Age Identity in Context Greeting - Tài liệu tham khảo | Đại học Hoa Sen

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/258189470 Age Identity in Context
ArticleinSocial Psychology Quarterly · September 2010 DOI: 10.1177/0190272510379751 CITATIONS READS 41 311 2 authors: Markus H Schafer Tetyana P Shippee University of Toronto
University of Minnesota Twin Cities
95 PUBLICATIONS2,220 CITATIONS
104 PUBLICATIONS2,784 CITATIONS SEE PROFILE SEE PROFILE
All content following this page was uploaded by Markus H Schafer on 30 September 2015.
The user has requested enhancement of the downloaded file. Social Psychology Quarterly Vol. 73, No. 3, 245–264
Ó American Sociological Association 2010 DOI: 10.1177/0190272510379751 http://spq.sagepub.com
Age Identity in Context: Stress and the Subjective Side of Aging MARKUS H. SCHAFER Purdue University TETYANA PYLYPIV SHIPPEE University of Minnesota
The passage of time is fundamentally experienced through people’s interaction with their
social worlds. Life-course scholars acknowledge the multiple aspects of time-based experience
but have given little attention to age identity in a dynamic context. Drawing from a stress-
process model, we expected that turbulence within people’s family relations and health
declines would produce increases in subjective age. Family role transitions were also exam-
ined as potential factors that would hasten subjective aging, but only to the extent that they
produced stress. We used the Midlife Development in the United States study, a nationally rep-
resentative longitudinal survey. Regression results show that turbulence within one’s family
accelerated age identity and that the effect of diminishing psychosocial resources explained
this relationship. For the most part, however, changes within family roles did not affect age
identity. On the other hand, the incidence of chronic health problems increased subjective
ages, and this relationship too was mediated by the depletion of psychosocial resources.
The findings demonstrate an interconnection between the stress process and age identity.
Broadening our conception of time-based experience with attention to the stress process offers
exciting directions for future theory and research in life course sociology.
Keywords: age identity, stress, life course, subjective age, health, family, role transition
The passage of time is so fundamental sense involves people’s active attempts to
a process that it is often taken for
slow down or speed up the passage of time
granted in the study of the life course.
(Flaherty 2002), as well as people’s construc-
How could we possibly describe key pro-
tion of an age-based identity based on present
cesses in sociology or in social gerontology
life conditions (Johnson, Berg, and Sirotzki
such as the accumulation of advantage, the
2007; Logan, Ward, and Spitze 1992; Wester-
transitioning through social roles, and the
hof and Barrett 2005) and pivotal life experi-
debilitating trajectories of chronic illness
ences from their pasts (Schafer 2009).
apart from references to time? Overlapping
This paper builds on age identity research
with the obvious chronological continuum
and considers stress-based shifts in age iden-
on which life events occur, however, is
tity over a ten-year period. It is one of the few
another, internalized clock. This subjective
empirical studies of which we are aware that
treats age identity not as a static measure
from a single point in time, but as a dynamic
We thank Nate Shippee, Ken Ferraro, and Tim Owens
quality which can fluctuate in relation to other
for their helpful feedback on earlier versions of this paper.
events in a person’s world. Past work empha-
The data were made available by the Inter-University
Consortium for Political and Social Research, Ann Arbor,
sizes that age identities are strongly tied to
MI. Neither the collector of the original data nor the
characteristics of the body, such as health sta-
Consortium bears any responsibility for the analyses or
tus (Barrett 2003; Charmaz 1991; Furman
interpretations presented here. Direct all correspondence
1997). This paper builds on that prior work
to Markus Schafer, Department of Sociology, Purdue
and directs attention to several levels of social
University, West Lafayette, IN 47907-2059; mhschafe@ purdue.edu.
abstraction: (1) health status (the individual 246 SOCIAL PSYCHOLOGY QUARTERLY
herself); (2) role occupancies (the individual
Before examining the empirical data, how-
in relation to her social world); and (3) turmoil
ever, we first provide an overview of the
within the lives of primary family ties (the
stress process and its relation to aging.
individual’s intimate social world itself).
