Early Life Circumstances and Later-Life Loneliness in Ireland
Kamiya, Yumiko, Doyle, Martha, Henretta, John, and Timonen, Virpi
The Gerontologist 54, (5), 2014, p773 - 783
Abstract
Purpose: This article examines the impact of early and later life circumstances on
loneliness among people aged 65+ in Ireland.
Design and Methods: Data are from the first wave of the Irish Longitudinal Study on
Ageing (TILDA), a nationally representative sample of community-dwelling adults
aged 50+. 2,645 participants aged 65+ were included in the analysis. Because of the
large number of never married persons in the older Irish population, we first used a
multinomial logistic model to examine which childhood circumstances are associated
with current marital status. We then estimated multiple regression models for
loneliness, in stages conforming to the life course, to examine the extent to which
early events are mediated by later events.
Results: Poor childhood socioeconomic status (for men and women) and parental
substance abuse (for men) have direct effects on loneliness at older ages.
Implications: The results indicate the significance of the childhood environment for
understanding loneliness in later life. Future research should examine possible
pathways not currently measured that may be responsible for the association of
early environment and later-life loneliness, and explore the links between childhood
and other measures of well-being in old age. The relationship of childhood
1
socioeconomic deprivation and parental substance abuse with adult wellbeing
should be an important consideration in social policy planning.
2
Introduction
Understanding who is affected by loneliness, and why, is important because
loneliness is linked to adverse outcomes that may be preventable or modifiable if
the causes of loneliness are understood (Berkman & Syme, 1979; Cornwell & Waite,
2009). Loneliness is associated with a decrease in physical health (House, 2001;
Perissinotto, Stijacic Cenzer, & Covinsky, 2012; Steptoe, Owen, Kunz-Ebrecht, &
Brydon, 2003; Tomaka, Thompson, & Palacios, 2006; Wilson et al., 2007) and quality
of life (Ekwall, Sivberg, & Hallberg, 2005). Loneliness is also an important predictor of
depression (Barg et al., 2006; Cacioppo, Hughes, Waite, Hawkley, & Thisted, 2006;
ÓLuanaigh and Lawlor, 2008).
An extensive literature documents relationships between late life characteristics and
loneliness, focusing to a substantial extent on later-life social relationships, social
networks, and social support (e.g., Fees, Martin, & Poon, 1999; Park, Jang, Lee,
Haley, & Chiriboga, 2013; Rote, Hill, & Ellison, 2012; van Baarsen, 2002). Examining
the topic more broadly, Victor, Scambler, Bowling & Bond (2005) identified six
independent vulnerability factors for loneliness among older adults: marital status,
increases in loneliness over the previous decade, increases in time alone over the
previous decade; elevated mental morbidity; poor current health; and poorer health
in old age than expected. Advanced age and possession of post-basic education
independently protected against loneliness. From this evidence, the authors propose
that there are three loneliness pathways in later life: continuation of a longestablished attribute, late-onset loneliness, and decreasing loneliness. These findings
3
of Victor et al. (2005) suggest the potential importance of a life course approach that
examines the role of early, mid-life and later events in understanding patterns of
later-life loneliness. The role of early childhood events in later-life loneliness,
however, has remained largely unexplored in existing research.
In this article, we develop a conceptual justification for why and how early life
experiences, marital status, education, and gender affect the level of loneliness in
later life. We utilize a life course approach to conceptualize how early life
circumstances (family status and financial well-being, urban/rural residence,
childhood health, and parental substance abuse) might have a direct effect on laterlife loneliness, or predispose individuals to later-life financial, health, and social
circumstances that in turn contribute to loneliness. We also examine the role of
marital status, which is highly variable for Irish birth cohorts born before the Second
World War, and gender differences.
A large body of literature takes a life course approach to understanding health
outcomes, focusing on the associations between childhood health / socioeconomic
circumstances and later-life health. The previously-discussed association of
loneliness with later-life physical health suggests that it, too, may be either a direct
or indirect result of childhood circumstances. Examination of this relationship offers
the potential of a richer understanding of the role of early events in later-life
physical and mental health.
4
Association of early and mid-life socioeconomic status with health and mortality in
later life has been explained using the accumulation or pathways model, a critical
event model (key events have latent effects) (Ben-Schlomo & Kuh, 2002) or a
cumulative stress model (Pollitt, Rose, & Kaufman, 2005). These three processes,
though not mutually exclusive, outline the simplest ways that early circumstances
might affect later outcomes, including loneliness. While these conceptual models are
not isomorphic with results from data analysis, indirect effects – life-long pathways
through education and marriage – would be consistent with the accumulation
model) while direct effects controlling for mid- and later life circumstances would be
consistent with a critical event model or a cumulative stress model. Much
sociological interest has been on the pathway model (O’Rand, 1996; Hatch, 2005).
The critical events model may result from some physiological change produced by
early stress (Galobardes, Lynch, & Davey Smith, 2004). In a statistical sense,
however, a direct effect cannot be equated with a conceptually-defined critical
event because it is possible that the appropriate pathway has not been identified or
measured and an additional alternative is the additive cumulative stress model.
Hence we prefer to focus our conceptualization on indirect pathways and direct
effects rather than equating them in any definitive way with a particular conceptual
model.
There is substantial evidence for direct effects of childhood health and
socioeconomic status on later-life health, including physical health (Case, Fertig, &
Paxson, 2004; Drakopoulos, Lakioti, & Theodossiou, 2011; Palloni, Milesi,White, &
Turner, 2009), functional health (Haas, 2008), and mental health (Kamiya, Doyle,
5
Henretta & Timonen, 2013). Childhood health also affects adult economic status
(Case et al., 2004; Drakopoulous et al., 2011; Palloni et al., 2009; Smith, 2009;).
