Self-Esteem Across The Second Half of Life The Role of Socioeconomic Status, Physical

────────────────────────────────────────────────────────────
This is an author generated postprint of the article:

von Soest, T., Wagner, J., Hansen, T., & Gerstorf, D. Self-esteem across the second half of life: The role of
socioeconomic status, physical health, social relationships, and personality factors. Journal of Personality
and Social Psychology. doi:10.1037/pspp0000123
Copyright 2017 APA. This article may not exactly replicate the final version published in the APA journal.
It is not the copy of record.
The final publication is available on http://dx.doi.org/10.1037/pspp0000123.
────────────────────────────────────────────────────────────
Running Head: SELF-ESTEEM ACROSS THE SECOND HALF OF LIFE
Self-Esteem Across the Second Half of Life: The Role of Socioeconomic Status, Physical
Health, Social Relationships, and Personality Factors
Tilmann von Soest

University of Oslo and Oslo and Akershus University College of Applied Sciences
Jenny Wagner
Leibniz Institute for Science Education and Humboldt University
Thomas Hansen
Oslo and Akershus University College of Applied Sciences
Denis Gerstorf
Humboldt University
Author Note
Tilmann von Soest, Department of Psychology, University of Oslo, Oslo, Norway, and
Norwegian Social Research, Oslo and Akershus University College of Applied Sciences, Oslo, Norway;
Jenny Wagner, Leibniz Institute for Science Education, Kiel, Germany, and Department of Psychology,
Humboldt University, Berlin, Germany; Thomas Hansen, Norwegian Social Research, Oslo and Akershus
University College of Applied Sciences, Oslo, Norway; Denis Gerstorf, Department of Psychology,
Humboldt University, Berlin, Germany.
Tilmann von Soest’s work on this project was facilitated by a longer-term stay at the Humboldt
University Berlin as a Visiting Scientist. The Norwegian Life Course, Ageing and Generation Study
(NorLAG) is financed by the Research Council of Norway (grant nos. 149564 and 168373); the Ministry
of Health and Care Services; the Ministry of Labour; the Ministry of Children, Equality and Social
Inclusion; the Ministry of Local Government and Regional Development; Norwegian Social Research
(NOVA); and Statistics Norway. The NorLAG dataset is part of the ACCESS Life Course infrastructure
project funded by the National Financing Initiative for Research Infrastructure at the Research Council of
Norway (grant no. 195403) and by NOVA. The data are distributed by the Norwegian Centre for
Research Data. None of the aforementioned institutions are responsible for the current analyses and
interpretations of the data.
Correspondence concerning this article should be addressed to Tilmann von Soest, Department of
Psychology, University of Oslo, P.O. Box 1094 Blindern, 0317 Oslo, Norway.
E-mail: [email protected].
SELF-ESTEEM ACROSS THE SECOND HALF OF LIFE 2
Abstract
Self-esteem development across adulthood has been in the center of interest for some time now.
However, not much is known about factors that shape self-esteem and its development in the
second half of life and whether the factors differ with age and gender. To examine these
questions, this study uses two-wave data from the population-based NorLAG study in Norway (N
= 5,555; Mage = 58 years; 51% women) and combines self-report data on self-esteem and
personality with registry-based information on socioeconomic status (education, income,
unemployment), health problems (sick leave, lifetime history of disability), and social
relationships (cohabitating partner, lifetime history of divorce and widowhood). Results from
latent change score models revealed that self-esteem peaked at around age 50 and declined
thereafter. More importantly, lower socioeconomic status, not having a cohabitating partner,
unemployment, and disability were each uniquely associated with lower levels of self-esteem
and/or steeper declines in self-esteem over the five-year study period. Over and above registry-
based information, personality characteristics were relevant, with a more mature personality
being associated with higher self-esteem level. Emotionally stable participants also showed less
pronounced declines in self-esteem. Moreover, associations of disability and of emotional
stability with self-esteem level were weaker with advancing age. Among women, self-esteem
level was more strongly associated with emotional stability and less strongly with openness,
compared to men. Our findings demonstrate the utility of registry-based information and suggest
that physical health, social relationships, and personality factors are in manifold ways uniquely
associated with self-esteem and its development later in life.
Keywords: self-esteem, register data, old age, physical health, relationships

Self-Esteem Across the Second Half of Life: The Role of Socioeconomic Status, Physical Health,
Social Relationships, and Personality Factors
Self-esteem is defined as an individual’s general attitude towards or evaluation of the self
and reflects people’s beliefs about how worthy they are as a person (Blascovich & Tomaka,
1991; Rosenberg, Schooler, Schoenbach, & Rosenberg, 1995). The importance of self-esteem as
a fundamental psychological construct has been emphasized by research showing self-esteem to
predict future outcomes such as good physical and mental health, satisfaction in close
relationships, social support, and also labor force participation and economic prospects across
adulthood (Orth & Robins, 2014; Trzesniewski et al., 2006). It is thus of key importance to better
understand the conditions under which self-esteem thrives. However, little is known about factors
that shape self-esteem and its development across adulthood and whether these operate in age-
and gender-specific ways. This report moves one step further by making use of self-report and
register-based information that cover socioeconomic status (SES), health problems, social
relationships, and personality to examine antecedents and correlates of self-esteem across the
second half of life. Specifically, we use two-wave data from the population-based NorLAG study
in Norway (N = 5,555; Mage = 58 years at baseline; 51% women) and examine three sets of
questions. First, how does self-esteem change across adulthood and old age? Second, what are
the unique roles of current conditions and lifetime history of SES (education, income,
unemployment), health problems (sick leave, disability), and social relationships (cohabitating
partner, divorce, widowhood) as well as Big Five personality traits for level and development of
self-esteem across the second half of life? Third, does the relevance of these factors differ across
age and gender?
Self-Esteem and Its Development Across the Second Half of Life

Theoretical accounts in personality and developmental psychology, including social
investment perspectives (Roberts, Wood, & Smith, 2005) and lifespan developmental
perspectives (Baltes, Lindenberger, & Staudinger, 2006) emphasize the importance of social role
changes for personality development across the lifespan. In accordance with these perspectives, it
has been proposed that through young and middle adulthood people invest heavily in socially
rewarding roles in the areas of intimate relationships, family life, and work. These investments, in
turn, are said to result in more mature personality and gradual increases in self-esteem (Orth,
Maes, & Schmitt, 2015). However, in old age, reductions in social networks and compromised
health may undermine people’s self-esteem (Orth et al., 2015). Several large-scale longitudinal
studies have largely supported age trends of that kind, reporting self-esteem to peak in midlife at
around age 50, followed by a phase of mean-level decline (for an overview, see Orth & Robins,
2014). However, the research studies do not agree on the average size of the decline in old age.
Some studies have reported steep forms of decline; others have found that the rate of decline is
rather minor. For example, Orth and colleagues (Orth, Robins, & Widaman, 2012; Orth,
Trzesniewski, & Robins, 2010) have reported from two long-term longitudinal studies that mean
levels of self-esteem decrease by about two thirds of a standard deviation from age 60 onwards.
In contrast, in three further independent longitudinal studies Wagner and colleagues (Wagner,
Gerstorf, Hoppmann, & Luszcz, 2013; Wagner, Hoppmann, Ram, & Gerstorf, 2015; Wagner,
Lang, Neyer, & Wagner, 2014) have found modest rates of decline, amounting to no more than
one fifth of a standard deviation per 10 years. In the context of these conflicting findings, this
study examines mean-level change in self-esteem from midlife to old age using long-term
longitudinal data from a population-based national sample from a Northern European country.
Predictors of Level and Change in Self-Esteem