Then, we provide a brief discussion of age
In considering why each of these three fac-
identity as an important aspect of aging and
tors may impinge on time-based evaluations
the factors, beyond chronological time, which
such as age identity, it is helpful to forge con-
are expected to make age identities grow
nections between life-course sociology and older over 10 years.
other areas of scholarship that investigate
how humans interact with time. Stress is
CHANGES IN AGE IDENTITY AND THE STRESS
a strong and unifying scientific concept for ad- PROCESS
dressing processes related to age and time.
Accordingly, we borrow from incipient life-
The idea that stress is a graying force tends
course thinking in the life sciences—alongside
to pop in and out of the media’s attention, and
recent findings from social psychology—and
has garnered public interest regarding whether
draw connections between the stress process
prominent people appear to be ‘‘aging well’’
and aging. Stress wears down reserves at mul-
(e.g., Sellers 2009). Though social scientists
tiple levels of scientific observation. From
are often rightfully wary of sensational claims
cells (Epel et al. 2004), to organs (Kelly,
splashed across newspaper headlines, the
Hurtzman, and Daniels 1997), to psychologi-
stress and aging connection has become
cal fortitude (Foster, Hagan, and Brooks- increasingly undergirded by pathbreaking
Gunn 2008), stress produces the likeness of
aging research in the life sciences. As one
being older for a given chronological age. The
example, an innovative line of research has
connection between physical and subjective
recently confirmed that chronically stressed
‘‘weathering’’ (Foster et al. 2008) suggests
people have older ‘‘biological ages,’’ as
that a stress-induced acceleration of aging is
demonstrated by evidence at the cellular
as plausible in the interview setting as it is
level (Damjanovic et al. 2007; Simon et al. under the microscope.
2006). Simply put, stress makes people
Our chief question is this: How do life
weaker and more vulnerable, though ‘‘the
events alter the age identity of adults over the
exact mechanisms that connect the mind to
course of a decade? In light of past research,
the cell are unknown’’ (Epel et al. 2004).
we are certainly mindful of how health declines
Though in early stages, social scientists
may correspond with changes in age identity.
have offered corroborating evidence of how
We also expect, however, that as fundamentally
stressful events precipitate similar aspects
social creatures, another part of what regulates
of premature aging. Two recent studies using
people’s age identity are significant events in
multiple waves of the Add-Health survey
their social worlds. Using a national longitudi-
find that hardship during childhood (e.g.,
nal survey, this paper therefore also focuses
neglect from parents, neighborhood instabil-
on changes in family role relations and strains
ity) hastens the sense of adulthood among
within families. We see the stress process as
people in their late teens to age 22 (Foster
a unifying perspective for explaining why
et al. 2008; Johnson and Mollborn 2009).
turbulence—whether located in the body or
Foster and colleagues (2008), in fact, show
within one’s intimate social world—alters age
that stressors trigger both early menarche
identity. Stress likely wears down the very psy-
in childhood and early identification as an
chosocial resources (e.g., sense of control and
adult, thus begetting a process of both physi-
positive affect) which enable the maintenance
cal and subjective ‘‘weathering’’ early in the
of a youthful identity (Barrett 2005). Thus,
life course. Similar to their life science col-
our analyses will be sensitive to factors which
leagues, however, the authors of both Add-
potentially mediate the relationship between
Health studies have yet to fully and conclu- stressors and age identity.
sively explicate the precise mechanisms
AGE IDENTITY AND THE STRESS PROCESS 247
linking stress and age identity. Exploring the
between stressor and outcome. This is a deteri-
psychosocial mechanisms will be at least
oration model, as Ensel and Lin (1991) termed a step in that direction.
it. Stressors, such as turbulence within one’s
One of the imperatives for understanding
family, essentially chip away at the resources
how stressors influence age constructs (bio-
which otherwise protect against stress; waning
logical and subjective alike) is to adopt a theo-
resources decrease well-being, thus serving as
retical perspective which clearly specifies the
the mediating factor between stressor and
stress process. In sociology, stress theory outcome.
maintains that negative events present a
Psychological distress is most commonly
burden that, unless shielded by protective
assessed as an outcome measure of the stress
resources, will compromise mental and phys-
process, though many studies also examine
ical well-being (Pearlin 1989; Thoits 1995).