A direct link between childhood characteristics and later-life loneliness might
operate through self-definition and role conception that are linked to loneliness in
both old age, and childhood (de Jong Gierveld, 1998). Childhood socioeconomic
adversity, poor health and parental substance abuse are detrimental to self-esteem
and self-efficacy, give rise to feelings of powerlessness, rejection, and self-perceived
lack of disclosure to others (de Jong-Gierveld, 1998), which in turn can have longlasting effects on loneliness, either directly or indirectly. Fry & Debats (2002) find
evidence for the importance of adult self-efficacy variables as predictors of
loneliness. Gender-specific variations revealed that women's stronger self-efficacy
domains in the interpersonal, social, and emotional realms, and men's stronger selfefficacy beliefs in the instrumental, financial, and physical realms predicted less
loneliness and psychological distress.
In addition to possible direct effects, low socioeconomic status, poor health and
parental substance abuse may inhibit the development of capacities, skills and
connections that are conducive to social engagement and achievement of
milestones and statuses such as forming relationships, marrying, gaining access to
higher education and obtaining high-status employment. That is, health and family
economic status in childhood may be consequential for later life loneliness through
the impact of early life events on later life social mobility and social capital (Deary et
6
al., 2005). This latter process might produce an indirect pathway to loneliness
through mid-life and later-life statuses.
The analysis presented here draws on data for persons aged 65 and older in Ireland
as of 2010. The youngest members of the analysis sample were born in 1945. This
and earlier Irish cohorts share two important contextual factors with implications for
the analysis. When these cohorts passed through childhood to early adulthood,
Ireland did not have free secondary schooling (introduced in 1967); free tertiary
(hospital) health care (introduced for lower-income groups in 1952, and expanded to
cover entire population in 1991); or free primary health care for low-income groups
(introduced in 1970). The cost of secondary schooling was prohibitive for children of
most farmers and working-class parents, and only a very small proportion of them
were able to access scholarships for secondary schools, or educational opportunities
in institutions that focused on training for the religious vocations (Keenan, 2006).
Thus effects of early life socioeconomic status and health may be particularly strong
for these cohorts because key welfare state measures were not available as
potential mechanisms for evening out life chances in the early stages of their life
course. In addition, marital status may be a particularly important link between early
and later-life experiences in these cohorts because a substantial portion of them
followed the traditional Irish marriage pattern (Kennedy, 1973) and never married.
Factors producing high rates of non-marriage in Ireland differed for men and
women. Given limited employment opportunities before 1950, marriage was
typically not possible for rural men who were not in line to inherit the family farm.
Women, in contrast to many other European countries, were more likely to
7
complete their schooling and to emigrate than men (MacCurtain, & Ó Corráin, 1979).
Married women were excluded from public employment between 1932 and 1973,
providing some incentive for more educated women to remain single (Kennedy,
1973; MacCurtain, & Ó Corráin, 1979). Different marital patterns for men and
women, together with large differences in the rate of later-life widowhood suggest
the possibility of different life course patterning of loneliness by gender.
Design and Methods
Sample
Data come from the first wave of the Irish Longitudinal Study on Ageing (TILDA).
TILDA is a nationally representative sample of 8,504 persons aged 50 and over living
in Ireland. TILDA was designed to mirror other longitudinal studies on aging such as
the Health and Retirement Study (HRS) in the United States, the English Longitudinal
Study on Ageing (ELSA), and the Survey of Health, Ageing and Retirement in Europe
(SHARE) (Whelan & Savva, 2013). Fieldwork for Wave 1 was completed between late
2009 and mid-2011. Data were collected using three modes: a face-to-face personal
interview, a leave-behind self-completion questionnaire, and a later clinic-based
health assessment. Our analysis uses data from the first two modes. The response
rate for the face-to-face interview was 62%. Of those, 83% returned the selfcompletion questionnaire which included the loneliness scale analyzed here.
Analysis is limited to respondents aged 65 and older. There are 3,517 respondents
over age 65 in TILDA, of whom 2,810 returned the self-completion questionnaire
with the loneliness scale completed. Of the 2,810 who provided the loneliness scale,
8
six percent are lost to missing data, leaving a final sample size of 2,645, consisting of
1,299 men and 1,346 women.
Measures
Outcome variable – Loneliness. Loneliness was assessed using a modified version of
the UCLA Loneliness Scale (Russell et al., 1980), including four negatively-worded
questions (e.g., “How often do you feel left out?”) and one positively-worded
question (“How often do you feel in tune with the people around you?”), each with a
three-point response scale of “hardly ever or never”; “some of the time”; or “often”
coded 0, 1, 2. The responses to the five items are summed, with higher scores
signifying greater loneliness. The average score for older men in Ireland is 1.9
(SD=2.1) and the average for older women is 2.2 (SD=2.1) on a scale from 0 (not
lonely) to 10 (extremely lonely). For the present study, the alpha reliability
coefficient is .78.
We compared loneliness means from TILDA with those from the 2004-2005 ELSA and
the 2006 HRS described above. Only three variables from the UCLA scale were
common to all three studies. For this subset of the questions, the means for TILDA,
ELSA, and HRS, respectively, were 1.04, 1.00, and 1.24 for men and 1.21, 1.34, and
1.49 for women. TILDA results are quite similar to those of ELSA, and the HRS scores
are higher. These differences may indicate true differences among the samples or
may reflect different contexts in which the questions were asked.
