson, 1992). For these women, interruption of the
work career because of family demands suggests that work
attachment is a less salient aspect of these womens lives
compared with women who are life-long members of the
labor force. Observed retirement patterns of women are
consistent with this interpretationwomen with uninterrupted careers during the childbearing years are more likely
to delay retirement compared with women with interrupted
careers in the childbearing years (ORand et al., 1992; Pienta,
Burr, & Mutchler, 1994). There is less evidence of the effects
of working conditions on womens retirement behavior.
Reinforcing the effects of differential family roles is the
gender inequity in resources derived from the work career.
Mens resources generally are greater than womens, suggesting that husbands plans and resources are more influential for womens plans than vice versa. Research findings,
however, are ambiguous on this point. Studies based on data
from the 1970s suggest the husbands expectations and resources are the dominant force in couples retirement decision making (Anderson et al., 1982; Shaw, 1984). More recent studies (Gustman & Steinmeier, 1994; Honig, 1996b)
point to a shift toward greater equity such that both partners
in a marriage have a strong predisposition toward joint retirement. This is reflected in the fact that the work status of
the husband (or wife) is strongly associated with delayed retirement of a spouse.
Our study brings these issues into empirical focus by
modeling retirement expectations as a function of individual, spousal, and household characteristics that arise from
the work career and family life course. What influences do
the individual and shared life-course experiences and resources of husbands and wives have on both their own and
their spouses retirement expectations? We are also able to
examine the additional implications of household retirement
planning behavior. As partners look toward retirement, we
anticipate that those who have discussed retirement plans
with their spouse have more crystallized expectations.
More than likely, such planful behavior is indicative of a
more general pattern of negotiation that evolves through
the balancing of work and family demands over the course
of a marriage.
METHODS
The 1992 Health and Retirement Study (HRS) is an excellent resource for examining the retirement expectations
of husbands and wives in the years preceding retirement
(Juster & Suzman, 1995). The target population of the HRS
is persons aged 5161 years and their spouses or partners;
spouses or partners are potentially younger or older than
age-eligible respondents. Comparable information is available for both husbands and wives on a variety of topics, including demographic characteristics, health and disability,
current/past job characteristics, retirement plans and expectations, net worth, income, and health insurance.
In this study, we restrict the analytic sample to age-eligible

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

pect to retire sooner when their nonmarket time is more

valuable, health insurance continues after retirement, pension wealth is high, health problems are present, their
spouse is retired, and their job tasks are mundane and repetitive (Ekerdt, DeViney, & Kosloski, 1996; Honig, 1996b;
Hall & Johnson, 1980; Shaw, 1984). Gender differences in
the characteristics associated with retirement expectations
also align with studies of gender differences in retirement
behavior. Like men, women expect to retire sooner if their
health insurance continues after retirement, a private pension is available, and pension wealth is sizeable. Somewhat
less clear are the effects of health status and job conditions
for womens retirement, but it appears that the effects generally parallel those for men (Honig, 1996b). Although there is
considerable evidence about the retirement expectations of individual workers, very few studies have focused explicitly on
retirement expectations as part of a family retirement process.
We take advantage of traditional theoretical frameworks
of retirement behavior to lay the foundation for our model
of retirement expectations within marriage. For example,
we expect that, among dual-earner couples, husbands and
wives formulate retirement expectations in response to their
work environment where they gather information and accumulate resources. Based on available information and resources, workers forecast their future work and nonwork income, pension entitlements, and the quality of life should
they retire. This model would fit actual planning behavior
were it not for the fact the most workers return home after
work and plan for retirement with their spouse. Marriage
provides a context in which the lives, decisions, and choices
of married partners are intertwined (Henretta et al., 1993;
Riley & Riley, 1993). Retirement decisions become joined by
virtue of the balance of family and work role negotiations that
originate from economic and social needs over the family
life course. An important consequence of the intersection of
work and family forces is that retirement expectations may
be gendered within marriagea point we consider next.

MARRIAGE AND RETIREMENT PLANNING

respondents who are married or cohabiting and both partners are in the work force (Chevan, 1996). This makes it
possible to examine the prospective retirement plans of dualearner partners and thus cross-spousal influences of work
characteristics. Regardless of whether a couple is married
or cohabiting, the term spouse is used to denote a partner.
We exclude couples if either spouse is younger than age 45
(n 5 276), because it is less likely that younger respondents have reasonable expectations about retirement timing and will be more prone to give information based on
normative retirement ages. About 8% of couples contain
missing data on the dependent variable for one of the
spouses (n 5 158 couples). The resulting analytic sample
includes 1,818 married and cohabiting couples for which
we have complete data.

Work Conditions and Resources

The HRS contains an array of work-related measures that
define the structural imperatives of work (i.e., factors defining the day-to-day qualities of work activity) that influence
retirement behavior (Hayward et al., 1998; Hayward, Grady,
Hardy, & Sommers, 1989; Hurd & McGarry, 1993; Spenner, 1988). Two measures of job demands (physical and
cognitive) are derived from a factor analysis of eight items
assessing current job requirements. Using the Promax rotation method, an exploratory factor analysis supports a two-