physical health as a consequence (e.g., Lin
Stress is typically conceptualized as a process
and Ensel 1989). Age identity, the focus of
in which stressors (e.g., negative life events)
this paper, is only beginning to emerge as
are the basic independent variables. Social
a dependent variable of the stress process
psychologists have provided much-needed
(Foster et al. 2008). The recent evidence dis-
nuance to the basic idea of stressors by
cussed above suggests that age identity is
addressing the meaning that people attach to
a very relevant outcome, particularly because
social identities, a factor which shapes the
it is an important life-course aspect of well-
stressfulness of life events (Marcussen, Ritter,
being and because the stress process is an
and Safron 2004; Thoits 1991). Family roles,
ideal perspective for aligning with social ger-
for instance, are particularly consequential for
ontology (see Pearlin 1996 for an expanded
identity (Reitzes and Mutran 2002), and it is
discussion). Identifying psychosocial mecha-
therefore plausible that problems involving nisms is a next step.
the lives of family would be more stressful
We focus on personal control and positive
than problems in the lives of acquaintances
affect as two psychosocial resources that may or celebrities.
be compromised by stressful conditions and
Though stressors increase stress-related
therefore increase vulnerability to feeling
outcomes, resources (e.g., psychosocial mea-
older. As we will briefly show in the sections
sures of well-being) act as protectors. A sense
below, both types of psychosocial resources
of personal control, for instance, endows peo-
can be dampened by health problems, by
ple with the confidence to address unfavorable
stressful role losses, and by misfortune in
life circumstances and to push on (Thoits
the lives of intimate others. Furthermore,
1995). Likewise, possessing positive affect
both psychosocial resources are associated
helps cast stressful events in a better light. Pos-
with the maintenance of youthful identities
itive affect is a fairly stable trait, but is respon-
(Barrett 2003, 2005; Bowling et al. 2005).
sive to life circumstances such as stressful
Barrett (2005), for instance, argues that
events (Watson and Walker 1996).
‘‘maintaining a younger—and more age-
As Ensel and Lin (1991) show in their
discrepant—identity as one ages is an active
influential piece, there are a number of hypo-
reinterpretation of one’s chronological age
thetical ways to configure the relationship
that requires a high degree of [perceived]
between these protective resources (e.g., per-
control over critical life domains’’ (178). Oth-
sonal control, positive affect) and social
ers have noted how self-enhancing, compen-
stressors in regards to an outcome of the
satory youthful identities are harder to
stress-process model. Consistent with their maintain when stressful conditions are findings and with subsequent empirical
induced (e.g., Zebrowitz 2003), suggesting
work (e.g., Martin, Grunendahl, and Martin
that a sense of powerlessness and emotional
2001), we proceed under the assumption
distress compromises people’s ability to pro-
that psychosocial resources—diminished by
ject their ideal self-presentations. However, stressful events—mediate the association
we have only been able to identify one 248 SOCIAL PSYCHOLOGY QUARTERLY
cross-sectional study on age identity among
changing as a linear function in our metric of
adults that has explicitly incorporated the
choice (e.g., the span of time it takes to orbit
concept of stress, and the measures used
the sun, as Westerners typically keep long-run
related mostly to perceived satisfaction with
social time). From the standpoint of social
work or romantic partnership rather than
psychology and the stress model, the more
drawing from the stress process in a more sys-
interesting theoretical question is how age
tematic way (Barrett 2005). Using Ensel and
identity may vary in correspondence with
Lin’s (1991) deterioration model, we seek to
other significant life events. We will start with
apply the stress model to age identity.
the most distal level and furthest outside of the
individual herself (turbulence within the lives AGE IDENTITY
of family relations), move to the intermediate
level that connects people to their social world
It perhaps comes as little surprise that in
(role positions and changes in roles), and wrap
Western societies where good health and
up with the most micro, individualistic level
youthfulness are celebrated, most adults seek
(problems within the body). Each of these lev-
to maintain a young age identity (Barrett
els represents locations of potential stressors
2003; Logan et al. 1992; Westerhof et al.
in social life, and by extension, possible influ-
2003). Subjective youthfulness in turn is asso-
ences on age identity change. These factors
ciated with well-being (Westerhof and Barrett
have also been considered, to various degrees,
2005), confidence about abilities (Schafer and in past age identity research.