The second and primary aim of this study is to examine variables from a variety of
different life domains as predictors of level and change in self-esteem across the second part of
life. We draw from two broad theoretical perspectives to provide insights into factors that may
influence self-esteem. First, according to the intrapersonal perspective, as originally proposed by
William James (1890), global self-esteem is based on people’s own perceptions of how
adequately they perform in domains where they consider success to be important. People’s
resources, abilities, and competence in important areas of life are thus expected to shape self-
esteem. Second, the interpersonal perspective, dating back to Cooley’s (1902) and Mead’s (1934)
notions in the framework of symbolic interactionism, emphasizes the role of social influences in
shaping self-esteem. This perspective largely considers self-esteem to be a result of
internalizations of others’ perceptions and evaluations of oneself. A prominent contemporary
variant of the interpersonal perspective is sociometer theory, which proposes that the major
function of self-esteem is to monitor threats of social exclusion (Leary, Tambor, Terdal, &
Downs, 1995).
Both perspectives provide important predictions about factors that may influence self-
esteem: The intrapersonal perspective’s focus on a person’s own abilities and resources would
predict that factors related to high SES, such as wealth, education, and high status occupations
are sources of self-esteem in adults. Good physical health would also be considered to be an
important individual resource relevant for self-esteem and its development. Likewise, the
intrapersonal perspective predicts that basic personality traits that promote one’s own status may
be of importance.
The interpersonal perspective and sociometer hypothesis would predict that factors
determining social integration influence self-esteem. Social relations and particularly close social
ties, such as those with one’s spouse, would then be of importance for self-esteem. Also for this
perspective, SES and physical health may be of importance, because low SES and health
problems may lead to social marginalization. Moreover, personality traits facilitating the
establishment and maintenance of close social ties are relevant for the interpersonal perspective.
In summary, the two broad conceptual perspectives predict that a variety of demographic,
health, and psychosocial factors are important sources of self-esteem. Despite differences in
fundamental assumptions underlying the two perspectives, their predictions are complementary
rather than competing. This study explores a comprehensive number of factors considered by the
two perspectives to influence self-esteem. The factors include indicators of SES (education,
income, unemployment), physical health (sick leave, disability), social ties (cohabitating partner,
divorce, widowhood), and measures of broad personality characteristics. A more detailed account
of the predictions of the two perspectives and previous research on predictors of self-esteem is
presented in the following.
Socioeconomic status. Following the intrapersonal perspective, SES is an evident
candidate as a source of self-esteem, because factors such as high education, high income, and
high status occupations are generally highly valued in Western societies. Moreover, low SES
may lead to social marginalization and may therefore be important for self-esteem according to
the interpersonal perspective. A meta-analysis (Twenge & Campbell, 2002) provided empirical
support for cross-sectional associations between SES and self-esteem, concluding that indicators
of SES had small but significant associations with individual differences in self-esteem.
Longitudinally, however, studies have typically not found evidence for associations of SES
indicators (education, income, occupational status) with rates of change in self-esteem (Orth et
al., 2010; Wagner et al., 2014).
Physical health. Physical health may also shape self-esteem, according to the
intrapersonal perspective, because health problems may make it difficult for individuals to live up
to their own expectations in domains where physical health is an advantage for successful
performance. Physical health problems may also compromise opportunities for rewarding social
engagement and relationships, which in turn may undermine self-esteem according to the
interpersonal perspective. These notions have received support in that cross-sectional
associations have been found between indicators of subjective health and levels of self-esteem
(Erol & Orth, 2011; Wagner et al., 2014). Longitudinally, empirical studies have found no or
only very small associations between self-reports of physical health and changes in self-esteem
across the adult lifespan (Orth et al., 2012; Wagner, Gerstorf, et al., 2013). However, a recent
study indicated that in old age, people with moderate to severe physical health conditions as
diagnosed by physicians reported considerably stronger declines in self-esteem compared to
people without such physical health issues (Wagner, Hoppmann, et al., 2015). With this
background, we conclude that a more comprehensive investigation of physical health that moves
beyond simple self-reports may help us better understand associations between physical health
and self-esteem.
Social relationships. Following the interpersonal perspective, social relationships are of
particular importance for self-esteem. A natural point of departure for examining the perspective
are close and intimate relationships, because temporally stable and affectively pleasant
relationships are considered to be fundamental for satisfying the basic human need of
belongingness (Baumeister & Leary, 1995). As a consequence, self-esteem serves as a monitor
for the degree of social inclusion. Empirical research has indeed shown that being married or
having a partner is cross-sectionally correlated with higher levels of self-esteem (Wagner et al.,
2014). Longitudinally, several studies on young adults have demonstrated that establishing
intimate relationships (in certain time windows) is associated with increases in self-esteem
(Lehnart, Neyer, & Eccles, 2010; Neyer & Asendorpf, 2001; Wagner, Becker, Lüdtke, &
Trautwein, 2015). However, differences of this kind in rates of self-esteem change have not been
observed with a lifespan sample (Wagner et al., 2014). Whether or not associations between
romantic relationships and self-esteem generalize to the second half of life is still an open
question. For example, close relationships may be of even greater importance for self-esteem in
old age, when social network size typically decreases as a result of changes in social roles and
life circumstances (Wrzus, Hanel, Wagner, & Neyer, 2013), such as retirement and children
leaving the parental home.
Experience of marital or partnership dissolution, be it through divorce or the death of a
partner, may constitute another powerful source of individual differences in self-esteem.
Obviously, partnership dissolution in itself represents a threat of social isolation, because
individuals lose a potentially important source of support. Divorce and widowhood among older
adults influence social ties to other persons as well. For example, adult offspring of divorced
parents have less contact with their parents and perceive the quality of their relationship to
parents to be poorer than offspring of married parents do (Herlofson, 2013). Smaller social
networks as a direct or indirect consequence of partnership dissolution may, in turn, reduce self-
esteem, because these decreases in social networks will probably be accompanied by increased
feelings of social isolation. However, the available empirical evidence is scarce and inconclusive.
Of the few cross-sectional studies available, one report indicated that self-esteem is indeed higher
among people who are currently married than among persons who were previously married
(including both divorced and widowed; McMullin & Cairney, 2004), whereas another report
found no association between self-esteem and relationship dissolution through divorce or death of
a partner (Lee & Shehan, 1989).
Personality. Both theoretical perspectives predict that broad personality traits are of
importance for self-esteem. The intrapersonal perspective’s emphasis on abilities and resources
suggests that personality traits promoting successful performance in society are relevant. For
example, extraversion and openness have been proposed to promote personal growth and agency
(Digman, 1997; Paulhus & John, 1998) and may thus be important sources of self-esteem.
Conscientiousness is also relevant for the intrapersonal perspective, because conscientiousness
predicts highly valued achievements such as academic and job performances (Barrick, Mount, &
Judge, 2001; Poropat, 2009). Theoretical accounts also relate neuroticism to the intrapersonal
perspective, because neuroticism is usually conceptualized as individual differences in the
general ability to handle stress (Denissen & Penke, 2008). Stress and negative experiences may
have less negative effects on performance and evaluation of one’s own abilities among
emotionally stable individuals (i.e., among persons with low degrees of neuroticism) than among
emotionally unstable individuals, leading in turn to the maintenance of higher levels in self-
esteem, particularly under stressful life conditions.
Most Big Five personality traits are also relevant for the interpersonal perspective,
because they are important for facilitating or maintaining social ties. Agreeableness and
conscientiousness have been found important because these personality traits are associated with
motives of nurturance and close relationships (Paulhus & John, 1998). Moreover, extraversion
may as well improve social integration, because it promotes social behavior like dating and
attending parties (Paunonen, 2003). Neuroticism may also be of importance, because it has been
suggested that neuroticism is closely related to individual differences in sensitivity to signs of
social exclusion (Denissen & Penke, 2008). In this case, threats to social inclusion may be of
lesser importance in determining self-esteem for emotionally stable individuals than for
emotionally unstable persons.
Empirical research has found considerable cross-sectional associations between high self-
esteem and high levels of emotional stability, extraversion, and conscientiousness (Erol & Orth,
2011; Robins, Tracy, Trzesniewski, Potter, & Gosling, 2001; Wagner et al., 2014), whereas
associations are typically weaker for agreeableness and openness (Erol & Orth, 2011; Robins et
al., 2001; Wagner et al., 2014). To the best of our knowledge, only two studies have examined
how Big Five personality traits are longitudinally related to self-esteem, with conflicting
findings: One study concluded that there were by and large no prospective associations (Erol &
Orth, 2011), but another study found that all Big Five personality traits were related to changes in
self-esteem over time (Wagner, Lüdtke, Jonkmann, & Trautwein, 2013). More specifically, the
results of the latter study indicated that low neuroticism and high extraversion, openness,
agreeableness, and conscientiousness are related to higher self-esteem across time. The striking
differences between the two studies may be due to different analytical approaches, because one
study examined the unique contribution of each personality trait to changes in self-esteem,
adjusting for all other traits (Erol & Orth, 2011), whereas the other study did not perform such
adjustments (Wagner, Lüdtke, et al., 2013). More research is thus needed to examine the link
between personality and self-esteem more thoroughly. Moreover, studies examining middle
adulthood and old age may be of particular importance because both of the existing studies
covered only the adolescent and young adult years.
The Role of Age and Gender
It is well possible that the above noted associations between psychosocial factors and self-
esteem differ with age. Drawing from lifespan psychological notions (Baltes et al., 2006; Gerstorf
& Ram, 2013), Wagner et al. (2014) have argued that certain resources are probably more
important for solving age-related challenges in one phase of life, whereas other resources are
more important for solving age-related challenges in a later phase of life. For example, resources
related to education may be of greater importance for self-esteem in adolescence and young adult
years than in older ages, because educational achievements in this period may be particularly
relevant for future career and work life. In a similar vein, with physical health challenges
becoming increasingly frequent and severe, it is important for older adults to gradually dissociate
their self-esteem from the health domain and turn towards domains over which they have more
control (Wagner et al., 2014). Initial evidence from cross-sectional studies can be interpreted to
be consistent with this notion. In particular, associations of subjective health and of neuroticism
with self-esteem have been found to be weaker in older adults than in younger adults (Wagner et
al., 2014). In a similar vein, associations of SES with self-esteem have been found to follow an
inverse u-shaped curve, in that associations were stronger at around age 60 than earlier in life and
were weaker thereafter (Twenge & Campbell, 2002). However, systematic research on age
differences in associations between a comprehensive number of potential predictors and self-
esteem is sparse.
Associations with self-esteem may also differ by gender, because in accordance with
cultural norms, men and women may value specific resources and abilities differently. For
example, gender differences in social relationship variables are plausible, because research
indicates that men rely to a greater degree on their spouse for emotional intimacy and that women
in general have a broader network of intimate relationships (Hatch & Bulcroft, 1992). In line with
this notion, empirical reports have found that being single and experiencing relationship
dissolutions are more detrimental to men’s psychosocial adjustment then women’s (Cooney &
Dunne, 2001). In contrast, studies have typically not found that associations between education
and self-esteem differ by gender (Orth et al., 2015; Orth et al., 2010).
Concerning personality, agentic traits (e.g., extraversion), which are typically ascribed to
men, may be of greater importance for men’s self-esteem, whereas communal traits (e.g.,
agreeableness), which are typically ascribed to women, may be of greater importance for
women’s self-esteem (Eagly, 2009). Empirical research has indeed found evidence consistent
with these ideas. To illustrate, in a large-scale cross-sectional study (N > 300,000), Robins et al.
(2001) found that correlations between self-esteem and personality traits differ by gender, with
correlations with extraversion, conscientiousness, and openness being stronger among men, and
correlations with agreeableness and emotional stability being stronger among women. Even
though the gender differences were modest in size, the results indicate gender stereotypic
patterns, with communal traits having stronger associations with self-esteem among women and
agentic traits having stronger associations with self-esteem among men (Robins et al., 2001). We
note, however, that the available evidence on moderating effects of gender is scarce and limited
to a selected set of sociodemographic and personality variables.
The Present Study
In this report, we (i) examine mean-level change in self-esteem from midlife to very old
age, (ii) test the role of SES, health, social relationships, and personality factors for predicting
self-esteem levels and its change, and (iii) explore whether associations differ by age and gender.
To do so, we apply latent change score models to a two-wave longitudinal national study of 5,555
Norwegian adults initially aged 40 to 80 years. Moving one step ahead, our study targets the
nature and correlates of self-esteem in the second half of life (rather than in adolescence and
young adulthood), examines the unique predictive effects of a comprehensive set of conceptually
motivated antecedents and correlates (rather than a few variables only), and uses registry data
that reliably cover the lifetime history of several major life events, including disability, divorce,
and widowhood (as opposed to self-reports about current conditions only). In particular, we
combine self-report data on self-esteem and personality with registry-based information about
SES (education, income, unemployment), health problems (sick leave, lifetime history of
disability), and social relationships (cohabitating partner, lifetime history of divorce, lifetime
history of widowhood). We expect that factors from all four domains will be correlated with
levels of self-esteem. Moreover, as indicated by previous research, we expect health problems
and social relationships to predict changes in self-esteem, whereas associations of SES and
personality with changes in self-esteem are less probable. Associations between health problems
and self-esteem may be particularly prone to be attenuated with increasing age, whereas gender
differences in associations with self-esteem are expected for relationship variables and
personality traits.
Methods
Sample and Procedure
Data from the population-based Norwegian Life Course, Aging, and Generation
(NorLAG) study was used in this study (for a detailed account of data collection procedures, see
Slagsvold et al., 2012). Data was collected in two waves. At the first wave (T1) in 2002, a
representative sample of non-institutionalized adults aged 40 to 79, stratified by age and gender,
was drawn from 30 municipalities in Norway representing different geographic regions.
Respondents were initially contacted through telephone interviews, and then followed up by
using a postal questionnaire. Data from nationwide official registries was added after respondents
gave informed consent. Of 8,298 individuals who were asked to participate, 5,555 (67.0%) were
interviewed, and 4,149 completed the questionnaire (74.6% of those who were interviewed;
combined response rate 50.0%).
The second wave (T2) was conducted in 2007, when all those who had participated in T1
were asked to be interviewed by telephone again (aged 45 to 84 years at T2). Between T1 and T2,
265 persons had died, and another 25 had moved abroad, reducing the sample to be approached
to 5,269 individuals. The same data collection methods as at T1 were used, and 3,774 persons
participated in the telephone interviews (71.6% of those eligible to participate at T2). Moreover,
2,984 completed the T2 questionnaire (79.1% of those participating in the telephone interview).
In this study, all persons who participated at least in the telephone interview at T1 were included
(N = 5,555; Mage = 57.9 years; 51.4% women).
As reported in prior publications (Slagsvold et al., 2012), participation bias at T1 was
small concerning gender, place of living, and age. However, participation rates were markedly
higher among those with high education compared to those with lower education (Slagsvold et
al., 2012). To examine sample attrition over the study period, we conducted two sets of
selectivity analyses. First, we compared respondents at T1 who completed the questionnaire after
the telephone interview with those who only participated in the telephone interview. Respondents
did not differ significantly in age (p > .05); however, participants who completed the
questionnaire had more years of education compared to non-participants (Cohen’s d = 0.29, p
< .001), they reported slightly higher earnings (d = 0.10, p < .01) and slightly higher self-esteem
on the four items assessed by telephone interview (d = 0.10, p < .001), they were more often
women (52.5% versus 48.2%, p < .01), they were less likely to have received a disability pension
(14.6% versus 17.6%, p < .01), and they were more likely to have a live-in partner (72.4% versus
64.5%, p < .001). Second, we examined if and how respondents who remained in the study at
both data waves differed from those who were not available for reassessment at T2. Respondents
who stayed in the study were younger (d = 0.61, p < .001), had more years of education (d =
0.46, p < .001), had higher earnings (d = 0.23, p < .001), reported slightly higher self-esteem (d =
0.17, p < .001), and more often had a live-in partner (74.3% versus 63.2%, p < .001). Gender and
a history of disability pension were not significantly related to attrition (p > .05). Conjointly, our
analyses indicated selectivity effects of moderate size for age and education, as measured by
Cohen’s d; group differences for other variables were of minor size.
Measures
Self-esteem. The Rosenberg Self-Esteem Scale (RSES; Rosenberg, 1965) was used at
both data collection waves. This 10-item measure is designed to assess general feelings of self-
worth and is the most widely used measure of self-esteem (Blascovich & Tomaka, 1991).
Participants used a 5-point response scale, ranging from 1 (strongly disagree) to 5 (strongly
agree). The scale was translated into Norwegian and back-translated into English by independent
translators to ensure the adequacy of the Norwegian translation. Norwegian translations of the
RSES have been shown to have good psychometric properties, comparable to those of the
English original version (von Soest, 2005). In the present study, four of the items were assessed
by telephone interview, and the remaining six items were assessed by questionnaire. Internal
consistency was α = .80 at T1 and α = .82 at T2.
Personality. At T1, Big Five personality traits were assessed by a short version of the 5-
Personality Factors – adjective (5-PFa) instrument (Engvik, 1993), which was included in the
questionnaire. The instrument contains adjective-anchored bipolar items (e.g., “friendly –
unfriendly”; “extravert – introvert”), which are rated on 7-point scales. Each of the Big Five
personality traits extraversion, agreeableness, conscientiousness, neuroticism, and openness were
assessed by four items. The instrument was developed in Norwegian and has shown good validity
and reliability (Engvik, 1993). For a less complex presentation of the results, the neuroticism
subscale was reversed, such that this subscale indicated emotional stability instead of
neuroticism. In this study, internal consistency was α = .74 for emotional stability, α = .58 for
extraversion, α = .71 for conscientiousness, α = .69 for agreeableness, and α = .66 for openness.
The moderate reliabilities reflect the heterogeneity of the items selected to measure relatively
broad constructs and are comparable to other brief personality scales (Donnellan, Oswald, Baird,
& Lucas, 2006; Lang, John, Lüdtke, Schupp, & Wagner, 2011).
Register data. Self-report data were linked with four sets of data from nationwide official
registries, as provided by Statistics Norway. To begin with, age and gender were assessed by
register data. Second, as an indicator of SES, education was divided into five categories, ranging
from 1 (completed junior high school or lower) to 5 (completed higher university degree).
Because the general level of education in the population had increased in more recent birth
cohorts, we standardized education separately for each year of birth, such that education scores
represent the level of education for a respondent measured in standard scores, compared to all
other respondents born in the same year.
Respondents’ income in the year of T1 (i.e., 2002) was also assessed by register data (in
Norwegian kroner per year). Because income is related to respondents’ age, we standardized
income for each birth cohort, in the same way as education was standardized. The measure
obtained thus provided an indication of income compared to other respondents’ incomes within
the same birth year. We also assessed whether respondents were unemployed the year that they
participated at T1.
Third, as indicators of health problems, we obtained register data on whether respondents
had a history of receiving a disability pension from 1991 or later (yes/no). Moreover, we assessed
by register data whether the respondent was on sick leave the year of T1 for two weeks or more
at a stretch (yes/no).
Finally, relationship variables were also assed with register data. We used the information
on whether the respondents had a cohabitating partner at the time of T1 (yes/no). Based on
register data, we constructed a variable indicating whether the respondents at any time after 1974
had experienced a divorce (yes/no) or experienced widowhood (yes/no). Data on divorce and
widowhood history were only available for respondents who participated in the study at T2.
Statistical Analyses
We conducted our analyses in a structural equation modeling framework, using Mplus 7.3
(Muthén & Muthén, 2012). Missing data were accommodated using full information maximum
likelihood under the usual missing at random assumptions underlying longitudinal designs (Little
& Rubin, 1987; Preacher, Wichman, MacCallum, & Briggs, 2008; Schafer & Graham, 2002). We
note that the correlates included in our models represent attrition-informative variables and so
helped to accommodate longitudinal selectivity under the assumption that incomplete data were
missing at random (i.e., missingness may have been related to these variables; McArdle, 1994).
Analyses were carried out in three steps. First, because measurement invariance is an
important requirement for longitudinal analyses, we constructed latent self-esteem measurement
models and tested for measurement invariance. Second, to model longitudinal change, we
constructed latent change score models, based on latent self-esteem factors. Third, to examine
potential sources of self-esteem, we regressed initial level and change of self-esteem, as
measured by change score models, on potential predictors of self-esteem.
Measurement invariance. To test for measurement invariance, we constructed self-
esteem measurement models. More specifically, based on the 10 self-esteem items, three parcels
were constructed that were used as indicators for a latent self-esteem factor for T1 and T2,
respectively. Reducing indicators through parceling has been recommended as an approach that
provides superior tests of structural model parameters because the constructs are defined more
precisely (Little, 2013; Little, Rhemtulla, Gibson, & Schoemann, 2013). Because four of the
indicators were assessed by telephone, these four items were used to construct one parcel. The
other six items were randomly assigned to one of two additional parcels, consisting thus of three
items each. Parcels were constructed in an identical manner at T1 and T2.
We conducted confirmatory factor analyses to test for measurement invariance across
time points. We first tested for configural invariance following Widaman, Ferrer, and Conger
(2010) by constructing latent factors based on the three parcels for each wave, and factors were
modeled to correlate across time points. Error variances from identical parcels created at different
time points were allowed to correlate. Model fit was evaluated by inspecting χ2 statistics, CFI,
TLI, and RMSEA. Following good practice, CFI and TLI values of .95 or greater and RMSEA
values of .06 or lower were considered as indicating good fit (Hu & Bentler, 1999). For a
graphical representation of the model, see Figure 1A. The fit for this model was adequate (see
Table 1). Next, we tested for weak invariance by comparing the basic configural invariant model
with a model in which factor loadings were forced to be equal across time points. Finally, strong
invariance was tested by additionally constraining intercepts of the parcels to be equal across
time. All models yielded adequate model fit (see Table 2).
Differences in model fit were tested by χ2 differences tests. Results showed no statistically
significant difference in fit between the configural invariant and the weak invariant model (Δχ2
[2] = 4.61; p < .05), whereas a significantly worsened fit was obtained when comparing the weak
invariant model with the strong invariant model (Δχ2 [3] = 29.03; p > .05). However, because χ2
values depend on sample size and a large sample was used in this analysis (N > 5,000), even
minor differences between time points may have yielded statistically significant differences. We
therefore assessed changes in model fit by the test of a small difference in fit (MacCallum,
Browne, & Cai, 2006); this showed the model fit to be not significantly worse for the weak
invariant or the strong invariant model. The test thus supported the self-esteem measure to be
invariant between T1 and T2.
Latent change score models. After establishing time invariance in self-esteem, we
constructed latent change score models based on the latent self-esteem factors at T1 and T2 with
strong invariance. The models were estimated to provide an indication of mean-level change in
self-esteem from T1 to T2 (see McArdle, 2009). The first latent self-esteem factor was scaled
according to Ferrer, Baluerka, and Widaman’s (2008) suggestion to establish an approximate
standard metric, such that the variance of the latent self-esteem factor at T1 was set to 1, and the
mean of the factor was set to 0. Change and initial values of self-esteem could thus be interpreted
in terms of a standardized metric with a mean of 0 and a standard deviation of 1 relative to self-
esteem at T1. As proposed by McArdle (2009) and Selig and Preacher (2009), the latent change
score in this model was controlled for the intercept or level of self-esteem. Controlling change for
the initial value of self-esteem provides an estimate of change that is not confounded by initial
mean-level differences in self-esteem. By adjusting for initial values of self-esteem, we were able
to accommodate that associations between potential predictors of change in self-esteem were not
an artifact of a correlation between initial value and change in self-esteem (see von Soest &
Hagtvet, 2011, for a detailed discussion of this issue).
Predictors of self-esteem. Next, level and change in self-esteem were regressed on
predictors. In a first model, only gender, age, and polynomials of age were included as predictors,
to examine age trends and gender-specific trends for self-esteem. For our analyses, age was
divided by 10, so as to express the rate of change by decade and centered at age 60, near the
average age of the sample.
Finally, to examine potential sources of self-esteem, all potential predictor variables were
included simultaneously as predictors of initial level and change in self-esteem, together with
age, its polynomials, and gender. All continuous predictor variables except age were mean
centered, whereas all dichotomous variables were effect coded. We used a weighted effect coding
scheme, such that the effect indicators were mean centered (see Cohen, Cohen, West, & Aiken,
2003). For the variables unemployment and sick leave, persons who already had retired at T1
were scored 0 (the mean score), because sick leave and unemployment are not applicable to
retired individuals. As a consequence, parameter estimates refer to age 60 and indicate sample
averages (across all individuals) and the extent of differences associated with a particular variable
(rather than for a particular group).
We also tested for interaction effects of gender and age by including interaction terms as
predictors of initial level and change. Only interaction terms that showed significant associations
with initial level or slope of self-esteem were kept in the final model.
Results
Table 2 presents descriptive statistics and intercorrelations of all variables included in the
study. At both T1 and T2, self-esteem scores were substantially higher than the midpoint of the
scale (i.e., larger than 3), thereby showing that on average participants reported rather high self-
esteem. The correlation of r = .61 between the two measurement points of self-esteem indicated
considerable rank-order stability of self-esteem across five years between the two data collection
waves. Correlations between self-esteem and register data were generally small, whereas
personality variables correlated more strongly with self-esteem, with the strongest correlations
being with emotional stability.
Self-Esteem and Self-Esteem Development Across the Second Half of Life
To examine change in self-esteem, change score models based on the latent self-esteem
factors at T1 and T2 were estimated, as Figure 1B depicts. The fit for this model was good (χ2[9]
= 67.76, CFI = .99, TLI = .99, RMSEA = .034). As defined by the model, the mean of the level
of self-esteem at T1 was 0, with a standard deviation (SD) of 1. The mean change of self-esteem
from T1 to T2 was estimated to .06 (p < .001; SD = .32, p < .001), indicating that self-esteem
scores on average increased by .06 standard deviations between T1 and T2 when considering the
entire sample.
Next, we estimated how age and gender predicted both initial status of self-esteem at T1
and change in self-esteem from T1 to T2. For this purpose, initial level of self-esteem and latent
change scores for self-esteem were regressed on age and gender. Moreover, quadratic and cubic
terms for age were included to examine potential non-linear age trends (see Figure 1C). Also for
this model, fit was adequate (χ2[25] = 247.08, CFI = .98, TLI = .96, RMSEA = .040).
As Table 3 shows (Model 1), when age and gender were included in the regression
analysis, we found that age, age2, and gender predicted significantly the initial status of self-
esteem. Men had significantly higher self-esteem, and level of self-esteem changed non-linearly
with age. Figure 2A shows that self-esteem levels peaked at about age 50 and declined gradually
thereafter. The size of the reduction in self-esteem from the peak of self-esteem at age 50 to the
estimated mean level of age 80 was d = 0.51, thereby indicating that self-esteem declined on
average about half a standard deviation between ages 50 and 80. When predicting changes in
self-esteem, age2 and age3 were significant predictors, whereas the linear components of age and
gender were not (see Table 3). The significant adjusted mean of the change score of .07 (p < .01)
indicated that the estimated self-esteem score increased with .07 standard deviations in the five
years between T1 and T2 for those who were 60 years of age at T1. However, as Figure 2B
shows, change differed across age, with increases in self-esteem in the youngest age group, when
self-esteem increased with d = 0.21 between T1 and T2 for people aged 40. Mean levels of self-
esteem remained rather stable between ages 50 and 70, with a mean change close to 0, whereas
declines in self-esteem between T1 and T2 from about age 70 onwards were observed, with a
considerable decline in self-esteem of d = .50 from age 80 (T1) to age 85 (T2). We conducted
additional analyses in which an age × gender interaction term was included as predictor.
However, no significant interactions were found for self-esteem level and change (p > .05),
indicating that the age trends in self-esteem level and change did not vary according to gender
and that the initial gender differences thus remained constant.