factor structure (eigenvalues: 2.45 and 1.98): physical demands and cognitive demands. Jobs that are physically
demanding involve physical effort, lifting heavy loads, and
bending over. Jobs that are cognitively demanding involve
good eyesight, intense concentration (or attention), analyzing data (or information), keeping a pace set by others, and
learning new things. Physical demands factor loadings
range from .85 (job requires stooping or bending) to .87
( job requires lifting heavy objects) explaining 21.6% of the
variance (21.4% variance is explained when the other factor
is eliminated). Cognitive demands factor loadings range
from .57 (job requires good eyesight) to .72 ( job requires
concentration) explaining 18.6% of the variance (18.5%
variance is explained when the other factor is eliminated).
The interfactor correlation is 2.076. Average scores for
each set of items are calculated with Cronbachs alphas of
.84 and .64 for physical and cognitive demands, respectively. Higher scores reflect greater job demands.
The retirement-related extrinsic rewards of work are measured in terms of pension eligibility status (no coverage
available; has pension coverage, but not yet eligible; and
currently receiving or eligible to receive benefits), private
pension wealth, and health insurance coverage. Pension
wealth is its present value at time of the 1992 interview.
Pension wealth is a constructed measure obtained from the
HRS web site (see www.umich.edu/~hrswww/center/rescont2.
html for a complete description of this measure). The
measure is based on self-report information and potentially reflects the promise or receipt of employer-provided
pension benefits from up to three current or prior jobs.
Health insurance references whether the respondent is uninsured, has employer-provided insurance either through
ones own employer or spouses employer, or has health
insurance from a nonemployer source (including privately
purchased and government-provided health insurance,
such as Medicaid).
Self-employment establishes whether retirement plans
are tied to ones position in a bureaucratic organization
(Ekerdt, Hackney, Kosloski, & DeViney, 2001; Hayward et
al., 1998). This is important because self-employed workers
encounter less social regulation of work careers (Ekerdt et
al., 2001). Also, self-employed men and women may work
longer because self-employed workers earn less than their
nonself-employed counterparts (Hamilton, 2000). Selfemployment is attractive to many workers because it includes many nonpecuniary benefits, such as greater autonomy, flexibility, intrinsic job satisfaction, and self-expression
(Carr, 1996; Hamilton, 2000). For all of these reasons, selfemployment is expected to increase expectations for delayed retirement. This is consistent with a recent study
showing that self-employed workers have greater uncertainty about retirement timing (Ekerdt et al., 2001). We also
control for part-time work (less than 35 hours/week) because part-time workers are likely to have lower labor market attachment than full-time workers.
Health Characteristics
All else being equal, health is a primary motive for retirement with those in the poorest health either retiring or becoming work disabled. In part, workers in poor health may

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

Retirement Expectations
We measure retirement expectations based on the following question: Thinking about work generally and not just
your present job, what do you think are the chances that you
will be working full time after you reach age 62? The response set is a scale ranging from 0 to 10, in which 0 equals
absolutely no chance and 10 equals absolutely certain.
We transform these values into subjective probabilities by
dividing each value by 10, thus a respondents probability
ranges from 0 to 1 (see Honig, 1996b, for an example of this
strategy). A value of one can be interpreted as having a high
expectation of continued work after reaching age 62, conditional on working currently. Values closer to zero represent
stronger expectations of early retirement. We also construct
a measure of delayed retirement expectations, i.e., expectations of working full time past age 65, conditional on working currently and having some expectation of working past
age 62.
Not surprisingly, retirement expectations are not normally distributedthe probabilities cluster at the extremes
of the distribution. We thus transform the subjective probabilities into a cumulative logistic function (Pindyck & Rubinfeld, 1976). Using the subjective probability of working
full time after age 62 for illustration, the transformation is
ln(P62/1 2 P62). The slope of the cumulative logistic distribution is greatest at P62 5 1/2, which implies that changes in
independent variables will have their greatest impact on the
probability of tending to be more neutral in plans for retirement. The low slopes near the endpoints of the distribution
require larger changes in independent variables to bring
about a small change in probability. This transformation allows us to use joint generalized least-squares (GLS) estimation, which is described below.

S201

S202

PIENTA AND HAYWARD

Household Planning Behavior

The HRS includes a measure of prior retirement planning
activities of couples. Respondents are asked how frequently
they have discussed retirement plans with their spouse. The
range of responses is hardly at all (scored as 1) to a lot
(scored as 4). Missing information is randomly assigned a
value based on the variables distribution. Although we anticipate that household planning behavior influences retirement expectations, it is also possible that an individuals retirement expectations also affect household planning. Thus,
we caution against interpreting the association between retirement planning behavior and retirement expectations in
strong causal terms, and we think it likely reflects the fact
that the effect is overestimated.
Household Family Characteristics
Family characteristics define the immediate and past
constraints and opportunity costs of both partners. Couples
share childrearing responsibilities, for example, although
there may be a division of tasks for husbands and wives. A
particularly salient issue, from a retirement planning perspective, is the presence of dependent children in the household. Couples who have chosen either to delay childbearing
or to postpone the end of their childbearing activities are
likely to enter their retirement years with children living at
home (or temporarily away at school).
We also examine the effects of the overall level of income
and wealth within the household. The household income of
dual-earner couples reflects both the resources immediately
available, as well as the opportunity costs associated with

early retirement. Net worth is a measure of assets less the

debts of a household.
Demographic Characteristics
We construct measures for various sociodemographic
factors that differentiate retirement experiences in the population. These measures include age, education (years of
completed schooling), and race/ethnicity (White, African
American, Hispanic, and other non-Whites [including Asian,
Pacific Islander, American Indian, or Alaskan native]). African American men typically exit the labor force at higher
rates than White men (Burr, Massagli, Mutchler, & Pienta,
1996; Hayward, Friedman, & Chen, 1996), whereas African
American women work longer than White women (Pienta et
al., 1994). Hispanic men have lower rates of labor force exiting than African American and White men. Hispanic
women have higher rates of exiting than other women (Flippen & Tienda, 2000).
Spouse Characteristics
The joint nature of husbands and wives retirement is
also captured by assessing the effects of the spouses expectations on the respondents own expectations, and the effects
of several other relevant spousal attributes. Because the
spouses expectation is potentially endogenous in our model
(i.e., the individuals and spouses expectations are simultaneously determined), we used an instrumental variables
approach to reduce the endogeneity problem. Our basic approach was one in which we attempted to select as many
predictors strongly associated with the spouses expectations, but not strongly associated with the individuals expectations, as possible (Heckman, 1997). In this analysis,
we regressed the spouses expectations on their age, education, self-reported health, pension wealth, expectations of
living to ages 7592, and reasoning ability (results available
from the authors), separately for expectations of early and
delayed retirement. Based on the estimated parameters, we
calculated a predicted score for the spouses expectations
and included this as our measure of spouses expectations in
the modeling of a persons retirement expectations.
Other key spouse characteristics include age, health, and
pension wealth. Spouses age, net of ones own age, captures the tendency to conform to a spouses age-related retirement schedule. When a spouse is older, we expect this to
increase the probability that a person will expect to retire
early. The effect of spousal health is ambiguous. A spouse in
poor health may increase the attractiveness of joint leisure, or it may have the opposite effect of requiring a
worker to remain in the labor force to meet the financial
needs of the couple. Finally, we consider whether access
to a spouses pension benefits increases early retirement
expectations.
Analytic Model
In estimating a retirement expectations model, we adopt
an analytic approach that allows us to gauge cross-spousal
influences allowing for the correlation of husbands and
wives unmeasured retirement preferences (Gustman &
Steinmeier, 1994). Another aspect of this problem is that
spouses may have retirement preferences that are unmea-