Shippee 2010), and even health stability in
older age (Demakakos, Gjonca, and Nazroo
Potential Influences on Age Identity
2007). Extant literature suggests that age iden-
tities are produced by a complex constellation
For a paper concerned with people in the
of factors which influence how people interact
context of time, change, and biographical
in society, such as health (Barrett 2003; Logan
development, life-course theory is a good per-
et al. 1992; Westerhof and Barrett 2005),
spective to emphasize from the outset (Elder
workforce participation (Johnson et al. 2007),
1994). In Elder’s thinking, significant others—
and familial stressors which limit one’s sense
particularly people’s spouses, children, and of control (Barrett 2005).1
parents—are collectively enmeshed in mutu-
Though scholarship on age identity has
ally influential life trajectories. Essentially,
brought insight to an important element of
primary social groups are foundational for
the self, it has until now considered the topic
understanding people’s experiences, and it
almost exclusively as a static state in a single
is the very durability and continuity of these
moment of time. 2 This is a peculiar predica-
relationships which make them so meaning-
ment because its subject is an inherently tem-
ful for people’s life chances and well-being
poral phenomenon; age is an anemic concept
(Gecas 2003). For the current inquiry, events
without a dynamic notion of time. In a chrono-
in the lives of significant others must
logical sense, age is continually in flux, albeit
2 We have identified only one empirical article with
our quantitative measure of age identity using two waves
1 Role transitions have long been considered seminal
of data, but it was essentially a descriptive study of elderly
in the social construction of age (Neugarten, Moore, and
Finnish adults (Uotinen et al. 2006). One recent study
Lowe 1965), but research demonstrates that they do not
observed a sample of 19-year-old Philadelphians over
have uniform effects on age identity across age groups
a 2-year period to see whether role and responsibility
or aspects of age identity. Mainly, transitions in family
changes (e.g., starting an independent household, parent-
and work life contribute to subjective adult attainment
hood) elicit the subjective attainment of adulthood (Ben-
among late adolescents and people in their early twenties
son and Furstenberg 2007). Several older descriptive
(Benson and Furstenberg 2007; Johnson et al. 2007). Most
studies utilizing longitudinal data have also used qualita-
studies among middle- and older-age adults, however,
tive categorical variables assessing whether people con-
find that age-graded transitions such as widowhood or
sidered themselves to be ‘‘middle-aged’’ or ‘‘very old’’
retirement do not influence age identity when health is
at two time points (Bultena and Powers 1978; Markides
controlled (e.g., Logan et al. 1992; Westerhof et al. 2003). and Boldt 1983).
AGE IDENTITY AND THE STRESS PROCESS 249
certainly be considered if we wish to under-
Besides turbulence within one’s primary
stand changes in people’s own identities
group, there are other types of interactions
over a decade. Components of well-being,
with the social world that may shift the moor- including age identity (Westerhof and
ings of age identity over a decade. The transi-
Barrett 2005), are not an individual’s own
tioning through age-graded social roles could
project—they are influenced by the lives of
be expected to change how people feel in important others.
relation to time. Prior research on middle-
What sorts of long-term relationships
and older-age adults, however, is inconsistent
might impinge on a person’s sense of age
on whether transitions such as widowhood
and well-being? Although potentially myriad,
and retirement produce older age identities
family ties are a logical set to examine and
(see Barak and Stern 1986; Logan et al.
are the focus for this study. The role of parent
1992; Westerhof et al. 2003). One problem
does not end once children are out of the
is that the earliest studies cited above did
house, but is better understood as a ‘‘life-
not control for health status, but a further lim-
long trajectory of shifting demands and
itation is the reliance on cross-sectional study responsibilities’’ (Milkie, Bierman, and
designs in which actual role transitions can-
Schieman 2008:87). When these linked lives
not be observed over time. The longitudinal
are perturbed, we expect a person to be
studies in print have been primarily con-
shaken. Not surprisingly, then, when adult
ducted on limited-age samples of late adoles-
children are burdened by a cumulative load
cent respondents and used only measurements
of problems, parents’ stress level is also
of discrete, categorical age identities (Benson increased (Greenfield and Marks 2006;
and Furstenberg 2007; Johnson et al. 2007).