Independent variables - Demographic variables. Current marital status is
represented by four dichotomous variables (never married, currently married,
9
divorced/separated and widowed). Currently married is the reference category. Age
is a continuous variable. Current residential location is coded by three dichotomous
variables (Dublin, city or town other than Dublin, and rural Ireland) with Dublin as
the reference category.
Independent variables - Early life circumstances. Early life circumstances assessed
childhood adversity using three measures of family socioeconomic conditions, a
measure of childhood health, and a measure of parental substance abuse. The
childhood socioeconomic measures are father’s social class, family’s relative income
level, and residential location. Father’s social class is measured by the respondent’s
report of father’s main occupation during childhood, coded using the Irish Census
Social Class Scale, as managerial or professional (the reference category), nonmanual, manual, and farmer (Appendix 2 of Census 2006). The family’s relative
income level was assessed by asking whether the respondent’s childhood family was
financially “pretty well off” (the reference category), “about average” or “poor.”
Childhood residential location is included among socio-economic conditions because
of the great disadvantages in rural areas in the early 20th century (although some
inner city areas of Dublin were also characterised by extreme poverty). It is coded 1
for rural childhoods and 0 for those who grew up in a city or town.
To measure poor childhood self-reported health, respondents were asked to rate
their health before age 14 as “poor,” “fair,” “good,” or “excellent.” In this analysis,
“poor,” coded 1, is contrasted with the combined other responses, coded 0. A
parental substance problem was measured by the following question: “Before you
10
were 18 years old, did either of your parents drink or use drugs so often that it
caused problems in the family?” It is coded into three categories: yes, no, and
question not answered. “No” is the reference category. We retained a missing
category in the analysis both to retain respondents we would otherwise lose and
because not answering this question may contain information about childhood.
Respondents may skip a question for many reasons, but when the topic is so
sensitive, it is possible that being missing indicates the presence of substance abuse.
If this is the case, we would expect missing and yes responses to have similar
associations with loneliness.
Independent variables. - Early adult circumstances. We consider educational
attainment to be a measure of achievement in late childhood and early adulthood.
Early school leaving was common in Ireland before the 1967 introduction of free
secondary schooling. The youngest respondents analyzed here were in their early
twenties in 1967 and hence were not affected by this change. Educational
attainment is a categorical variable measuring the respondent’s highest degree of
education obtained: primary, secondary, or tertiary, with primary as the reference
category.
Independent variables. - Later life circumstances. Later life socioeconomic status was
measured by current income and home ownership as a proxy for wealth. Income is
total household income in euro. Those who did not answer the income question
were asked a series of questions to identify their income bracket. We imputed
missing data on this variable using the mean of valid responses that fell in the same
11
bracket as the missing respondent plus a random error with a variance equal to that
of the valid responses. Home ownership is a dichotomous variable that seeks to
capture lifetime wealth. Current health is measured with two variables. Selfreported health equals 1 if the respondent reported “very poor” or “bad” health (vs.
“good” or “excellent” health, the reference category). IADL is measured as the
number of Instrumental Activities of Daily Living with which the respondent reported
having difficulty. The activities included in TILDA are preparing a hot meal, doing
household chores, shopping, making phone calls, taking medications, and managing
money. We also include a measure of the respondent’s current ability to drive,
coded as never drove (reference category), past driver, and current driver.
Model
We base our model for loneliness on typical time ordering of events. The
presentation of the regression results in Tables 4 and 5 with their four equations
reflect this conceptual model.
Early childhood events precede completion of schooling which typically precedes
marriage. In turn, these three sets of events typically precede the late-life
characteristics measured at the interview. Multiple variables are added in the first
and fourth equations, and we do not make any assumptions about time ordering or
causation among the variables added in these equations.
Results
As noted earlier, Irish marriage patterns in the cohorts analyzed are quite distinct.
Table 1 indicates the high never-married rates. Among men, 14.7% of those aged 6512
74 years never married, compared with 17.9% of those aged 75 years and older.
Among women, 11.5% of those aged 65-74 years never married compared with
17.7% of those aged 75 and over. These patterns reflect the gradual increase in
marriage rates in Ireland from the 1960s onwards. Still, most men over age 65 are
currently married (66.7%), while the most prevalent status among women is
widowed (42.6%), reflecting the average age difference between spouses and
women’s higher life expectancy. Only four percent of the population aged 65 and
over is divorced or separated, reflecting the relatively late introduction of divorce in
Ireland (1997), a development that is too recent to have had major impact among
the current cohorts of older people (Fahey, 2012). Divorce is considered less
acceptable by older than younger cohorts in Ireland, suggesting that divorce might
carry stigma among some groups of older adults (Garry, Hardiman, & Payne, 2006).
---Table 1 around here
In Table 2 we present the descriptive statistics for variables used in the analysis, both
dependent and covariates, by sex. Focusing on the dependent variable, there is a
difference in loneliness by gender and marital status. Overall mean loneliness scores
for men and women are similar. However, while currently married men report less
loneliness than currently married women, men’s loneliness scores are higher than
women’s in each of the other marital groups.
----Table 2 around here
13
We first used a multinomial logistic model to examine whether childhood
circumstances are associated with current marital status. Given the different life
experiences of men and women, we examined whether the models were
significantly different by gender and found that they are (Χ2 = 86.4 / 11 df. p< .001).