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

attach higher value to retirement than work. A related issue

is that health problems may make it more difficult to keep
up with the daily demands of work. Most retirement research examining the effects of health relies on a work limitations measure (i.e., Does the respondent report a health
condition that limits the amount or type of work they can
do?) Like prior studies, we make use of the health measure
evaluating the presence of a health condition that limits
work ability (1 5 limited in the amount or kind of work;
0 5 otherwise).
We expand the measurement of health to include serious
chronic disease conditions that affect not only functional
status, but also a persons motivation to work (Verbrugge &
Jette, 1994; Wolinsky, Armbrecht, & Wyrwich, 2000). We
have chosen to focus on serious chronic health conditions
(Ferraro & Farmer, 1999) for the following reasons. First,
serious chronic illnesses may either increase the value of
leisure relative to work and contribute to early retirement
expectations or shorten ones time horizon for achieving
ones goals. Conversely, delayed retirement expectations
may result when a person attempts to manage a serious
chronic illness, thereby increasing the need for health care
insurance or income. Serious chronic conditions are measured using self-report of whether a doctor has ever told the
respondent that he/she has hypertension, coronary heart
disease, stroke, cancer (nonskin), diabetes, or chronic lung
disease. For simplicity, we create a measure in which each
condition is given equal weight (1 5 condition present, 0 5
not present) and summed together to form a single measure.

MARRIAGE AND RETIREMENT PLANNING

Ph
Ln 1------------- P h = h + h Work ih + h Health jh + h Plan h
+ h Family kh + h Demog lh + h Spouse mh + h .
Pw
Ln --------------1 P w = w + h Work iw + w Health jw + w Plan w
+ w Family kw + w Demog lw + w Spouse mw + w .
Ph represents the husbands subjective probability of working full time after age 62 (or age 65), and Pw is the wifes
subjective probability. Recall that Ph and Pw are transformed
using a cumulative logistic distribution function so that ordinary least squares (OLS) can be used for estimation. Each
spouses expectation is presumed to be a function of a vector of i work characteristics, j health characteristics, k
family characteristics, and so on. Note that the model includes a vector of m spouse characteristics described
previously.
The method allows the disturbances (h and w) to be correlated across the two equations, because there is reason to
suspect that this is the case. The estimator for this set of
equations is obtained in two steps. In the first step, a crossequation covariance matrix is computed from residuals from
the two separate equations estimated with OLS. This matrix
is then used to weight the final estimates retirement expectations. The SURE model is estimated using PROC MODEL
in SAS. The SAS procedure produces a system-weighted R2
computed using the cross-equation correlation matrix.
In the analysis that follows, nested models are estimated
in which the base model includes only individual characteristics. A second model introduces spousal and couple-level
characteristics (spouses plans, spouses characteristics, and
household constraints/resources). The nested models allow
us to assess how the effects of person-specific retirement
predictors are altered when taking into account spousal resources and expectations. The full model is the basis on
which we evaluate the direct effects of spousal and household influences on a partners expectations. Because we estimate models for both husbands and wives, we are able to

compare whether factors influencing retirement expectations are dependent on gender. We use an F test in the two
nested models of husbands and wives to formally test for the
equivalence of effects. Values in the tables are underlined to
indicate statistically different predictors of husbands and
wives plans (significant at the .05 level and adjusted for
shared covariance across the models), providing insights
into the gendering of retirement expectations among married couples.
Two sets of nested models are estimated: one for early retirement (i.e., the subjective probability of working full time
after age 62) and the other for delayed retirement (the subjective probability of working full time past age 65). The
analysis of delayed retirement expectations is based on a
subset of the sample who had at least some expectation of
working past age 62. In other words, we analyze couples
that attach some probability to working past age 62 (Ph and
Pw are both greater than zero). For couples thinking about
working past age 62, the probability of working past age 65
is regressed on comparable independent variables. Again,
positive coefficients indicate stronger plans for delaying retirement past age 65. Finally, the system-weighted R2 (joint
GLS) is compared with the adjusted R2 (OLS) to evaluate
whether we improve upon the individual estimation of retirement expectations.
Selection
Because couples are excluded when at least one spouse
has already left the labor force, we estimated a selection
model to evaluate whether the selection of dual-earner couples biases our results. For example, it is possible that more
traditional couples or couples that are more age heterogamous are excluded in our models. Because the potential exists for introducing selectivity into our analysis, we estimated a two-stage selection model (Berk, 1983; Heckman,
1979). In the first stage, a number of independent variables
were chosen that were related to sample inclusion, but not
the dependent variable (retirement expectations). The full
set of predictor variables included the following: both partners ages and education, the race/ethnicity of the household
head, number of offspring, number of children living in the
household or at school, and equality in decision making.
The results showed that excluded couples differ from those
in our analysis mainly in that the husband tends to be older,
the wife tends to be younger, the wife tends to be less educated, the couple has a larger number of children, and the
couple does not engage in egalitarian decision making. The
second-stage model (see results in Tables 2 and 3) included
a selection correction term calculated from the parameter
estimates of the first model. Although our results remain the
same whether we correct for selection or not, we present
the results correcting for sample selection.
RESULTS
Included in Table 1 are descriptive characteristics for
HRS husbands and wives. Although cohabiters tend to be
younger and have less education than those who are married
(results not shown), retirement expectations do not differ
between cohabiting and married persons. As such, we do
not distinguish between married and cohabiting persons. We