Milkie et al. 2008). Likewise, the role of
From these studies, it is apparent that transi-
a child is not abdicated when a person begins
tions from school to the workforce and from
their own family; studies document the
unattached life to family responsibilities
linked-lives concept by demonstrating the
bring feelings of adult attainment. But more
negative consequences associated with seeing
age-representative studies with controls for
one’s father or mother in declining health or
health status and continuous measures of
other problematic experiences (e.g., Willson,
felt age suggest that age identity is relatively
Shuey, and Elder 2003). Finally, despite
impervious to role changes (e.g., Westerhof
changing social norms toward marriage,
et al. 2003). For this reason, we do not expect
Americans continue to view spouses among
role changes in general to produce marked
their most significant others (McPherson,
changes in age identity over a decade.
Smith-Lovin, and Cook 2001). Accordingly,
A certain category of role transitions,
experiences of trouble in the life of one’s
however, may produce substantial changes
spouse reverberate in husbands’ and wives’
in age identity, those being ones that are
well-being and generate stress.
drastically off-time in family life. A recent
In all of the cases mentioned above—
article, for instance, found that death of
problems in the lives of children, parents,
a mother during childhood produced three
spouses—the stress induced by turmoil is
extra years of age identity, on average—up expected to wear down psychosocial
to half a century later in the life course
resources. Roles attached to family life are
(Schafer 2009). Death of a parent during
central for personal identities (Reitzes and
adulthood, however, did not have any effect.
Mutran 2002), and so the welfare and misfor-
Based on these findings, we anticipate that
tune of family members impinge on people’s
extremely off-time and unexpected role tran-
well-being. This includes a broad set of fac-
sitions may produce older age identities.
tors in the experience of aging (e.g., health,
From a demographic standpoint, the clearest
mental health, the availability and receipt of
example of such a role transition in the cur-
caregiving, the transmission of emotional
rent study would be the loss of a child (Ha-
support; Bengtson and Silverstein 1993). gestad 1988). 250 SOCIAL PSYCHOLOGY QUARTERLY
The immutable ending of any family rela- METHODS
tionship, however, can be deeply unsettling.
The analyses for this study are based on
And so to qualify our expectation above, we
data from the National Survey of Midlife
anticipate that if the loss of a spouse, sibling,
Development in the United States (MIDUS).
or parent diminishes positive psychosocial
In 1995, random-digit-dialing was used to
resources, age identity will also be affected.
obtain a representative sample of English-
The discontinuation of roles that are central
speaking non-institutionalized adults aged
for identity poses particular risks for well-
25 to 74 in the 48 contiguous states, with
being (Marcussen et al. 2004; Thoits 1991).
oversampling for males between 65 and 74.
Again, in light of their importance (Reitzes
The response rate from these initial telephone
and Mutran 2002), the rupture of familial
interviews was 70 percent. The final stage
roles likely induces stress by diminishing pos-
included a questionnaire mailed to those
itive affect and a sense of control (Chi and
who participated in the telephone interview,
Chou 2001). Studies on widowhood, divorce,
yielding an 86.6 percent response rate.
and parental and sibling death all attest to the
Thus, the overall response rate was 61 percent
damaging effects of these experiences—
(.70 x .87 = .61), producing a total sample of
though psychological health is also quite
3,034 participants who completed both the
capable of resilience (Bonanno 2004).
telephone and the mail portions of Wave I.
Finally, negative changes within a person’s
Approximately 10 years later, a second
body over the course of a decade can be a har-
wave of data was collected, using both phone
binger of older age. The age-inducing effect
and mail questionnaires. Of the complete
of poor health is by far the most consistent
Wave II sample, 2,103 individuals were fol-
finding of past quantitative research—in
lowed up (69 percent), but not all tracked re- cross-national studies (Westerhof et al.
spondents returned the mail questionnaires in
2003, Westerhof and Barrett 2005), across
which age identity was asked. Our final sam-
a wide variety of health measures (Barrett
ple thus consists of 1,668 respondents who
2003), and in studies using categorical and
completed both phone and mail surveys of
continuous age identity outcome variables
Wave II and after we used multiple imputa-
alike (Barak and Stern 1986). Qualitative
tion for 323 subjects who had missing data
studies, too, have demonstrated the centrality
on independent variables.3 Findings were
of health for the self-concept and the ways
robust to alternate strategies of handling miss-
that people strive to maintain a vigorous,
ing data. All analyses used weights to adjust
healthful, and ultimately youthful presenta-
for differential probabilities of selection and
tion of self (Charmaz 1991; Clarke 2001; Fur-
differential non-response with Stata 10.0.
man 1997). These projects demonstrate the
ways that health is central to a sense of auton- Measures
omy, purpose, and self-efficacy. The inci-
dence of health problems, then, is expected
Age identity. We operationalize the con-
to accelerate age identity over a decade.
cept of age identity with a measure that traces
Some of this effect may result from health
the rate of participants’ subjective aging. Spe-
being a reminder that one is a finite mortal
cifically, the question asked respondents,
with a limited lifespan, but part of the effect
‘‘Many people feel younger or older than
of declining health may be due to its associa-
tion with frustration and disillusionment.