In a final set of analyses, we additionally included all potential predictor variables
simultaneously into our models. Table 3 shows the results of these analyses (Model 2).
Concerning measures of SES, results showed that education, income, and unemployment were all
significantly related to the initial level of self-esteem, indicating that higher SES was associated
with higher self-esteem. None of the SES indicators showed significant associations with change
in self-esteem. In a similar vein, none of the tested interaction effects of SES with age and gender
were significant.
As an indicator of health problems, having received a disability pension was significantly
related to both lower levels of and smaller increases in self-esteem, compared to having never
received a disability pension. The association with the level of self-esteem was qualified by a
significant interaction effect with age. As Figure 3 shows, relative to people without disability,
those with disability pensions reported lower self-esteem in midlife, but not in old age. No
association of sick leave with self-esteem was found.
As an indicator of social relationships, having a cohabitating partner was related to both a
higher level of self-esteem and a stronger increase in self-esteem, as compared to people without
a live-in partner. This effect was not moderated by age or gender. A history of divorce or
widowhood was not related to either level or change in self-esteem.
As indicators of personality, high scores on extraversion, conscientiousness, and openness
were each related to higher levels of self-esteem, but no such relation was found for
agreeableness. In addition, a significant gender × openness interaction emerged. As Figure 4
shows, more open participants reported higher levels of self-esteem, but openness made a greater
difference for self-esteem among men than among women. None of the four personality traits
were related to changes in self-esteem.