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

sured in our model, but that are highly correlated. Gustman

and Steinmeier (1994) give an example of correlated retirement preferences arising from assortative mating (e.g., men
and women with similar preferences for early retirement
may marry).
Because it is not possible to include all theoretically relevant information about husbands and wives retirement
preferences, our modeling approach adjusts the error structure of both spouses by taking into account a correlated error structure in the husbands and wives expectations
models. We do this by estimating husbands and wives retirement expectations with a joint GLS modelalso referred
to as a seemingly unrelated regression (SURE) model
(Zellner, 1962). An example of an application of the SURE
model is a study by Simpson and Wilson (1987), who attempted to capture the correlation between spouses plans
within dual-earner couples from unobserved heterogeneity.
The joint GLS model takes the following form based on
the conceptualization of retirement expectations:

S203

S204

PIENTA AND HAYWARD

Table 1. Descriptive Characteristics of Husbands and Wives Aged

4561 (N 5 1,818 Couples)
Characteristic
Demographic
Age (mean)
White (%)
African American (%)
Latino (%)
Other (%)
Education (mean years)
Serious chronic condition (mean no.)
Work-disabled (%)

Retirement Plans
Expectation work after 62
Expectation work after 65
Household
Discussed retired spouse
Household income (mean $)
Household net worth (mean $)
Children home/school (%)

Wives

55.0
87.0
6.9
4.2
1.9
13.0
.7
10.9

52.2
87.2
6.6
4.4
1.8
13.0
.5
8.6

9.2
2.1
2.8
22.9
24.6
24.9
50.5
86,919
10.6
80.2
9.1

32.4
1.9
2.9
16.5
40.2
7.5
52.3
33,498
12.6
79.8
7.6
.36
.29

.51
.38
2.5

2.4
70,359
29,905
44.4

Notes: Values are calculations of means and frequencies of weighted data from
the Health and Retirement Study (1992). T test or x2 statistic was used to determine statistically significant differences across husbands and wives (underlined).

also generated a parallel set of results that excluded cohabiters, but these did not differ substantially from the results
presented later.
As one might expect, wives are younger than husbands,
but do not differ in educational attainment. Wives are generally in better health than husbandswives report fewer fatal
chronic conditions and less work disability. Life cycle work
experiences of husbands and wives are uniformly different.
Husbands tend to work in jobs that are more physically demanding and less cognitively demanding than wives jobs; husbands also tend to be self-employed. Wives tend to work part
time. Pension characteristics differ with men having greater
pension wealth and access to employer-provided pensions.
Husbands attach a much higher probability to working
full time after age 62 than do wives (Ph 5 .51 compared
with Pw 5 .36). Husbands are also more likely to expect to
continue working full time after age 65 than are wives (Ph 5
.38, whereas Pw 5 .29). Both findings reinforce the idea that
women maintain a somewhat weaker attachment to the
labor forcein part this may be because wives in these birth
cohorts have had substantially less work experience throughout their lives. These findings also demonstrate the enormous popularity of early retirement given that neither husbands or wives report strong probabilities of working full
time after age 62 or age 65. Husbands and wives expecta-

Early Retirement Expectations

Early retirement expectations of husbands and wives are
explored in greater detail in Table 2 to determine similarities
and differences in the factors associated with husbands and
wives early retirement expectations. The results for Models
1 and 2 show that, although there is considerable similarity
across spouses, age and health influence the early retirement
plans of husbands and wives differently. Age has a much
smaller impact on wives early retirement expectations than
it does on husbands expectations (.053 and .259, respectively, in Model 2). Both husbands and wives who are older
are more likely to expect to work full time after reaching
age 62, but this effect is statistically significant for men and
not women. The overall age pattern may reflect a combination of factors, such as age selectivity (i.e., who is still in the
labor force when queried about expectations), more accurate forecasts among older workers, and potential cohort
differences.
We expected chronic health conditions and work disability to be positively associated with the early retirement expectations of both husbands and wives based on the idea
that declining health signals a shorter work career. Our results for husbands are consistent with this expectation. Husbands who have been diagnosed with a serious chronic condition have weaker expectations of working past age 62.
Wives in poorer health, however measured, do not differ
significantly from wives in good health in their retirement
expectations. Health is surprisingly not associated with
wives expectations.
Attributes of work have equivalent effects on husbands
and wives early retirement expectations. Our results show
that men and women expect to work longer in jobs with
greater cognitive demands and jobs in which workers are
self-employed. Greater pension wealth is negatively associated with husbands and wives expectations of working full
time after age 62. As shown in Model 1, the absence of
health insurance increases the probability of husbands and
wives early retirement, although this effect is reduced
and no longer significant once the spouses characteristics
are included as control variables.
Spouse and household characteristics provide additional
insight into the retirement plans of couples (Model 2, Table
2). Husbands work disabilities, pension eligibility, and pension wealth all decrease wives expectations of working
after age 62 (see Model 2, Table 2). Husbands retirement
expectations, however, are not influenced by wives health
and pension characteristics. Although the pattern of these
spousal effects suggests that husbands play a more significant role in womens retirement decisions than vice versa,
we note that the differences are not statistically significant.
Discussing retirement plans with a spouse is associated with
lower expectations of working full time after age 62 for both
husbands and wives, but particularly so for husbands. Although potential endogeneity problems make it difficult to
make strong inferences about this association, it strongly

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

Work
Part time (%)
Physical demands (mean)
Cognitive demands (mean)
Self-employed (%)
No private pension (%)
Eligible/receiving pension (%)
Covered pension (%)
Pension wealth (mean $)
Uninsured (%)
Employer health insurance (%)
Other health insurance (%)

Husbands

tions about the timing of retirement are interrelated, with

correlation coefficients of .28 and .23 (expectations about
retirement upon reaching ages 62 and 65, respectively) that
are statistically significant (results not shown).