3 We did not impute missing data on the dependent
Indeed, declining health is instrumental in
variables (von Hippel 2007). We used Royston’s (2004)
the stress process by engendering distress
approach to multiple imputation in Stata, creating 5 cop-
(Kelley-Moore and Ferraro 2005; Schieman
ies of the data, each of which has missing values imputed
and Plickert 2007). Therefore, health decline
by different regression equations (depending on the prop-
erty of the variable to be imputed), analyzing each dataset
is expected to hasten age identity to the extent
separately, and then generating a single estimate which is
that it wears down psychosocial resources.
an average across the five datasets.
AGE IDENTITY AND THE STRESS PROCESS 251
they actually are. What age do you feel most
others had experienced any of the following:
of the time?’’ Subjective aging was measured
chronic disease/disability; frequent minor ill-
by subtracting subjective age at Wave I from
ness; emotional problems; alcohol/substance
subjective age at Wave II. This results in neg-
use; financial problems; problems at school
ative values reflecting younger subjective
or work; difficulty finding or keeping a job;
ages and positive values reflecting older sub-
marital or partner relationship problem; legal jective ages.
problems; or difficulty getting along with peo-
Mediating variables.Personal control (often
ple. We use baseline family adversity as a con-
called mastery in the stress literature) is a key
trol, while incident family adversity between
resource in the stress model. Personal control
Wave I and Wave II is of primary interest. For
was measured at both waves as a 12-item index
Wave I family adversity, responses were
(Prenda and Lachman 2001). Respondents were
summed, with scores ranging from 0 to 17.
given statements such as the following: ‘‘In gen-
For incident family adversity, respondents
eral I feel I am in charge of the situations in
were coded as experiencing an event(s) if their
which I live.’’ Each item ranged from 1 =
significant others had not faced an adversity at
strongly disagree to 7 = strongly agree, and
Wave I but had experienced it by Wave II.
the 7 negatively worded questions were each re-
This approach allowed us to isolate the timing
coded so that high values would correspond
of the events to the 10-year span between
with high sense of personal control. Taken
waves rather than anytime in their lives prior.
together, the index items had a reliability of
Responses were summed, producing a possible
alpha = 0.85. The second psychosocial resource
range from 0 to 31 (although the highest score
included was positive affect. Positive affect is
was 29, it was reported by only one respon-
a fitting variable to capture emotional well-
dent, with the next highest score resting at
being because it is a fairly stable trait, yet is
14). Because cumulative adversity scores
responsive to life circumstances (Watson and
often demonstrate non-linear effects in regres-
Walker 1996). Positive affect was measured at
sion analysis, we use family adversity’s
both waves as a six-item scale (Mroczek and
square to avoid potential model misspecifica-
Kolarz 1998). Questions relate to whether re-
tion (Pollard, Hawkins, and Arthur 1999; spondents felt cheerful, in good spirits,
Schilling, Aseltine, and Gore 2008). Models
extremely happy, calm and peaceful, etc. over
presented herein include the quadratic term
the last 30 days. Each item ranged from 1 =
where statistically significant.
strongly disagree to 5 = strongly agree with
Family roles. A second interest for this
high values corresponding to high sense of pos-
study is change in family roles, which repre-
itive affect. The index items had a reliability of
sent a key relation that an individual has to 0.91 at each wave.