Emotional stability evinced an even more complex pattern of associations with self-
esteem. First, we observed significant associations between emotional stability and self-esteem
level and significant interactions with age and gender. As Figure 5 shows, emotionally stable
participants reported higher self-esteem, but the differences in level of self-esteem between
persons high and low in emotional stability were more pronounced in midlife than in old age and
more pronounced in women than men. Second, emotional stability was also significantly
associated with changes in self-esteem. Emotional stability also interacted with age, indicating
that the rate of change in self-esteem was more positive (i.e., less of a decrease, if not an increase
in self-esteem) among the emotionally stable in midlife, but differences by emotional stability
disappeared with increasing age.
Discussion
In this study, we applied latent change score models to two-wave longitudinal self-esteem
data obtained as part of the population-based NorLAG study in Norway and examined the role of
self-report and registry-based predictors. Results revealed that self-esteem peaked at around age
50 and declined thereafter, particularly among those aged 70 and older. Most important for our
questions, lower SES, not having a cohabitating partner, and lifetime experiences of
unemployment and disability were each uniquely associated with a lower self-esteem level and/or
steeper self-esteem declines over the five-year study period. Over and above the registry-based
information, personality characteristics were also found to be relevant, with high levels of
emotional stability, extraversion, conscientiousness, and openness each being related to higher
levels of self-esteem. Emotionally stable participants also showed less pronounced declines in
self-esteem. Finally, we found that associations of self-esteem level with disability and emotional
stability were weaker with advancing age. Gender interactions indicated that emotional stability
was more strongly associated with self-esteem level among women than men, whereas openness
was more strongly associated with self-esteem level among men. Our discussion will consider
how the findings demonstrate the utility of registry-based information and suggest that physical
health, social relationships, and personality factors are in manifold ways uniquely associated with
self-esteem and its development later in life.
Self-Esteem and Self-Esteem Development Across the Second Half of Life
Age trends of self-esteem as found in this study are in accordance with trends found in
several longitudinal studies showing a peak of self-esteem in midlife, at around age 50, and then
a decrease in old age (Orth & Robins, 2014). Moreover, our study indicates considerable declines
in self-esteem in old age, as we have observed a rather large reduction in the level of self-esteem
of about half a standard deviation from age 50 to age 80, as well as accelerating longitudinal
reductions in self-esteem particularly from age 70 and thereon. Our results are thus in line with
studies reporting rather strong declines in self-esteem in old age (Orth et al., 2012; Orth et al.,
2010). However, we note that the empirical literature is divided, as other studies have reported
considerably smaller mean-level declines in self-esteem in old age (Orth et al., 2015; Wagner,
Gerstorf, et al., 2013; Wagner et al., 2014). According to Orth et al. (2015), the diversity of the
findings may be explained by cross-cultural differences, because studies reporting large drops in
self-esteem primarily stem from the United States (Orth et al., 2012; Orth et al., 2010), and
studies reporting minor decreases used samples from Germany (Orth et al., 2015; Wagner et al.,
2014) or Australia (Wagner, Gerstorf, et al., 2013). However, there are considerable similarities
in cultural values and health care systems between a Northern European country such as Norway
and Germany. The greater decreases in self-esteem in old age in our study compared to German
samples may therefore not be easily explained by cultural differences. The nature and causes of
differential trajectories of self-esteem in old age should be examined in greater detail in future
studies.
Socioeconomic status. Extending earlier reports that made use of self-reported SES, our
study utilized registry-based information on key SES indicators: education, income, and
unemployment. The results of our study are largely in accordance with previous research: High
SES was consistently associated with high levels of self-esteem, as was reported in meta-analytic
studies (Twenge & Campbell, 2002). Moreover, SES did not predict changes in self-esteem in
our study, which is again in line with results from previous prospective research (Orth et al.,
2012; Orth et al., 2010; Wagner et al., 2014).
The temporal ordering of associations between SES and self-esteem remains unclear,
however: Even though the literature in the field typically proposes that SES is an antecedent of
self-esteem, it may also be possible that high self-esteem influences SES (Twenge & Campbell,
2002). Because the size of associations between SES and self-esteem is small, studies examining
the mechanisms through which SES is related to self-esteem may be underpowered, even though
large samples were used both in our study and in other longitudinal studies examining
prospective SES–self-esteem associations.
Physical health. Concerning physical health, a history of receiving a disability pension
was related to both the level of and changes in self-esteem, thereby supporting findings from a
recent longitudinal study reporting that moderate to severe physical health conditions as
diagnosed by physicians predicted stronger decreases in self-esteem in late life (Wagner,
Hoppmann, et al., 2015). In our study, recent sick leaves were not related to either the level of or
change in self-esteem. This may indicate that short-term variations in physical health, which are
typically measured by recent histories of sick leaves, are not associated with self-esteem, whereas
far-reaching, long-term health conditions, which are typically indicated by receiving disability
pensions, in fact show associations with poor self-esteem.