MARRIAGE AND RETIREMENT PLANNING

S205

Table 2. Joint-Generalized Least-Squares Analysis of Husbands and Wives Expectations of Full-Time Work After Age 62
Model 1

Characteristic

Husbands
(N 5 1,818)
b (SE)

Model 2

Wives
(N 5 1,818)
b (SE)

F Value

Husbands
(N 5 1,818)
b (SE)

Wives
(N 5 1,818)
b (SE)

F Value

0.116 (.064)*
20.932 (.429)**
0.217 (.569)
20.549 (.807)
0.157 (.052)***
20.718 (.140)***
20.988 (.342)***

0.063 (.028)**
20.508 (.419)
0.371 (.509)
0.773 (.776)
0.089 (.058)
0.060 (.145)
20.115 (.357)

0.61
0.61
0.05
1.60
0.77
15.04
3.13

0.259 (.084)***
21.395 (.423)***
0.009 (.567)
20.808 (.805)
0.179 (.052)***
20.657 (.139)***
20.963 (.343)***

0.053 (.039)
20.943 (.412)**
0.182 (.520)
0.488 (.822)
0.067 (.110)**
0.005 (.145)
20.112 (.357)

4.72
0.69
0.06
1.39
0.84
10.76
2.95

Work
Part time
Physical demands
Cognitive demands
Self-employed
Covered pension
Eligible/received pension
Pension wealth
Uninsured
Other health insurance

22.238 (.384)***
20.051 (.132)
0.626 (.187)***
1.455 (.297)***
20.503 (.307)
20.558 (.352)
20.543 (.085)***
0.505 (.359)
0.634 (.400)

22.231 (.233)***
20.178 (.127)
0.483 (.172)***
1.644 (.300)***
20.128 (.250)
20.284 (.429)
20.620 (.134)***
0.660 (.311)**
20.023 (.396)

0.00
0.49
0.32
0.21
0.91
0.25
0.24
0.11
1.55

22.354 (.376)***
20.035 (.130)
0.636 (.184)***
1.504 (.299)***
20.312 (.306)
20.302 (.352)
20.513 (.086)***
0.478 (.353)
0.610 (.393)

22.148 (.231)***
20.236 (.126)*
0.506 (.169)***
1.683 (.297)***
20.064 (.249)
20.322 (.428)
20.471 (.135)***
0.382 (.309)
20.279 (.400)

0.22
1.26
0.28
0.19
0.39
0.00
0.07
0.04
2.81

Spouse
Expects to retire by 62
Age
No. of serious chronic conditions
Work-disabled
Covered by pension
Eligible/receiving pension
Pension wealth

20.266 (.344)
20.028 (.035)
20.026 (.153)
0.044 (.374)
0.024 (.239)
20.587 (.440)
20.096 (.233)

20.040 (.106)
0.140 (.127)
0.152 (.138)
20.720 (.329)**
20.269 (.270)
20.665 (.322)**
20.321 (.116)**

0.40
1.58
0.75
2.34
0.65
0.02
0.75

Household
Discussed retired spouse
Income
Net worth
Children home/school

20.869 (.095)***
0.073 (.101)
20.118 (.042)***
0.506 (.223)**

20.588 (.088)***
20.007 (.096)
20.121 (.039)***
0.079 (.216)

5.14
0.39
0.00
2.25

Selection correction
Intercept

1.494 (1.622)

System-weighted R2
OLS adjusted R2

20.803 (2.011)

0.551 (0.872)

29.598 (2.786) ***

213.232 (3.592)***

0.08

0.17

0.10
0.12

22.118 (3.111)

26.962 (1.711)***

28.908 (3.496)**

0.15
0.12

Notes: Values are calculations of weighted data from the Health and Retirement Study (1992). Underlined values represent differences across the husbands and
wives models that are statistically significant ( p , .05). OLS 5 ordinary least squares.
*p , .10; **p , .05; *** p , .01.

suggests that retirement decisions are the outcome of household planning involving both the husband and the wife.
Both husbands and wives living in wealthier households
are less likely to plan to continue working after age 62
(Model 2, Table 2). Financial constraints, particularly those
defined by household wealth, clearly limit choices about the
timing of retirement. Household wealth allows both partners
in a marriage greater freedom to choose leisure over work.
Another economic challenge faced by many retirement-age
couples is providing for dependent children in the household. Our results suggest that husbands, but not wives, plan
to postpone retirement when dependent children live in the
homea possible product of mens greater earnings.
Delayed Retirement Expectations
The availability of full Social Security benefits at age 65
raises the possibility that the mix of factors influencing de-

layed retirement differs from that influencing retirement expectations under conditions of reduced benefits. To evaluate
this question, we modeled expectations of working full time
after age 65 among persons who, at a minimum, are uncertain about early retirement. In other words, this model includes couples where both partners report the probability of
working full time after age 62 is greater than zero.
Similar to our results for early retirement, the age gradient is steeper for men, and health has little effect on
womens expectations. More years of education increase expectations of delayed retirement for both husbands and
wives (see Model 1, Table 3), although the education effect
is attenuated for wives once spouse characteristics are controlled for.
Work characteristics are related to delayed retirement
expectations. Cognitive job demands significantly increase
the delayed retirement expectations, although no such ef-

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

Demographic and Health

Age
African American
Latino
Other
Education
No. of serious chronic conditions
Work-disabled

S206

PIENTA AND HAYWARD

Table 3. Joint-Generalized Least-Squares Analysis of Husbands and Wives Expectations of Full-Time Work After Age 65
Model 1

F Value

Husbands
(n 5 936)
b (SE)

12.95
0.44
0.15
0.04
0.03
0.80
0.12

0.254 (.093)***
0.134 (.602)
20.125 (.732)
0.399 (.864)
0.154 (.066)**
0.237 (.176)
20.470 (.442)

20.045 (.047)
0.587 (.587)
0.143 (.607)
0.181 (.909)
0.144 (.136)
20.061 (.184)
20.266 (.435)

7.94
0.34
0.09
0.03
0.00
1.36
0.11

3.23
0.26
3.92
2.01
0.14
0.22
2.36
2.43
3.97

0.193 (.497)
0.233 (.159)
0.421 (.227)*
1.815 (.335)***
20.477 (.352)
20.495 (.427)
20.155 (.123)
1.279 (.407)***
0.896 (.449)**