Table 3 presents the results of the marital status analysis separately for men and
women. Never married men, contrasted to those currently married, show a
distinctive pattern. Consistent with the earlier discussion of Irish marital patterns,
growing up in a rural area, having a father who was a farmer, and having only a
primary education are strongly associated with men’s being never married. Poor
childhood health and not responding to the substance abuse question are also
significantly associated with being never married. Among women, having a tertiary
education and being older, a measure of birth cohort, are associated with remaining
single. The point estimate for childhood health on the married-never married
contrast is virtually the same as for men but it is not significant. Returning to the
bivariate frequencies in Table 2, having poor childhood health is much more
common among the never-married than among the currently married.
--- Table 3 around here
We next estimated multiple regression models for loneliness, estimating the model
in several stages conforming to typical time ordering in order to examine the extent
to which early events are mediated by later events. These results are presented in
14
Tables 4 (for men) and 5 (for women), We discuss these two tables together and
discuss each variable across equations..
Model 1 includes only early life characteristics; model 2 adds level of education; and
model 3 adds marital status. The full model, model 4, adds later-life characteristics.
We estimated the full interaction model of gender with all the covariates included in
model 4 and found significant interactions with gender (F = 1.82 with 23/2597 df.
p=.01). We therefore present models separately for men and women. Among
individual variables, marital status (p=.02) and current place of residence (p=.03)
have significantly different effects for men and women.
---Tables 4 and 5 around here
In model 1, two early life characteristics, family economic status and parental
substance abuse, are associated with loneliness. Men and women who report poor
financial status as a child (contrasted to being well-off) are more lonely in later life.
The association is also significant for women who report average financial status.
These coefficients retain significance across all equations; the coefficients in model 4
retain more than 85% of their significant model 1 associations for women and nearly
70 percent for men, indicating that relatively little of the association is mediated by
later events.
In model 1, parental substance abuse is significant for men, and 80 percent of its
association is direct, even in model 4. For both men and women, missing response
15
to the parental substance abuse question has a strong association with loneliness
across all equations with some decline in the final equation.
Model 2 incorporates the effects of education. Having more education, secondary
education for women and tertiary education for men, is associated with lower levels
of loneliness. These associations are largely accounted for by the addition of marital
status in model 3 and later-life events in model 4. They do not differ significantly
between men and women in model 4.
In model 3, being never married or widowed (contrasted with currently married) are
associated with higher loneliness scores for men and women; and for men, being
separated or divorced is also associated with loneliness. Very little of the relationship
of marital status and loneliness is accounted for by other late life statuses; 80
percent or more of the association is direct in model 4. Both men and women who
are never married or widowed, as well as men who are divorced or separated, report
more loneliness than currently married persons of the same sex, but the association
of marital status with loneliness is significantly greater for men. As noted earlier, the
overall test of a gender difference in the effect of marital status is significant; among
individual coefficients, the contrast between married and widowed is significantly
different by gender (p=.01) and the married-never married is borderline significant
(p=.06). Among later life characteristics, both current poor self-rated health and
number of IADL limitations are associated with an increase in loneliness for both
men and women. Being a current or past driver, compared with never having driven,
is not associated with loneliness for either men or women. As noted earlier, the
16
overall test of gender differences in place of residence is significant as is the
individual contrast of rural versus Dublin residence (p<.01). Rural residence is
associated with significantly greater loneliness for women than for men.
As noted earlier, most of the effects of the early childhood measures of poverty and
parental substance abuse in model 1 remain direct in model 4, indicating relatively
little mediation. Standard mediation tests are not very useful when applied to our
model because of the number of multinomial variables that are mediators. Instead,
using a Wald test of differences across equations, we examined the related issue of
whether addition of education, marital status, and health each produced a
significant decline in effects of childhood poverty and parental substance abuse.
Among men, addition of marital status produced a significant decline (p=.05) in the
effect of substance abuse on loneliness. For marital status and education, these
tests can be directly related to the models presented. To test health, we used a
simplified model that added only the two health measures in model 4. Overall,
these tests of change in the effect of childhood variables when possible mediators
are added to the model strengthen the conclusion that most of the effect of
childhood variables on loneliness in our models is direct.
Discussion; Limitations and Future Directions
There are three central findings. First, childhood characteristics have direct effects
on later-life loneliness, and, with a few exceptions, only a small portion of their initial
17
effect is accounted for by intervening statuses. Second, compared to married
persons, the never married and widows are substantially lonelier, and these
relationships are stronger for men than for women. Third, the later-life
characteristic of poor health is associated with greater later-life loneliness.
Growing up in relative poverty and, for men, having parents with substance abuse
are both associated with greater later-life loneliness. Most of the initial association
remains direct after controlling for adult and late-life characteristics. While it is not
possible to unequivocally attribute this finding to one of the conceptual explanations
presented earlier, the finding does provide substantial evidence of the continuing
impact of childhood on later-life loneliness. The finding suggests that the research
literature on the long-term implications of childhood characteristics for health can
be expanded to related areas of morale.
Interestingly, men and women who did not answer the substance abuse question
also showed high levels of loneliness. The interpretation of this result is unclear. It
may be that current loneliness leads respondents to be less willing to answer
sensitive questions or it could reflect actual parental substance abuse about which
the respondent is too conflicted to provide a report. The relationship of childhood
socioeconomic deprivation and parental substance abuse with adult wellbeing
should be an important consideration in social policy planning.