Social relationships. Our results extend earlier reports on the relevance of social factors
that have primarily drawn from samples in early adulthood to midlife and old age. Our study
indicates the importance of having a cohabitating partner in middle and late adulthood. A
cohabiting partner was associated with both higher levels of and more favorable changes in self-
esteem. Of interest is that a history of divorce and widowhood was not associated with self-
esteem over and above current partner status. One way to interpret the finding is that current
relationship status is of greater importance than people’s dissolution history. This interpretation is
in accordance with empirical studies reporting that relationship dissolution is indeed followed by
a period of pragmatic, physical, and emotional problems; however, the problems typically
subside eventually, as persons often make significant adjustments to their new life circumstances
(Chase-Lansdale & Hetherington, 1990). In the long run, then, current relationship status may be
of greater importance for one’s own self-esteem than people’s histories of relationship
dissolutions.
Personality. All personality traits except agreeableness showed associations with level of
self-esteem. In particular, high levels of self-esteem were associated with a mature personality
profile consisting of high scores on emotional stability, conscientiousness, and extraversion,
which provide individuals with adequate capacities to be productive and involved contributors to
society (Caspi, Roberts, & Shiner, 2005). This pattern is in accordance with other studies
showing associations between personality and levels of self-esteem (Erol & Orth, 2011; Robins et
al., 2001; Wagner et al., 2014).
As to predicting changes in self-esteem, only emotional stability showed significant
associations; this diverges somewhat from previous studies in which all Big Five traits were
related to changes in self-esteem (Wagner, Lüdtke, et al., 2013). The conflicting results may be
due to different analytical approaches, because we included all personality traits simultaneously
as predictors of self-esteem to assess their unique predictive capabilities, whereas Wagner et al.
(2013) examined personality–self-esteem associations for each trait separately. We also note that
all earlier reports examined adolescents and young adults, whereas our study assessed
personality–self-esteem associations in the second half of life. Because our study found age
differences in associations between emotional stability and changes in self-esteem, prospective
associations between these two variables may be specific to particular age periods.
Our study provides novel information that could inform theoretical perspectives on
sources of self-esteem. By and large, our results on predictors of level of self-esteem are in
accordance with expectations from both the intrapersonal and the interpersonal perspective, in
that we found unique associations of all broad factors of SES, physical health, social
relationships, and most personality traits with levels of self-esteem.
The results are less conclusive with respect to associations with changes in self-esteem.
Overall, fewer associations with changes in self-esteem were found, yet there is some further
support for the interpersonal perspective based on the findings of partnership status. This
relationship shows the importance of long-term, close relationships for self-esteem development
across middle and late adulthood. Associations of emotional stability with changes in self-esteem
provide additional support for this perspective, being in line with notions suggesting that high
emotional stability desensitizes people to being affected by signs of social exclusion (Denissen &
Penke, 2008). The finding is also in accordance with the intrapersonal perspective, because high
emotional stability probably makes persons also less vulnerable to other negative experiences and
losses that accompany later life. Finally, a history of having received a disability pension was
also associated with changes in self-esteem. Physical health problems may hinder individuals
from living up to their own expectations in important domains of life and may increase the risk of
social marginalization. Accordingly, this finding could support both the intrapersonal and
interpersonal perspectives. However, the lack of associations of self-esteem change with a more
comprehensive set of potential predictors – including indicators of SES – appears to limit the
explanatory value of both perspectives, and particularly the intrapersonal perspective.
The Role of Age and Gender
Two cross-sectional interactions with age emerged, in that associations of self-esteem
level with disability and emotional stability were reduced with advancing age. These reductions
are in accordance with a resource perspective (Baltes et al., 2006). Older adults may have better
chances to maintain self-esteem if they are capable of increasingly dissociating themselves from
resources that typically decrease in old age, such as physical health (Wagner et al., 2014). The
age interaction effects are also in accordance with a recent study reporting similar decreases of
associations of emotional stability and subjective health with self-esteem in older age (Wagner et
al., 2014). An intriguing aspect of our findings is that associations between emotional stability
and change in self-esteem also decreased in size with increasing age. This finding supports the
notion that aspects of emotional stability may be less likely to affect the evaluation of the self
among old adults.
Gender interactions were limited to personality traits, and our results are largely in
agreement with previous findings from Robins et al. (2001). We also found a gender stereotypic
pattern, with openness being more strongly correlated with self-esteem level among men and
emotional stability being more strongly correlated with self-esteem level among women in cross-
sectional analyses. Contrary to our expectations, we did not find significant gender interactions
with social relationship variables. Even though research indicates that being single or divorced is
more detrimental for men’s psychosocial adjustment than for women’s (Cooney & Dunne, 2001),
single, divorced, and widowed men may experience fewer financial problems than their female
counterparts, which in turn may counteract potential gender differences.