20.779 (.288)***
0.310 (.158)*
20.082 (.207)
1.207 (.338)***
20.771 (.300)**
21.050 (.552)*
20.449 (.173)***
0.229 (.360)
20.411 (.447)

2.87
0.12
2.74
1.69
0.40
0.63
1.87
3.88
4.75

Spouse
Expects to retire by 62
Age
No. of serious chronic conditions
Work-disabled
Covered by pension
Eligible/receiving pension
Pension wealth

0.218 (.275)
0.024 (.047)
20.047 (.196)
0.465 (.463)
20.150 (.287)
20.297 (.576)
0.119 (.249)

0.324 (.273)
20.080 (.167)
20.031 (.165)
21.067 (.414)**
20.308 (.306)
20.167 (.388)
0.151 (.173)

0.08
0.35
0.00
6.04
0.14
0.03
0.01

Household
Discussed retired spouse
Income
Net worth
Children home/school

20.526 (.118)***
0.124 (.134)
20.035 (.044)
0.407 (.269)

20.377 (.109)***
20.114 (.125)
20.087 (.041)**
0.058 (.256)

0.95
1.98
0.85
1.03

22.574 (2.153)

21.718 (3.911)

Characteristics

Husbands
(n 5 936)
b (SE)

Model 2

Wives
(n 5 936)
b (SE)

0.233 (.076)***
0.237 (.597)
20.065 (.719)
0.738 (.857)
0.129 (.064)**
0.226 (.174)
20.274 (.434)

20.055 (.033)
0.747 (.591)
0.279 (.588)
0.527 (.798)
0.112 (.067)*
0.003 (.181)
20.488 (.430)

Work
Part time
Physical demands
Cognitive demands
Self-employed
Covered pension
Eligible/receiving pension
Pension wealth
Uninsured
Other health insurance

0.223 (.496)
0.228 (.159)
0.492 (.225)**
1.805 (.324)***
20.619 (.346)*
20.700 (.418)*
20.191 (.119)
1.238 (.408)***
0.903 (.448)**

20.805 (284)***
0.339 (.157)**
20.108 (.207)
1.158 (.336)***
20.791 (.296)***
21.019 (.543)*
20.502 (.168)***
0.411 (.358)
20.265 (.434)

Selection correction
Intercept

22.579 (1.927)

System-weighted R2
OLS adjusted R2

0.511 (1.032)

216.330 (3.310)***

20.734 (2.040)

218.473 (3.550)***

0.10
0.12

F Value

5.662 (5.168)

0.13
0.06

0.14

0.08

Notes: Values are calculations of weighted data from the Health and Retirement Study (1992). Underlined values represent differences across the husbands and
wives models that are statistically significant ( p , .05). OLS 5 ordinary least squares.
*p , .10; **p , .05; ***p , .01.

fects are observed for wives. The effect of cognitive demands for men is consistent with prior research documenting lower rates of retirement among male workers in these
positions (Hayward et al., 1998). Physical job demands, on
the other hand, are positively associated with expectations
of wives working full time after age 65. The positive effect of physical job demands on womens expectations is
contrary to what one would expect under the wear-andtear effects of a demanding work environment. Other research, however, indicates that physical demands in late
life jobs have positive benefits for mens health, after controlling for selection into these jobs (Moore & Hayward,
1990). As such, physical demands may actually increase
the relative value of work over leisure. Self-employed husbands and wives, net of all other factors, are more likely
to anticipate delayed retirement. Because we control for
health, and pension and asset wealth, we suggest that this
effect reflects self-employed individuals flexibility to ad-

just their work behavior in accordance with their health or

preferences for work.
Wives pension wealth significantly reduces expectations
to delay retirement, although no comparable effect is observed for husbands (but this is not statistically significant).
Among husbands, the absence of health insurance is positively associated with expectations for delayed retirement as
is nonemployer health insurance coverage. Husbands with
employer-provided health insurance are the least likely to
expect to delay retirement. Unlike our observations of early
retirement expectations, health insurance does not significantly shape wives delayed retirement expectations, and
the difference between husbands and wives is statistically
significant.
Like early retirement expectations, delayed retirement expectations are related to retirement planning with spouse.
Household planning behavior is associated with expectations
of retiring earlier. Again, the husbands work disability sta-

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

Demographic and Health

Age
African American
Latino
Other
Education
No. of serious chronic conditions
Work-disabled

Wives
(n 5 936)
b (SE)

MARRIAGE AND RETIREMENT PLANNING

DISCUSSION
Like previous research, our analysis demonstrates that
husbands retirement expectations are shaped by pension
wealth, health, and characteristics of the work environment
relationships reflecting the institutionalization of retirement
(Szinovacz & DeViney, 1999). Although the factors shaping
wives expectations are less well known, like husbands,
womens plans respond to working conditions and pension
wealth. Such equivalent pension wealth effects, however,
should be considered alongside studies that report on aggregate gender differences in pension wealth. In our study,
wives have accumulated pension wealth that is one-third the
value of the pension wealth of husbands. Hardy and Shuey
(2000) noted persistent structural differences in women and
mens employer-provided pensions that place women at a
disadvantage. Our results suggest that marriage may provide
an important financial safety net that heightens the importance of wives pension wealth in their retirement decision
making, even though their levels of pension wealth lag behind those for men.
Consistent with prior research, we find an association between early retirement and workers labor market vulnerabilities. Vulnerabilities defined by lower education and being
African American weaken connections to work. In contrast to
Flippen and Tienda (2000), our results indicate that Hispanics retirement expectations do not differ from those of
Whites. One possible explanation is that the present study
selects only married couples and investigates retirement expectations rather than retirement behavior.
Characteristics of the spouse and household are important
predictors of retirement expectations. In terms of household
factors, husbands and wives expectations of early retirement increase as household wealth increases. Husbands,
though not wives, expect to delay retirement if dependent
children reside in the household. Also, household planning
behaviors tend to accompany spouses plans to retire early.
This finding suggests that planful behavior in household decision making is important for the joint maximization of leisure among husbands and wives. We observe asymmetry,
however, in the effects of the spouses health and pension
characteristics on their partners expectations. Although
most gender differences in spousal effects do not differ sta-