her social world. Besides enabling us to
Family adversity. Since a key interest of
directly consider whether changes in these
this paper is to examine how negative events
roles correspond with changes in age identity,
in one’s network of intimate ties influence
including family roles also controls for the
change in subjective age, our first main inde-
configuration of family relations that are per-
pendent variable is a score for family adver-
tinent for family adversity. We therefore
sity. This variable is intended to capture the
include the composition of people’s families
extent of adversity experienced in people’s
at the beginning of the study as well as
intimate social worlds, or what Turner and
changes in the structure of families from
Lloyd (1995) have called ‘‘the joint or cumu-
Wave I to Wave II. Marital status at baseline
lative effects of multiple traumas’’ (268). The
was operationalized with four dummy varia-
MIDUS study included a series of items about
bles, including currently married, widowed,
problems faced by respondents’ spouse(s)/
divorced, and never married. Those who
partner(s), children, and/or parent(s). At both
were married at Wave I comprised the omit-
waves, respondents were asked whether, in
ted reference group in regression analyses.
the past twelve months, these significant
We further differentiated between those 252 SOCIAL PSYCHOLOGY QUARTERLY
respondents who, between waves, stayed
chronic illness, respondents received a score
married, got married, got divorced, got wid-
if they had not experienced a particular condi-
owed, or never married, with remained mar-
tion it at Wave I but had experienced it by
ried as a reference group. Respondents’
Wave II. Responses were summed and range
status as a parent at baseline was also
from 0 to 5 for incident serious illness and
a dummy (1 has any biological children,
from 0 to 17 for incident chronic illness.
0 does not), and those who became parents
Control variables. The analyses controlled
between waves were also identified (1
for a number of factors collected at Wave I
became a parent, 0 did not).4 We differenti-
and identified in the literature as related to
ated between respondents who had both par-
age identity and to personal control and pos-
ents alive during baseline, those who had
itive affect. Chronological age was included
one parent alive, and those who had neither
in all multivariate models, coded as year of
parent alive. Variables for loss of one parent
birth subtracted from 1995. To capture the
and loss of both parents between waves versus
possibility of nonlinear effects of chronologi-
both parents being still alive were also gener-
cal age on subjective age, an age-squared
ated to chart potential change in family roles.
term was also included in regression analyses.
Finally, we included measures for other inti-
Female and black were binary variables with
mate loss, such as death of a child (1 has
1 equal to the name of the variable. Males
a child die between waves, 0 does not) and
were the reference group for female, and
death of a sibling (1 has a sibling die between
self-identified white respondents were the waves, 0 does not). 5
reference group for black. Education was
Health decline. To account for effects of
coded with three dummy variables: less
health decline, our third domain of age iden-
than high school, high school/GED, and
tity influence, we included variables for both
bachelor’s degree and higher, with high
serious and chronic illnesses at Wave I as
school/GED as the reference group in regres-
well as incidents of serious and chronic ill-
sion analyses. Household income was mea-
ness between waves. Specifically, the count sured by summing the participants’
of serious illnesses included 7 conditions,
personal income with income from spouse
including cancer (ever), diabetes, hyperten-
and from anyone else living in the house-
sion, heart trouble (ever), HIV/AIDS, neuro-
hold. We also created change in household
logical problems, and stroke, while the income between waves by subtracting
summary of chronic illnesses included 25
household income at Wave II from house-
problematic conditions, including hay fever,
hold income at Wave II. Working was
back and skin problems, and migraine head-
a dummy variable which measured whether
aches. Except for cancer and heart problems,
the respondent was working for pay at
all conditions were asked in relation to the
Wave I (1 = working, 0 = not) while retired
past 12 months. For incident serious and
corresponds to those who became retired between Wave I and Wave II. 4
Adjustment for nonresponse. Using longi-
We included controls for quality of relationship with
tudinal data has key advantages, but introdu-
spouse and children. The variables were not significant
and the findings were not substantially altered. Therefore,
ces the problem of attrition. To ensure that
we use a more parsimonious approach with fewer inde-
our estimates were not biased by selective pendent variables.
attrition between waves, we followed the
5 Also, in supplementary analyses, we explored recent
Heckman (1979) method of correcting for
role transitions. Specifically, we differentiated between
nonresponse bias. First we estimated a probit
transitions that occurred two or fewer years ago versus
model predicting likelihood of W2 response,
those that happened three to ten years ago. Among recent
transitions, only the death of both parents emerged as
using a variety of demographic and psychoso-
a significant predictor of age identity change. Thus, for
cial variables related to survey response (e.g.,
parsimony, we did not include differentiation for recent-
age, report of past discriminatory behavior,
ness of role transitions in our final models, but these anal-
smoking). Next we calculated a hazard
yses are available upon request.