In contrast to the currently married, all non-married statuses report greater
loneliness. These effects are greater for men than women, perhaps because men’s
18
earlier mortality means that older unmarried women have more potential friends
among other widowed women (Blau, 1961). Ireland is unique in the developed
country context because a high proportion of the birth cohorts examined here are
never married and relatively few are separated or divorced. There are strong
gender-specific patterns to the never married. Among men, parental substance
abuse, primary education, being the son of a farmer, living in a rural area, and poor
childhood health are associated with being never married. Among women, it is the
highly educated who are unmarried. In addition, the strong effect of age indicates
the rapid decline in the never married status in the younger members of the cohorts
studied. Yet, while marital status, including being never married, has a substantial
association with loneliness, there’s relatively little evidence that marital status is the
pathway between early events and later-life loneliness. Of the factors affecting
marriage, only parental substance abuse (for men) and education are associated
with loneliness. Most of the relationship between parental substance abuse and
loneliness is direct. However, education effects in model 2 are accounted for by
marital status as well as late life status.
Marital status effects retain almost all their effects in model 4, and this continued
association is different from previous findings relating marital status to health
outcomes. Research has indicated that the effects of marital status on health (as
measured by mortality) are, to a substantial extent, accounted for by economic
status (Lillard & Waite, 1995; Rogers, 1995). Our findings indicate that this is not the
case for loneliness in Ireland.
19
Finally, late life characteristics of poor health and low social support are associated
with loneliness. Poor health is likely to increase social isolation by reducing the
ability of the individual to participate in social relationships, and limited social
support is also likely to produce social isolation. Hence these findings are not
surprising and provide some evidence of the validity of the measure of loneliness.
The results are subject to several limitations. Information on childhood
circumstances was gathered retrospectively when the respondents were aged 65
years and over, and such retrospective accounts may be compromised by poor recall
(Looker, 1989). A further limitation is that the study is a cross-sectional analysis
which fails to capture changes over time, and is not fully able to capture the
direction of causality. We cannot rule out reverse causation between loneliness and
some of the late life variables such as current health and social support. Moreover,
current loneliness may affect recall of childhood events, leading respondents to
emphasize their negative experiences. Finally, the response rate to the TILDA study
was 62%. This response rate is quite good for a European study though lower than
similarly well-designed samples in the U.S. 83% of those interviewed returned the
self-completion questionnaire. This is also a solid response rate. However, taken
together, 51.5% of the original sample are represented in this analysis. The results
can be generalized to the population sampled only on the assumption that nonresponse was random. The issue and the assumption are not unique to this study but
are found in all survey research.
The findings reported here add to what is already known about the relation of
childhood and well-being in later life. Future research should be directed to
20
exploring the links between childhood and other measures of well-being as well as
continuing to examine possible pathways not currently measured that may be
responsible for the association of early environment and later-life loneliness.
Acknowledgements
The authors would like to acknowledge the contribution of the participants in the
study, members of the TILDA research team, and administrators. Funding was
gratefully received from the Atlantic Philanthropies, the Irish Government, and Irish
Life plc.
21
References
Barg, F.K., Huss-Ashmore, R., Wittink, M.N., Murray, G.F., Bogner, H.R. & Gallo, J.J.
(2006). A mixed-methods approach to understanding loneliness and depression in
older adults. Journal of Gerontology: Social Sciences, 61B, S329-S339.
Ben-Shlomo, Y. & Kuh, D. (2002). A life-course approach to chronic disease
epidemiology: Conceptual models, empirical challenges, and interdisciplinary
perspectives. International Journal of Epidemiology, 31, 285-293.
Berkman, L. & Syme, S. (1979). Social networks, host resistance, and mortality: A
nine-year follow-up study of Alameda County residents. American Journal of
Epidemiology, 109, 186-204.
Blau, Z.S. (1961). Structural constraints on friendships in old age. American
Sociological Review, 26, 429-439.
Cacioppo, J. T., Hughes, M.E., Waite, L. J., Hawkley, L. C., & Thisted, R.A. (2006).
Loneliness as a specific risk factor for depressive symptoms: Cross-sectional and
longitudinal analyses. Psychology and Aging, 21, 140–151.
Case, A., Fertig, A. & Paxson,C. (2004). The lasting impact of childhood health and
circumstance. Journal of Health Economics, 24, 365-389
Cornwell, E. & Waite, L. (2009). Social disconnectedness, perceived isolation, and
health among older adults. Journal of Health and Social Behavior, 50, 31-48.
Deary, I.J., Taylor, M.D., Hart, M.D., Wilson, V., Davey Smith, G., Blane, D. & Starr, J.
M (2005). Intergenerational social mobility and mid-life status attainment: Influences
22
of childhood intelligence, childhood social factors, and education. Intelligence, 33,
454-472.
De Jong Gierveld, J. (1998). A review of loneliness: Concepts and definitions, causes
and consequences. Reviews in Clinical Gerontology, 8, 73-80.
Drakopoulos, S.A., Lakioti, E. & Theodossiou, I. (2011). Childhood socioeconomic
deprivation and later adulthood health. International Journal of Social Economics,
38, 23 – 38.
Ekwall, A.K., Sivberg, B. & Hallberg, I.R. (2005). Loneliness as a predictor of quality of
life among older caregivers. Journal of Advanced Nursing, 49, 23-32.
Fahey, T. (2012). Small bang? The effect of divorce legislation on marital breakdown
in Ireland. International Journal of Law, Policy and the Family, 26, 242-258.