Limitations and Outlook
With this study, we were in the unique position to utilize long-term longitudinal data from
a relatively large population-representative national sample, making use of the extensively used
Rosenberg Self-Esteem Scale, and linking self-report data with information obtained from
registry records. At the same time, we acknowledge several limitations of our measures, study
design, and sample. Concerning limitations in measures, we measured Big Five personality traits
with an instrument that has been used in Norway only, and some of the personality traits
measured had somewhat low reliability. Even though the scale has been shown to have good
validity (Engvik, 1993) and reliability is comparable to other Big Five scales with a similar
number of items (Donnellan et al., 2006), a more widely used instrument assessing the Big Five
personality traits would have been an advantage so as to compare findings directly with reports
from other studies. Moreover, factors not assessed in the present study could provide additional
theoretical insights. For example, the intrapersonal perspective emphasizes that people’s own
perception of their performance is an important source of self-esteem. However, this study only
assessed outcomes that are typically desirable in our society but did not assess how individuals
themselves evaluate their performance or their perceptions of the importance of their
performance in different areas. In future research, it may also be feasible to use registry-based
information about the end of life, so as to examine more thoroughly how self-esteem changes in
the last years of life and what role presumably relevant factors, such as place of death and cause
of death, play (for an overview, see Gerstorf & Ram, 2013). We also note that it would be highly
intriguing to examine whether our findings regarding self-esteem change and predictors thereof
would also generalize to rank-order stability. For example, self-esteem may fluctuate more
among emotionally unstable than among stable individuals, because self-esteem may be
determined to a larger degree by external stressors among persons with emotional instability.
Predictors of rank-order stability of self-esteem should thus be examined more closely in future
studies.
Second, as limitations of our study design, predictors of self-esteem were assessed at one
time point only; as a consequence, longitudinal analyses were limited to examining how
predictors at the first time point were associated with change in self-esteem, whereas it was not
possible to examine how changes in predictors were related to changes in self-esteem. For
example, it was not possible to examine how changes in indicators of SES – such as income and
unemployment – were related to changes in self-esteem, even though such analyses would
provide valuable information about the longitudinal association of these two concepts. Moreover,
we note that the size and strengths of associations found for the socioeconomic, physical health,
social relationships, and personality factors reported here may be specific to the five-year time
scale. It is quite possible that the relevance of these variables is different or that other factors may
emerge to be important when associations are considered over shorter time intervals, such as
fluctuations from one situation to the next or one day to the next (Gerstorf, Hoppmann, & Ram,
2014). For example, although in our study sick leaves were not related to changes in self-esteem
over a long time span such as five years, it is quite possible that shorter-term sick leaves may
predict temporary reductions in self-esteem. Our study design is also limited in that only two
waves of data were available. As a result, it was not possible, for example, to examine complex
longitudinal change patterns (e.g., non-linear change) in how self-esteem develops across the
second half of life.
As limitations of the sample, selective participation may have biased our results, because
no institutionalized older adults were included in the sample and persons with low educational
status were underrepresented. Likewise, we acknowledge selective attrition as another limitation,
even though we used contemporary missing data routines to reduce the impact of selective
attrition on the results. It remains to be seen whether the findings obtained here generalize to less
positively select segments of the general population. For example, it is possible that emotional
stability is particularly important for self-esteem under the challenging living conditions of
population segments with low SES. As another sample limitation, the nature and correlates of
change in self-esteem may have been shaped by the specific historical time monitored. Although
research has shown that such cohort differences in levels and rates of change in self-esteem
appear negligible (see Orth & Robins, 2014), it will also be intriguing to see what happens with
self-esteem trajectories when its presumed antecedents and correlates change historically (e.g.,
better cognitive performance and psychosocial functioning among 75-year-olds nowadays than
among same-aged peers two decades ago: Gerstorf et al., 2015).
Conclusions
In conclusion, this study provides novel information on registry and self-report predictors
of self-esteem level and change in the second half of life. The study provides some support for
both intrapersonal and interpersonal perspectives; the associations of self-esteem levels found are
by and large in agreement with both theoretical frameworks. However, results concerning
predictors of self-esteem change provided more mixed support. Only partnership status,
emotional stability, and receiving a disability pension were predictors of self-esteem change,
whereas other factors did not show associations. This study thus indicates that both perspectives
may be of some importance for self-esteem development in middle and late adulthood; however,
the role of these perspectives may be limited to certain life domains.
Future longitudinal studies corroborating findings from this study are needed. The
reduced associations with self-esteem in old age observed for health variables and emotional
stability/neuroticism also deserve future research attention: Knowledge about mechanisms that
are at play in these reduced associations may inform intervention studies that aim to boost self-
esteem in vulnerable groups by dissociating people’s own self-evaluations from biopsychosocial
characteristics, such as poor physical health or low social status.

References
Baltes, P. B., Lindenberger, U., & Staudinger, U. M. (2006). Life span theory in developmental
psychology. In W. Damon & R. M. Lerner (Eds.), Handbook of child psychology: Vol. 1.
Theoretical models of human development (6th ed., pp. 569–664). New York: Wiley.
Barrick, M. R., Mount, M. K., & Judge, T. A. (2001). Personality and performance at the
beginning of the new millennium: What do we know and where do we go next?
International Journal of Selection and Assessment, 9, 9-30. doi:10.1111/1468-2389.00160
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal
attachments as a fundamental human motivation. Psychological Bulletin, 117, 497-529.
doi:10.1037/0033-2909.117.3.497
Blascovich, J., & Tomaka, J. (1991). Measures of self-esteem. In J. P. Robinson, P. R. Shaver &
L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (pp.
115-160). San Diego, CA: Academic Press.
Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and change.
Annual Review of Psychology, 56, 453-484.
doi:10.1146/annurev.psych.55.090902.141913
Chase-Lansdale, P. L., & Hetherington, E. M. (1990). The impact of divorce on life-span
development: Short and long term effects. In P. B. Baltes, D. L. Featherman & R. M.
Lerner (Eds.), Life-span development and behavior: Vol. 10 (pp. 105-150). New York:
Psychology Press.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Erlbaum.
Cooley, C. H. (1902). Human nature and the social order. New York: Charles Scribner's Sons.
Cooney, T. M., & Dunne, K. (2001). Intimate relationships in later life. Current realities, future
prospects. Journal of Family Issues, 22, 838-858. doi:10.1177/019251301022007003
Denissen, J. J. A., & Penke, L. (2008). Motivational individual reaction norms underlying the
Five-Factor model of personality: First steps towards a theory-based conceptual
framework. Journal of Research in Personality, 42, 1285-1302.
doi:10.1016/j.jrp.2008.04.002
Donnellan, M. B., Oswald, F. L., Baird, B. M., & Lucas, R. E. (2006). The Mini-IPIP scales:
Tiny-yet-effective measures of the big five factors of personality. Psychological
Assessment, 18, 192-203. doi:10.1037/1040-3590.18.2.192
Eagly, A. H. (2009). The his and hers of prosocial behavior: An examination of the social
psychology of gender. American Psychologist, 64, 644-658. doi:10.1037/0003-
066X.64.8.644
Engvik, H. (1993). "Big Five" på norsk [Big Five in Norwegian]. Journal of the Norwegian
Psychological Association, 30, 884-896.
Erol, R. Y., & Orth, U. (2011). Self-esteem development from age 14 to 30 years: A longitudinal
study. Journal of Personality and Social Psychology, 101, 607-619.
doi:10.1037/a0024299
Ferrer, E., Balluerka, N., & Widaman, K. F. (2008). Factorial invariance and the specification of
second-order latent growth models. Methodology, 4, 22-36. doi:10.1027/1614-
2241.4.1.22
Gerstorf, D., Hoppmann, C. A., & Ram, N. (2014). The promise and challenges of integrating
multiple time-scales in adult developmental inquiry. Research in Human Development,
11, 75-90. doi:10.1080/15427609.2014.906725