tistically, the pattern of effects points to a stronger influence

of husbands on wives retirement expectations than vice
versa. Inequity remains a defining feature of retirement decision making within marriage.
An important gender difference to emerge from our analysis is the extent to which a persons health influences their
retirement expectations. Unlike husbands, wives expectations do not appear to be sensitive to health. Given the preeminence of health effects in the retirement literature, the
absence of such effects for wives was unexpected. One possible explanation is that health problems are less severe
among wives of the ages considered here. On the other hand,
wives are more likely to expect early retirement when their
husbands have a work disability. Wives appear to take into
account their spouses health rather than their own when
formulating retirement expectations.
Increasingly, economists and sociologists point to the importance of investigating retirement as a household decision (e.g.,
Anderson et al., 1982; Blau, 1998; Gustman & Steinmeier,
1994; Henretta et al., 1993). Despite growing recognition of the
theoretical importance of household decision-making perspective, much of the research on retirement behavior continues
to focus on individual resources. The results shown here,
however, demonstrate that retirement expectations are not
simply borne out of an individuals own achievements in the
labor market, but they very much reflect the resources and
expectations of their spouse (particularly for women),
household factors such as asset wealth, and household planning behavior.
Considerable theoretical and empirical research reinforces the idea that husbands and wives invest in marriage.
Forms of family organization develop early in marriages out
of negotiations involving work/family activities. Because
lifelong family organizational strategies maintain and enhance family well-being, it should not be surprising that retirement decisions are intimately tied to planning behavior,
household factors, and the spouses expectations and resources. This conclusion speaks to institutional factors that
reinforce retirement as a family decision. Social Security
benefits in particular are structured not on individuals earnings histories, but on the earnings history of the husband
and wife within a family. As policy makers plan to address
concerns surrounding a growing pool of older workers, it is
important to be mindful of the pairing of retirement plans
and retirement behavior (Henretta & ORand, 1983) that occurs through marriage. Very often, it is neither his retirement nor her retirement but their retirement that alters the
landscape of pension beneficiaries and the timing of retirement in the population.
Acknowledgments
Partial support for this research has been provided by the AARP Andrus
Foundation, the National Institute on Aging (R03 AG16127), and the National Institute of Child Health and Human Development (5 P30 HD28263).
We thank Angela ORand and Diane McLaughlin for insightful comments on an earlier draft of this manuscript, and Kristi Rahrig Jenkins for
her research assistance.
Address correspondence to Amy Pienta, Institute on Aging, P.O. Box
103505, University of Florida, Gainesville, FL 32610-3505. E-mail:
pienta@ufl.edu

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

tus reduces his wifes expectations of working after age 65.

Whether the wifes response to her husbands health needs
reflects the need for caregiving is not clear in our analysis.
When comparing the model fit of the joint GLS estimation (system-weighted R2) with OLS estimation (adjusted
R2), we find that joint GLS may be a somewhat more efficient estimator. For example, in the fully specified models
in Table 2 (Model 2), only 17% (husbands) and 12%
(wives) of the variance in retirement expectations was explained using OLS, whereas 15% of the variance was
explained with joint GLS estimation. Individually estimated models explained a smaller portion of the variance in
wives retirement expectations at age 65 than joint models,
as well. This suggests that our models have tapped into important couple-level dimensions of the planning process,
in particular improving the explanatory model for wives in
dual-earner couples.

S207

S208

PIENTA AND HAYWARD

References

ties (HRS/AHEAD Working Paper No. 96036). Ann Arbor, MI: Population Studies Center.
Honig, M. (1996b). Retirement expectations: Differences by race, ethnicity, and gender. The Gerontologist, 36, 373382.
Hurd, M. D. (1990). The joint retirement decisions of husbands and wives.
In D. A. Wise (Ed.), Issues in the economics of aging (pp. 231254).
Chicago: University of Chicago Press for the National Bureau of Economic Research.
Hurd, M., & McGarry, K. (1993). The relationship between job characteristics and retirement (National Bureau of Economic Research
Working Paper No. 4558). Cambridge, MA: National Bureau of Economic Research.
Juster, F. T., & Suzman, R. (1995). An overview of the Health and Retirement Study. The Journal of Human Resources, 30, S7S56.
Moore, D. E., & Hayward, M. D. (1990). Occupational careers and mortality of elderly men. Demography, 27, 3153.
ORand, A. M., Henretta, J. C., & Krecker, M. (1992). Family pathways to
retirement. In M. Szinovacz, D. Ekerdt, & B. Vinick (Eds.), Families
and retirement (pp. 8198). Newbury Park, CA: Sage Publications.
Pienta, A. M., Burr, J. A., & Mutchler, J. E. (1994). Womens labor force
participation in later life: The effects of early work and family experiences. Journal of Gerontology: Social Sciences, 49, S231S239.
Pindyck, R. S., & Rubinfeld, D. L. (1976). Econometric models and economic forecasts. New York: McGraw-Hill.
Pozzenbon, S., & Mitchell, O. S. (1989). Married womens retirement behavior. Journal of Population Economics, 2, 3953.
Quinn, J. F., & Burkhauser, R.V. (1994). Retirement and labor force behavior of the elderly. In S. H. Preston & L. G. Martin (Eds.), Demography of aging (pp. 50101). Washington, DC: National Academy
Press.
Riley, M. W., & Riley, J. (1993). Connections: Kin & cohort. In V. L.
Bengtson & W. A. Achenbaum (Eds.), The changing contract across
the generations (pp. 169189). New York: Aldine De Gruyter.
Shaw, L. (1984). Retirement plans of middle-aged married women. The
Gerontologist, 24, 154159.
Simpson, I. H., & Wilson, J. (1987). Church activism among farm couples:
Measuring the impact of the conjugal unit. Journal of Marriage and the
Family, 49, 875882.
Smith, D. B., & Moen, P. (1998). Spousal influence on retirement: His, her,
and their perceptions. Journal of Marriage and the Family, 60, 734
744.
Spenner, K. I. (1988). Social stratification, work, and personality. Annual
Review of Sociology, 14, 6997.
Szinovacz, M. E. (1989). Decision making on retirement timing. In D.
Brinberg & J. Jaccard (Eds.), Dyadic decision making (pp. 286304).
New York: Springer.
Szinovacz, M. E., & DeViney, S. (1999). The retiree identity: Gender and
race differences. Journal of Gerontology: Social Sciences, 54B, S207
S218.
Szinovacz, M., Ekerdt, D. J., & Vinick, B. H. (1994). Families and retirement: Conceptual and methodological tools. In M. Szinovacz, D. J.
Ekerdt, & B. Vinick (Eds.), Families and retirement (pp. 119). Newbury Park, CA: Sage Publications.
Treas, J. (1991). Common pot or separate purses? A transaction cost interpretation. In R. L. Blumberg (Ed.), Gender, family, and economy: The
triple overlap (pp. 211224). Newbury Park, CA: Sage Publications.
Verbrugge, L. M., & Jette, A. M. (1994). The disablement process. Social
Science and Medicine, 38, 114.
Wolinsky, F. D., Armbrecht, E. S., & Wyrwich, K. W. (2000). Rethinking
functional limitation pathways. The Gerontologist, 40, 137146.
Zellner, A. (1962). An efficient method of estimating seemingly unrelated
regressions and tests for aggregation bias. Journal of the American Statistical Association, 58, 348368.