Fees, B., Martin, P. & Poon, L (1999). A model of loneliness in older adults. Journal of
Gerontology, Series B: Psychological Sciences and Social Sciences, 54, 231-239.
Fry, P.S. & Debats, D.L. (2002). Self-efficacy beliefs as predictors of loneliness and
psychological distress in older adults. International Journal of Aging and Human
Development, 55, 233-269.
Galobardes, B. Lynch, J.W & Davey Smith, G. (2004). Childhood socioeconomic
circumstances and cause-specific mortality in adulthood: systematic review and
interpretation. Epidemiologic Review, 26, 7-21.
Garry, J., Hardiman, N. & Payne, D. (2006). Irish social and political attitudes.
Liverpool, Liverpool University Press.
23
Haas, S. (2008). Trajectories of functional health: The ‘long arm’ of childhood health
and socioeconomic factors. Social Science and Medicine, 66, 849-861.
Hatch, S.L. (2005). Conceptualizing and Identifying Cumulative Adversity and
Protective Resources: Implications for Understanding Health Inequalities. Journal of
Gerontology, Series B: Psychological Sciences and Social Sciences, 60, S130-S134
House, J.S. (2001). Social isolation kills, but how and why? Psychosomatic Medicine,
63, 273-274.
Kamiya, Y., Doyle, M., Henretta, J. & Timonen, V. (2013). Depressive symptoms
among older adults: The impact of early and later life circumstances and marital
status. Aging & Mental Health, 17, 349-357.
Keenan, D. (2006). Post-Famine Ireland: Social Structure Ireland as it really was.
Xlibris Corporation.
Kennedy, R.E. (1973). The Irish: Emigration, Marriage and Fertility. London,
University of California Press.
Lillard, L.A. & Waite, L.J. (1995). ‘Til death do us part: Marital disruption and
mortality. American Journal of Sociology, 100, 1131-1156.
Looker, E. D. (1989). Accuracy of proxy reports of parental status characteristics.
Social Education, 62, 257-276.
MacCurtain, M. and ÓCorráin, D (1979). Women in Irish society - the historical
dimension. Naas, Arlen House.
ÓLuanaigh, C. & Lawlor, B.A. (2008). Loneliness and the health of older people.
International Journal of Geriatric Psychiatry, 23, 1213-1221.
24
O’Rand, A.M. (1996). The precious and the precocious: Understanding cumulative
disadvantage and cumulative advantage over the life course. The Gerontologist, 36,
230–238.
Palloni, A., Milesi, C.,White, R.G. and Turner, A. (2009).Early childhood health,
reproduction of economic inequalities and the persistence of health and mortality
differentials. Social Science and Medicine, 68, 1575-1582.
Park, N.S., Jang, Y, Lee, B.S., Haley, W.E. & Chiriboga, D.A. (2013). The mediating role
of loneliness in the relation between social engagement and depressive symptoms
among older Korean Americans: Do men and women differ? The Journals of
Gerontology, Series B: Psychological Sciences and Social Sciences, 68, 193-201.
Perissinotto, C.M., Stijacic Cenzer, I. & Covinsky, K.E. (2012). Loneliness in older
persons - A predictor of functional decline and death. Archives of International
Medicine, 172, 1078-1083.
Pollitt, R.A., Rose, K.M., & Kaufman, J.S. (2005). Evaluating the evidence for models
of life course socioeconomic factors and cardiovascular outcomes: A systematic
review. BMC Public Health, 5, 7.
Rogers, R.G. (1995). Marriage, sex, and mortality. Journal of Marriage and the
Family, 57, 515-526.
Rote, S., Hill, T.D. & Elllison, C.G. (2012). Religious attendance and loneliness in later
life. The Gerontologist, 53, 39-50.
Russell, D., Peplau, L.A., & Cutrona, C.E. (1980). The revised UCLA Loneliness Scale:
Concurrent and discriminant validity evidence. Journal of Personality and Social
Psychology, 39, 472-480.
25
Smith, J.P. (2009). The impact of childhood health on adult labor market outcomes.
Review of Economics and Statistics, 91, 478-489.
Steptoe, A., Owen, N., Kunz-Ebrecht, S.R. & Brydon, L. (2004). Loneliness and
neuroendocrine, cardiovascular, and inflammatory stress responses in middle-aged
men and women. Psychoneuroendrocrinology, 29, 593-611.
Tomaka, J., Thompson, S. & Palacios, R. (2006). The relation of social isolation,
loneliness and social support to disease outcomes among the elderly. Journal of
Aging and Health, 18, 359 -384.
van Baarsen, B. (2002). Theories on coping with loss: The impact of social support
and self-esteem on adjustment to emotional and social loneliness following a
partner’s death in later life. Journals of Gerontology: Social Sciences, 57, S33-S42.
Victor, C.R., Scambler, S.J., Bowling, A. & Bond, J. (2005). The prevalence of, and risk
factors for, loneliness in later life: A survey of older people in Great Britain. Ageing &
Society, 25, 357–375.
Wilson, R.S., Krueger, K.R., Arnold, S.E., Schneider, J.A., Kelly, J.F., Barnes, L.L., &
Bennett, D.A. (2007). Loneliness and risk of Alzheimer Disease. Archives of General
Psychiatry, 64, 234-240.
Whelan, B.J. and Savva, G. M. (2013). Design and Methodology of the TILDA Study.
Journal of the American Geriatrics Society 61, S265-68.