Gerstorf, D., Hülür, G., Drewelies, J., Eibich, P., Duezel, S., Demuth, I., . . . Lindenberger, U.
(2015). Secular changes in late-life cognition and well-being: Towards a long bright
future with a short brisk ending? Psychology and Aging, 30, 301-310.
doi:10.1037/pag0000016
Gerstorf, D., & Ram, N. (2013). Inquiry into terminal decline: Five objectives for future study.
Gerontologist, 53, 727-737. doi:10.1093/geront/gnt046
Hatch, L. R., & Bulcroft, K. (1992). Contact with friends in later life: Disentangling the effects of
gender and marital status. Journal of Marriage and Family, 54, 222-232.
doi:10.2307/353289
Herlofson, K. (2013). How gender and generation matter: Examples from research on divorced
parents and adult children. Families, Relationships and Societies, 2, 43-60.
doi:10.1332/204674313X664699
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.
doi:10.1080/10705519909540118
James, W. (1890). The principles of psychology. New York: Henry Holt.
Lang, F. R., John, D., Lüdtke, O., Schupp, J., & Wagner, G. G. (2011). Short assessment of the
Big Five: Robust across survey methods except telephone interviewing. Behavior
Research Methods, 43, 548-567. doi:10.3758/s13428-011-0066-z
Leary, M. R., Tambor, E. S., Terdal, S. K., & Downs, D. L. (1995). Self-esteem as an
interpersonal monitor: The sociometer hypothesis. Journal of Personality and Social
Psychology, 68, 518-530. doi:10.1037/0022-3514.68.3.518
Lee, G. R., & Shehan, C. L. (1989). Social relations and the self-esteem of older persons.
Research on Aging, 11, 427-442. doi:10.1177/0164027589114002

Lehnart, J., Neyer, F. J., & Eccles, J. (2010). Long-term effects of social investment: The case of
partnering in young adulthood. Journal of Personality, 78, 639-670. doi:10.1111/j.1467-
6494.2010.00629.x
Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data. New York: Wiley.
Little, T. D. (2013). Longitudinal structural equation modeling. New York: Guilford.
Little, T. D., Rhemtulla, M., Gibson, K., & Schoemann, A. M. (2013). Why the items versus
parcels controversy needn’t be one. Psychological Methods, 18, 285-300.
doi:10.1037/a0033266
MacCallum, R. C., Browne, M. W., & Cai, L. (2006). Testing differences between nested
covariance structure models: Power analysis and null hypotheses. Psychological Methods,
11, 19-35. doi:10.1037/1082-989X.11.1.19
McArdle, J. J. (1994). Structural factor analysis experiments with incomplete data. Multivariate
Behavioral Research, 29, 409-454. doi:10.1207/s15327906mbr2904_5
McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal
data. Annual Review of Psychology, 60, 577-605.
doi:10.1146/annurev.psych.60.110707.163612
McMullin, J. A., & Cairney, J. (2004). Self-esteem and the intersection of age, class, and gender.
Journal of Aging Studies, 18, 75-90. doi:10.1016/j.jaging.2003.09.006
Mead, G. H. (1934). Mind, self, and society. Chicago, IL: University of Chicago Press.
Muthén, L. K., & Muthén, B. O. (2012). Mplus user's guide (7th ed.). Los Angeles, CA: Muthén
& Muthén.
Neyer, F. J., & Asendorpf, J. B. (2001). Personality-relationship transaction in young adulthood.
Journal of Personality and Social Psychology, 81, 1190-1204. doi:10.1037/0022-
3514.81.6.1190
Orth, U., Maes, J., & Schmitt, M. (2015). Self-esteem development across the life span: A
longitudinal study with a large sample from Germany. Developmental Psychology, 51,
248-259. doi:10.1037/a0038481
Orth, U., & Robins, R. W. (2014). The development of self-esteem. Current Directions in
Psychological Science, 23, 381-387. doi:10.1177/0963721414547414
Orth, U., Robins, R. W., & Widaman, K. F. (2012). Life span development of self-esteem and its
effects on important life outcomes. Journal of Personality and Social Psychology, 102,
1271-1288. doi:10.1037/a0025558
Orth, U., Trzesniewski, K. H., & Robins, R. W. (2010). Self-esteem development from young
adulthood to old age: A cohort-sequential longitudinal study. Journal of Personality and
Social Psychology, 98, 645-658. doi:10.1037/a0018769
Paulhus, D. L., & John, O. P. (1998). Egoistic and moralistic biases in self-perception: The
interplay of self-deceptive styles with basic traits and motives. Journal of Personality, 66,
1025-1060. doi:10.1111/1467-6494.00041
Paunonen, S. V. (2003). Big Five factors of personality and replicated predictions of behavior.
Journal of Personality and Social Psychology, 84, 411-424. doi:10.1037/0022-
3514.84.2.411
Poropat, A. E. (2009). A meta-analysis of the five-factor model of personality and academic
performance. Psychological Bulletin, 135, 322-338. doi:10.1037/a0014996
Preacher, K. J., Wichman, A. L., MacCallum, R. C., & Briggs, N. E. (2008). Latent growth curve
modeling. Thousand Oaks, CA: Sage.
Roberts, B. W., Wood, D., & Smith, J. L. (2005). Evaluating Five Factor Theory and social
investment perspectives on personality trait development. Journal of Research in
Personality, 39, 166-184. doi:10.1016/j.jrp.2004.08.002

Robins, R. W., Tracy, J. L., Trzesniewski, K., Potter, J., & Gosling, S. D. (2001). Personality
correlates of self-esteem. Journal of Research in Personality, 35, 463-482.
doi:10.1006/jrpe.2001.2324
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University
Press.
Rosenberg, M., Schooler, C., Schoenbach, C., & Rosenberg, F. (1995). Global self-esteem and
specific self-esteem: Different concepts, different outcomes. American Sociological
Review, 60, 141-156. doi:10.2307/2096350
Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art.
Psychological Methods, 7, 147-177. doi:10.1037//1082-989X.7.2.147
Selig, J. P., & Preacher, K. J. (2009). Mediation models for longitudinal data in developmental
research. Research in Human Development, 6, 144-164.
doi:10.1080/15427600902911247
Slagsvold, B., Veenstra, M., Daatland, S. O., Hagestad, G., Hansen, T., Herlofson, K., . . . Solem,
P. E. (2012). Life-course, ageing and generations in Norway: The NorLAG study.
Norwegian Journal of Epidemiology, 22, 95-102. doi:10.5324/nje.v22i2.1554
Trzesniewski, K. H., Donnellan, M. B., Moffitt, T. E., Robins, R. W., Poulton, R., & Caspi, A.
(2006). Low self-esteem during adolescence predicts poor health, criminal behavior, and
limited economic prospects during adulthood. Developmental Psychology, 42, 381-390.
doi:10.1037/0012-1649.42.2.381
Twenge, J. M., & Campbell, W. K. (2002). Self-esteem and socioeconomic status: A meta-
analytic review. Personality and Social Psychology Review, 6, 59-71.
doi:10.1207/s15327957pspr0601_3
von Soest, T. (2005). Rosenbergs selvfølelsesskala: Validering av en norsk oversettelse [The
Rosenberg Self-Esteem Scale: Validation of a Norwegian translation]. Journal of the
Norwegian Psychological Association, 42, 226-228.
von Soest, T., & Hagtvet, K. A. (2011). Mediation analysis in a latent growth curve modeling
framework. Structural Equation Modeling, 18, 289-314.
doi:10.1080/10705511.2011.557344
Wagner, J., Becker, M., Lüdtke, O., & Trautwein, U. (2015). The first partnership experience and
personality development: A propensity score matching study in young adulthood. Social
Psychological and Personality Science, 6, 455-463. doi:10.1177/1948550614566092
Wagner, J., Gerstorf, D., Hoppmann, C., & Luszcz, M. A. (2013). The nature and correlates of
self-esteem trajectories in late life. Journal of Personality and Social Psychology, 105,
139-153. doi:10.1037/a0032279
Wagner, J., Hoppmann, C., Ram, N., & Gerstorf, D. (2015). Self-esteem is relatively stable late
in life: The role of resources in the health, self-regulation, and social domains.
Developmental Psychology, 51, 136-149. doi:10.1037/a0038338
Wagner, J., Lang, F. R., Neyer, F. J., & Wagner, G. G. (2014). Self-esteem across adulthood: The
role of resources. European Journal of Ageing, 11, 109-119. doi:10.1007/s10433-013-
0299-z
Wagner, J., Lüdtke, O., Jonkmann, K., & Trautwein, U. (2013). Cherish yourself: Longitudinal
patterns and conditions of self-esteem change in the transition to young adulthood.
Journal of Personality and Social Psychology, 104, 148-163. doi:10.1037/a0029680
Widaman, K. F., Ferrer, E., & Conger, R. D. (2010). Factorial invariance within longitudinal
structural equation models: Measuring the same construct across time. Child Development
Perspectives, 4, 10-18. doi:10.1111/j.1750-8606.2009.00110.x

Wrzus, C., Hanel, M., Wagner, J., & Neyer, F. J. (2013). Social network changes and life events
across the life span: A meta-analysis. Psychological Bulletin, 139, 53-80.
doi:10.1037/a0028601
Table 1.
Fit Indices for Measurement Models with Increasing Degree of Invariance Across Time
χ2 df CFI TLI RMSEA 90% CI RMSEA