Received November 15, 2000

Accepted September 13, 2001
Decision Editor: Fredric D. Wolinsky, PhD

Downloaded from http://psychsocgerontology.oxfordjournals.org/ by guest on July 20, 2016

Anderson, K., Burkhauser, R. V., & Quinn, J. F. (1986). Do retirement

dreams come true? The effects of unanticipated events on retirement
plans. Industrial and Labor Relations Review, 39, 518526.
Anderson, K., Clark, R. L., & Johnson, T. (1982). Retirement in dualcareer families. In R. Clark (Ed.), Retirement policy in an aging society
(pp. 109127). Durham, NC: Duke.
Berk, R. A. (1983). An introduction to sample selection bias in sociological
data. American Sociological Review, 48, 386398.
Blau, D. M. (1998). Labor force dynamics of older married couples. Journal of Labor Economics, 16, 595629.
Burr, J. A., Massagli, M. P., Mutchler, J. E., & Pienta, A. M. (1996). Labor
force transitions among older African American and White men. Social
Forces, 74, 963982.
Carr, D. (1996). Two paths to self-employment? Womens and mens self-employment in the United States, 1980. Work and Occupations, 23, 2653.
Chevan, A. (1996). As cheaply as one: Cohabitation in the older population. Journal of Marriage and the Family, 58, 656667.
Ekerdt, D. J., DeViney, S., & Kosloski, K. (1996). Profiling plans for retirement. Journal of Gerontology: Social Sciences, 51B, S140S149.
Ekerdt, D. J., Hackney, J., Kosloski, K., & DeViney, S. (2001). Eddies in
the stream: The prevalence of uncertain plans for retirement. Journal of
Gerontology: Social Sciences, 56B, S162S170.
England, P., & Farkas, G. (1986). Households, employment and gender: A
social economic, and demographic view. New York: Aldine.
Ferraro, K. F., & Farmer, M. M. (1999). Utility of health data from social
surveys: Is there a gold standard for measuring morbidity? American
Sociological Review, 64, 303315.
Flippen, C., & Tienda, M. (2000). Pathways to retirement: Patterns of labor
force participation and labor market exit among the pre-retirement population by race, Hispanic origin, and sex. Journal of Gerontology: Social Sciences, 55B, S14S27.
Gradman, T. J. (1994). Masculine identity from work to retirement. In E. H.
Thompson, Jr. (Ed.), Older mens lives (pp. 104121). Thousand Oaks,
CA: Sage Publications.
Gustman, A. L., & Steinmeier, T. L. (1994). Retirement in a family context:
A structural model for husbands and wives (National Bureau of Economic Research Working Paper No. 4629). Cambridge, MA: National
Bureau of Economic Research.
Hall, A., & Johnson, T. R. (1980). The determinants of planned retirement
age. Industrial and Labor Relations Review, 33, 241254.
Hamilton, B. H. (2000). Does entrepreneurship pay? An empirical analysis
of the returns of self-employment. The Journal of Political Economy,
108, 604631.
Hardy, M. A., & Shuey, K. (2000). Pension decisions in a changing economy: Gender, structure, and choice. Journal of Gerontology: Social
Sciences, 55B, S271S277.
Hatch, L., & Thompson, A. (1992). Family responsibilities and womens
retirement. In M. Szinovacz, D. Ekerdt, & B. Vinick (Eds.), Families
and retirement (pp. 99113). Newbury Park, CA: Sage Publications.
Hayward, M. D., Friedman, S., & Chen, H. (1996). Race inequities in mens
retirement. Journal of Gerontology: Social Sciences, 51B, S1S10.
Hayward, M. D., Friedman, S., & Chen, H. (1998). Career trajectories and
older mens retirement. Journal of Gerontology: Social Sciences, 53B,
S91S103.
Hayward, M. D., Grady, W. R., Hardy, M. A., & Sommers, D. (1989). Occupational influences on retirement, disability, and death. Demography,
26, 393409.
Heckman, J. J. (1979). Sample selection bias as a specification error.
Econometrica, 47, 153162.
Heckman, J. J. (1997). Instrumental variables: A study of implicit behavioral assumptions used in making program evaluations. Journal of Human Resources, 32, 441462.
Henretta, J. C., & ORand, A. M. (1983). Joint retirement in the dual
worker family. Social Forces, 62, 504520.
Henretta, J. C., ORand, A. M., & Chan, C. G. (1993). Joint role investments and synchronization of retirement: A sequential approach to couples retirement timing. Social Forces, 71, 9811000.
Honig, M. (1996a). Changes over time in subjective retirement probabili-