26
Table 1: Distribution of marital status among people aged 65 and over in Ireland
%
Never
married
%
Sep/divorce
d
%
72.3
57.7
66.7
14.7
17.9
15.9
4.1
0.8
2.9
55.7
22.6
40.6
11.5
17.7
14.4
63.9
37.4
52.7
13.1
17.8
15.1
Married
Male
65-74
>=75
Total
Female
65-74
>=75
Total
Total
65-74
>=75
Total
Widowed
Total
Number in
sample
8.9
23.5
14.5
100
100
100
852
447
1299
4.4
0.0
2.4
28.4
59.8
42.6
100
100
100
841
505
1346
4.3
0
2.6
18.7
44.4
29.6
100
100
100
1693
952
2645
%
Estimates are weighted
27
Table 2: Descriptive statistics
married
Demographic characteristics
Age (mean)
Current residence
Dublin
Other city or town
Rural area
Early life circumstances
Father's occupation
Professional
Manual
Non-manual
Farmers
Family financial well-being
well-off
average
poor
Residential location (=rural)
Childhood health (=poor)
Parental substance problem (=yes)
Missing
men
separated/
never
married
divorced
widowed
married
women
separated/
never
married
divorced
N
widowed
72.4
73.7
70.1
77.2
71.2
75.6
68.5
77.3
2645
25.6
29.6
44.8
15.9
25.2
58.8
29.7
32.6
37.6
26.6
32.1
41.3
28.1
29.2
42.7
33.7
37.1
29.1
30.9
41.1
27.9
25.2
28.1
46.7
670
738
1237
10.2
56.9
8.9
23.6
2.7
48.8
5.6
42.8
21.8
48.4
11.3
18.3
5.5
59.8
7.4
27.1
10.3
56.4
9.2
24.1
18.6
40.1
11.1
30.2
24.4
59.2
6.1
10.3
6.3
53.8
6.1
33.7
327
1297
248
703
9.6
63
27.4
58.1
5.9
4.5
1.8
7.3
65.9
26.8
76.5
10.7
6.6
8.2
9.7
69.1
21.8
63.8
8.6
9.6
0
7.7
59
33.3
63.2
5.7
5
2.5
7.8
70.6
21.6
59.1
6.2
4.5
1.8
12
69.3
18.7
61.5
9.7
6.6
8.2
21.8
55.9
22.3
51.8
9.8
9.6
0
9
71.6
19.4
70.7
9.1
5
2.5
306
1774
565
1651
178
143
70
28
Early adult life circumstances
Education
primary
secondary
tertiary
Late adult life circumstances
Household Income (mean)
Own a house
Self-reported health (=poor)
IADL (mean)
Driving
never drove
used to drive
drive
Mental Health
Loneliness (mean)
Sample size (N)
Estimated are weighted
Sample size is unweighted
52.2
32.3
15.4
72.8
18.6
9
49.5
32.1
18.5
66.6
24.9
8.5
47.9
39.4
12.8
31.4
44.7
23.9
40.8
38.7
20.6
68.7
23.8
7.5
1083
833
663
32920
97.8
16.7
0.14
16602
78.8
17
0.13
17333
50.7
14.9
0.12
27607
87.7
15.2
0.24
29514
96.3
14.9
0.13
21560
78.9
16.2
0.32
17000
74.2
25.3
0.41
17187
92
18.4
0.49
2645
2645
396
2645
5.1
87.6
7.3
16.8
61.3
21.8
14.8
71.7
13.4
6.7
77.4
15.9
34.1
56.6
9.2
33.4
55.2
11.4
31.6
58.8
9.6
49.8
41.3
8.9
524
1882
239
1.4
931
3.1
144
2.7
42
2.9
182
1.7
686
2.2
123
2.2
43
2.7
491
2645
2645
29
Table 3: Multinomial logistic regression of early life circumstance and marital status (results in
Odds Ratio)
men
women
never
never
married
divorced
widowed
married
divorced
widowed
1.014
0.937*
1.111***
1.094***
0.874**
1.145***
(0.0159)
(0.0309)
(0.0146)
(0.0186)
(0.0371)
(0.0129)
childhood health
1.947*
1.179
0.982
1.941
1.811
1.463
(=poor)
(0.639)
(0.741)
(0.360)
(0.694)
(0.937)
(0.370)
childhood residence
1.889**
1.729
0.945
1.057
0.947
1.127
(=rural)
(0.461)
(0.618)
(0.182)
(0.252)
(0.321)
(0.172)
2.085
0.546
1.321
0.583
0.502
1.280
(0.980)
(0.262)
(0.424)
(0.177)
(0.211)
(0.298)
1.896
0.938
1.213
0.970
0.515
1.330
(=non-manual)
(1.055)
(0.530)
(0.470)
(0.372)
(0.315)
(0.402)
father occupation
3.257*
0.377
1.028
0.897
0.217*
1.438
(=farmer)
(1.574)
(0.220)
(0.378)
(0.296)
(0.129)
(0.366)
1.145
0.961
1.237
0.894
0.435*
0.781
(0.395)
(0.477)
(0.383)
(0.262)
(0.184)
(0.166)
0.859
0.818
1.611
0.948
0.456
0.722
(0.335)
(0.500)
(0.553)
(0.376)
(0.265)
(0.189)
age
father occupation
(=manual)
father occupation
childhood family
(=average)
childhood family
(=poor financially)
30
parental substance abuse
1.824
1.636
1.384
0.749
1.695
0.800
(0.685)
(0.915)
(0.515)
(0.372)
(0.835)
(0.246)
5.402***
0.00000836
1.360
2.564
1.035
1.099
(=missing)
(2.270)
(0.00404)
(0.743)