Configural invariance 50.72 5 1.00 .98 .041 .031 – .051
Weak invariance 55.33 7 .99 .99 .035 .027 – .044
Strong invariance 84.36 10 .99 .99 .037 .030 – .044
Note. N = 5,555. df = degrees of freedom; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; RMSEA = root mean
square error of approximation. 90 % CI RMSEA = 90 % confidence interval of RMSEA.
Table 2.
Descriptive Statistics for the Variables Under Study
Intercorrelations
Parameter M SD N (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
Self-esteem (range 1-5)
(1) Self-esteem T1* 4.08 0.61 5,554
(2) Self-esteem T2* 4.14 0.59 3,554 .61
Socio-demographics
(3) Age (range 40–80) 57.91 11.11 5,555 -.10 -.11
(4) % women 51.4 5,555 -.09 -.08 -.01
Register data
(5) Education (in years) 11.62 2.76 5,508 .11 .09 .01 -.07
(6) Income 293.53 269.67 5,555 .16 .13 .01 -.27 .38
(7) % unemployment 6.7 4,101 -.07 -.06 -.06 -.02 -.07 -.10
(8) % sick leave 19.5 4,101 -.01 -.01 -.03 .01 -.07 -.02 .03
(9) % history disability 15.4 5,555 -.12 -.12 .02 .07 -.18 -.22 -.03 -.08
(10) % cohabitating partner 70.4 5,514 .09 .10 -.16 -.14 .02 -.01 -.10 .00 -.04
(11) % history divorce 24.4 3,562 .03 .02 -.07 .04 .01 .04 .06 .02 .08 -.32
(12) % history widowhood 4.0 3,559 -.08 -.07 .35 .13 -.06 .04 -.01 -.02 -.04 -.39 -.09
Personality (range 1-7)
(13) Emotional stability 4.77 1.26 4,100 .49 .40 -.02 -.15 .01 .09 -.03 -.04 -.10 .06 -.01 -.01
(14) Extraversion 4.87 1.17 4,107 .32 .28 -.03 .12 -.01 .01 -.06 .00 .02 .01 .03 .01 .28
(15) Conscientiousness 5.02 1.01 4,093 .27 .22 -.02 .04 .08 .10 -.06 -.01 -.05 .03 .01 .01 .22 .25
(16) Agreeableness 5.65 0.94 4,109 .17 .17 .01 .23 -.06 -.11 -.03 .01 .05 -.05 .03 .08 .21 .40 .32
(17) Openness 4.87 0.87 4,096 .27 .23 -.04 -.01 -.01 .03 -.02 -.01 .04 .02 .04 -.01 .26 .36 .32 .36
Note. Intercorrelations of r = .06 or above are statistically significantly different from zero at p < .001. Continuous measures are
presented in their original metric. N = number of participants. Income in 1,000 Norwegian kroner (NOK). *Descriptive statistics of
self-esteem scores are based on mean scores of all 10 items of the Rosenberg Self-Esteem Scale. However, all analyses in this paper
are based on latent self-esteem scores.
Table 3.
Standardized Beta Coefficients from Regression Analyses Predicting Initial Status and
Subsequent Change in Self-Esteem
Model 1 Model 2
Age and gender included All predictors included
Parameter Initial status Change Initial status Change
Intercepts (adjusted means) .00 .07*** .00 .08**
Socio-demographic variables
Age –.21*** .09 –.14*** .11
Age2 –.05*** –.11* –.03 –.15**
Age3 .06 –.27** .06 –.33***
Female gender –.10*** .01 –.02 .02
Register data
Education .10*** .05
Income .08*** .02
Unemployment –.03* .01
Sick leave –.01 –.01
Disability –.06*** –.06*
Disability × age .04** –
Cohabitating partner .04* .06*
History of divorce .03 .01
History of widowhood –.04 .01
Personality
Extraversion .17*** .05
Openness .11*** .04
Openness × gender –.04* –
Conscientiousness .12*** .03
Agreeableness –.01 .04
Emotional stability .37*** .09*
Emotional stability × age –.06*** –.07*
Emotional stability × gender .06*** –
Note. Included were all participants with valid self-esteem data at T1 (N = 5,555).
* p < .05; ** p < .01; *** p < .001.
Figure 1. Latent difference score models for self-esteem as estimated in this study.
Figure 2. Initial level of self-esteem (A) and change in self-esteem (B) by age and gender.
Standardized scores of self-esteem (z scores) are presented. Self-esteem peaked at around age 50 and
declined thereafter, particularly among those aged 70 and older. Men reported higher self-esteem by
about 0.19 SD, but men and women changed in parallel. Gender differences thus remained constant.
Figure 3. Initial level of self-esteem (z scores) for people with and without a history of receiving
a disability pension. Relative to people without disability, those with disability pensions reported
lower self-esteem in midlife, but not in old age.

Figure 4. Initial level of self-esteem (z scores) by gender and openness. More open participants
reported higher self-esteem, but openness made a greater difference for self-esteem among men
than among women.

Figure 5. Initial level of self-
esteem (A) and change in self-esteem (B) by gender and emotional stability. Standardized scores
of self-esteem (z scores) are presented. Emotionally stable participants reported higher self-
esteem, but differences in level of self-esteem between persons high and low in emotional
stability were more pronounced in midlife than in old age and more pronounced in women than
in men. Emotional stability also interacted with age, indicating that the rate of change in self-
esteem was more positive (i.e., less of a decrease, if not an increase in self-esteem) among the
emotionally stable in midlife, but differences by emotional stability disappeared with increasing
age.

Self-Esteem Across The Second Half of Life The Role of Socioeconomic Status, Physical

Uploaded by

Copyright:

Available Formats

Self-Esteem Across The Second Half of Life The Role of Socioeconomic Status, Physical

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Self-Esteem Across The Second Half of Life The Role of Socioeconomic Status, Physical

Uploaded by

Copyright:

Available Formats

────────────────────────────────────────────────────────────

This is an author generated postprint of the article:

Running Head: SELF-ESTEEM ACROSS THE SECOND HALF OF LIFE

Tilmann von Soest

personality with registry-based information on socioeconomic status (education, income,

pronounced declines in self-esteem. Moreover, associations of disability and of emotional

associated with self-esteem and its development later in life.

Keywords: self-esteem, register data, old age, physical health, relationships

Social Relationships, and Personality Factors

Self-esteem is defined as an individual’s general attitude towards or evaluation of the self

a fundamental psychological construct has been emphasized by research showing self-esteem to

age and gender?

Self-Esteem and Its Development Across the Second Half of Life

Theoretical accounts in personality and developmental psychology, including social

Predictors of Level and Change in Self-Esteem

influence self-esteem. First, according to the intrapersonal perspective, as originally proposed by

shaping self-esteem. This perspective largely considers self-esteem to be a result of

internalizations of others’ perceptions and evaluations of oneself. A prominent contemporary

presented in the following.

Socioeconomic status. Following the intrapersonal perspective, SES is an evident

al., 2010; Wagner et al., 2014).

interpersonal perspective. These notions have received support in that cross-sectional

diagnosed by physicians reported considerably stronger declines in self-esteem compared to

Social relationships. Following the interpersonal perspective, social relationships are of

belongingness (Baumeister & Leary, 1995). As a consequence, self-esteem serves as a monitor

leaving the parental home.

Experience of marital or partnership dissolution, be it through divorce or the death of a

partner, may constitute another powerful source of individual differences in self-esteem.

Obviously, partnership dissolution in itself represents a threat of social isolation, because

a partner (Lee & Shehan, 1989).

Conscientiousness is also relevant for the intrapersonal perspective, because conscientiousness

perspective, because neuroticism is usually conceptualized as individual differences in the

esteem, particularly under stressful life conditions.

suggested that neuroticism is closely related to individual differences in sensitivity to signs of

emotionally unstable persons.

covered only the adolescent and young adult years.

The Role of Age and Gender

differences in associations between a comprehensive number of potential predictors and self-

to a selected set of sociodemographic and personality variables.

The Present Study

levels of self-esteem. Moreover, as indicated by previous research, we expect health problems

Sample and Procedure

was drawn from 30 municipalities in Norway representing different geographic regions.

combined response rate 50.0%).

(N = 5,555; Mage = 57.9 years; 51.4% women).

As reported in prior publications (Slagsvold et al., 2012), participation bias at T1 was

questionnaire had more years of education compared to non-participants (Cohen’s d = 0.29, p

Cohen’s d; group differences for other variables were of minor size.

consistency was α = .80 at T1 and α = .82 at T2.

questionnaire. The instrument contains adjective-anchored bipolar items (e.g., “friendly –

personality traits extraversion, agreeableness, conscientiousness, neuroticism, and openness were

other respondents born in the same year.

Third, as indicators of health problems, we obtained register data on whether respondents

important requirement for longitudinal analyses, we constructed latent self-esteem measurement

potential sources of self-esteem, we regressed initial level and change of self-esteem, as

measured by change score models, on potential predictors of self-esteem.

Measurement invariance. To test for measurement invariance, we constructed self-

items each. Parcels were constructed in an identical manner at T1 and T2.

We conducted confirmatory factor analyses to test for measurement invariance across