(1.245)
(1.098)
(0.446)
secondary education
0.536**
0.797
0.886
1.716
0.739
0.641**
(0.126)
(0.330)
(0.182)
(0.477)
(0.301)
(0.0978)
0.632
0.922
0.681
3.076***
0.910
0.749
(0.169)
(0.405)
(0.170)
(0.915)
(0.397)
(0.139)
(=yes)
parental substance abuse
tertiary education
Exponentiated coefficients; Standard errors in parentheses
*p<0.05, **p<0.01, ***p<0.001
31
Table 4: Effects (OLS regression) of early and late life circumstances on loneliness, men
men
age
childhood health
(=poor)
childhood residence
(=rural)
father occupation
(=manual)
father occupation
(=non-manual)
father occupation
(=farmer)
childhood family
(=average)
childhood family
(=poor financially)
parental substance abuse
parental substance abuse
(=missing)
secondary education
tertiary education
Model 1
0.00897
(0.00940)
0.413
(0.236)
0.147
(0.131)
-0.0188
(0.194)
-0.0143
(0.247)
0.0663
(0.220)
0.257
(0.192)
0.728***
(0.218)
0.718**
(0.252)
1.272***
(0.369)
Model 2
0.00719
(0.00946)
0.421
(0.236)
0.117
(0.132)
-0.161
(0.202)
-0.0512
(0.247)
-0.0852
(0.229)
0.204
(0.193)
0.639**
(0.223)
0.738**
(0.252)
1.273***
(0.369)
-0.0889
(0.141)
-0.382*
(0.161)
never married
sep/divorced
widow
current poor health
IADLs
other city than Dublin
32
Model 3
-0.0118
(0.00927)
0.348
(0.225)
0.0624
(0.126)
-0.211
(0.192)
-0.101
(0.235)
-0.151
(0.218)
0.162
(0.184)
0.586**
(0.212)
0.612*
(0.240)
0.983**
(0.354)
0.00157
(0.134)
-0.269
(0.153)
1.328***
(0.180)
1.229***
(0.308)
1.608***
(0.163)
Model 4
-0.0191*
(0.00935)
0.291
(0.220)
0.0913
(0.131)
-0.182
(0.188)
-0.0693
(0.230)
-0.0821
(0.216)
0.201
(0.180)
0.528*
(0.208)
0.638**
(0.235)
0.976**
(0.346)
0.0797
(0.133)
-0.0900
(0.155)
1.173***
(0.186)
0.963**
(0.320)
1.570***
(0.161)
0.833***
(0.154)
0.348***
(0.0909)
-0.0521
(0.148)
rural
log(income)
own house
drive
used to drive
_cons
N
0.622
(0.721)
1299
1.067
(0.757)
1299
Standard errors in parentheses
*p<0.05, **p<0.01, ***p<0.001
33
2.122**
(0.735)
1299
-0.0787
(0.152)
-0.0696
(0.0467)
-0.416
(0.232)
-0.205
(0.239)
-0.00687
(0.286)
3.689***
(0.941)
1299
Table 5: Effects (OLS regression) of early and late life conditions on loneliness, women
age
childhood health
(=poor)
childhood residence
(=rural)
father occupation
(=manual)
father occupation
(=non-manual)
father occupation
(=farmer)
childhood family
(=average)
childhood family
(=poor financially)
parental substance abuse
parental substance abuse
(=missing)
secondary education
tertiary education
never married
sep/divorced
widow
Model 1
0.0191*
(0.00928)
0.401
(0.229)
0.0814
(0.139)
-0.00778
(0.194)
-0.435
(0.263)
-0.280
(0.218)
0.470*
(0.188)
0.821***
(0.231)
0.308
(0.266)
1.037**
(0.358)
women
Model 2
Model 3
0.0166
-0.0110
(0.00941)
(0.00994)
0.387
0.313
(0.229)
(0.225)
0.0843
0.0599
(0.139)
(0.136)
-0.0784
-0.108
(0.200)
(0.197)
-0.427
-0.472
(0.264)
(0.259)
-0.334
-0.393
(0.222)
(0.218)
0.418*
0.460*
(0.192)
(0.188)
0.709**
0.773***
(0.239)
(0.234)
0.302
0.345
(0.266)
(0.260)
1.041**
1.010**
(0.357)
(0.351)
-0.314*
-0.234
(0.142)
(0.140)
-0.233
-0.194
(0.168)
(0.166)
0.636**
(0.211)
0.268
(0.325)
1.038***
(0.135)
current poor health
IADLs
other city than Dublin
34
Model 4
-0.0204*
(0.0103)
0.0723
(0.221)
-0.152
(0.141)
-0.173
(0.192)
-0.519*
(0.253)
-0.448*
(0.212)
0.397*
(0.184)
0.686**
(0.229)
0.429
(0.254)
0.828*
(0.343)
-0.0371
(0.141)
0.0683
(0.171)
0.644**
(0.211)
0.177
(0.323)
1.038***
(0.134)
1.219***
(0.167)
0.150
(0.0777)
0.252
rural
log(income)
own house
drive
used to drive
_cons
N
0.236
(0.713)
1346
0.701
(0.761)
1346
Standard errors in parentheses
*p<0.05, **p<0.01, ***p<0.001
35
2.242**
(0.780)
1346
(0.157)
0.509**
(0.155)
-0.0437
(0.0531)
-0.0345
(0.238)
-0.186
(0.144)
-0.241
(0.214)
3.129**
(1.020)
1346