Content uploaded by Dimitri Mortelmans
Author content
All content in this area was uploaded by Dimitri Mortelmans on Jun 09, 2015
Content may be subject to copyright.
ORIGINAL PAPER
Retirement Timing of Dual-Earner Couples in 11 European
Countries? A Comparison of Cox and Shared Frailty Models
Hanne De Preter •Dorien Van Looy •
Dimitri Mortelmans
ÓSpringer Science+Business Media New York 2014
Abstract Retirement is not only an individual decision, it
is a decision taken within a family system. The extent to
which estimates of retirement predictors are biased by the
clustering of older workers (50 years of age or older) into
dual-earner households remains unclear. Using longitudi-
nal data on 11 European countries from the first (2004/
2005) and second (2006/2007) waves of the Survey on
Health, Ageing and Retirement in Europe (SHARE), this
study investigated the retirement timing of older workers
aged 50?living in dual-earner households in 11 European
countries. We questioned how measures that are signifi-
cantly related to retirement timing differ according to
whether one takes into account the household context in the
retirement analysis. The results showed that covariates that
are dependent among partners (e.g., household size, edu-
cational level) changed the least between the Cox model
and the shared frailty model. The effects of controlling for
clustering between partners, although fairly modest,
appeared to be important as they revealed a systematic
pattern in the direction of bias in estimates that assume an
independent sample.
Keywords Dual-earner households Retirement timing
Cox models Shared frailty models SHARE
Introduction
Retirement has typically been viewed as an event that men
experience at the end of their careers. In this sense, a
household has only one retirement: that of the husband and
breadwinner (Ho and Raymo 2009). This pervasive ‘‘male
bias’’ has become a subject of increasing criticism, given
the rise in female employment. The focus has gradually
shifted from male household heads to individual men and
women in the labour market (Blossfeld and Drobnivc
2001). In recent years, a substantial body of research has
investigated the retirement transitions of both men and
women (Dahl et al. 2003; Ginn and Arber 1996; Ruhm
1996), focussing largely on factors affecting the retirement
timing of individuals (Loretto and Vickerstaff 2012). One
major conceptual limitation of this individualistic approach
seems to be its failure to acknowledge the extent to which
lives are linked. In addition to the links between lives,
however, life course theory has emphasised the importance
of interdependence between the spheres of work and family
(Elder 1999; Moen 1996; Raymo and Sweeney 2006).
Another fundamental drawback of the individualistic line
of thought has therefore been that, by ignoring familial and
household roles, it implicitly suggests that there are no
differences between households (Blossfeld and Drobnivc
2001).
It is well understood that couples develop divisions of
labour with respect to family and paid work throughout the
life course (Loretto and Vickerstaff 2012). There are
numerous possible conceptualisations of the different types
of household divisions, ranging from role specialisation
(e.g., male breadwinner) to role combination in dual-earner
households (Hall and MacDermid 2009; Loretto and
Vickerstaff 2012). Economic welfare growth, educational
expansion, female emancipation and the increased
H. De Preter (&)D. Van Looy D. Mortelmans
Research Centre for Longitudinal and Life Course Studies
(CELLO), University of Antwerp, St. Jacobstraat 2,
2000 Antwerp, Belgium
e-mail: hanne.depreter@uantwerpen.be
123
J Fam Econ Iss
DOI 10.1007/s10834-014-9403-6
availability of childcare have all stimulated the labour
market participation of women, resulting in an increase in
dual-earner couples (Van Gils and Kraaykamp 2008).
According to life course theory, contemporary dual-earner
couples are living in a historical period in which they have
become the norm for maintaining an adequate standard of
living (Orrange et al. 2003).
Given the increasing number of dual-earner couples
(Gustman and Steinmeier 2000; Kapur and Rogowski
2007), the most obvious unit of analysis for retirement
decisions has become the couple (Bingley and Lanot 2007).
The retirement behaviour of couples has therefore been
studied with growing interest, due to the increasing prev-
alence of observations on working women of retirement
age in self-weighted samples (Hurd 1990). Although most
contemporary workers are part of a dual-earner couple
(thus navigating two career paths and facing two retire-
ments), most studies modelled the passage into retirement
status as an individual event. Moreover, organisational and
public policies and practices have treated retirement deci-
sion-making as an individual endeavour, and not one
undertaken by a couple (Moen et al. 2006).
From the family perspective, it is known that the char-
acteristics and actions of one family member can be
influenced by those of other family members. The fact that
family members are nested within social systems stems in
part from Bronfenbrenner’s (1977) ecological perspective
that individuals and their environment are in constant
negotiation (Pedersen and Minnotte 2012). From this per-
spective, the family unit is a system, i.e., a complex,
multilateral and integrated structure in which family
members are necessarily interdependent and in which they
exert a continuous and reciprocal influence on one another
(Cox and Paley 1997, p. 246). Since individuals are
embedded within larger family systems, their behaviour
cannot be fully understood independent of the context of
that system (Cox and Paley 1997,2003; Kreppner and
Lerner 1989). Family system theory thus provides a broad
contextual framework for studying the retirement decisions
of dual-earner couples within the household context. The
integration of the life-course approach and family system
theory can promote the understanding of ways in which
retirement decision-making is always a social-relational
process. This unified theoretical framework captures the
dynamic aspects of work and family careers, emphasising
the embeddedness of relationships within a system of
family ties (Orrange et al. 2003).
Previous research has modelled the influence of several
individual and family characteristics on retirement timing
(Gustman and Steinmeier 2000; O’Rand and Farkas 2002;
Pienta and Hayward 2002; Smith and Moen 1998). The
results of these studies have demonstrated that retirement is
not only an individual decision, but one that is taken within
a family system (Loretto and Vickerstaff 2012). The extent
to which the clustering of individuals into households
affects factors influencing individual retirement timing
nevertheless remains unclear. Consequently, our aim is to
understand the retirement timing of older workers aged
50?living in dual-earner households in 11 European
countries. We question how measures that are significantly
related to retirement timing differ according to whether one
takes into account the household context in the retirement
analysis. In order to answer this research question, this
study compared a proportional hazard (Cox) model to a
shared frailty model.
A major assumption of the Cox model is that the
observations are independent. Models that ignore cluster-
ing in the survival experiences of couples could be biased,
as they violate the statistical assumption of independence.
This leads to small standard errors and falsely significant
estimates. Frailty models are important in this regard,
because they allow for correlation in the survival experi-
ences of couples while producing accurate estimates of
retirement determinants (Omariba et al. 2007). Given that
individuals within the same household have a similar
likelihood of retirement (Denney 2010), our model allowed
for shared frailty at the household level. The study has
drawn upon data from the first (2004–2005) and second
(2006–2007) waves of the large cross-national Survey of
Health, Ageing and Retirement in Europe dataset. In
SHARE, the information on respondents is sampled in
households, and individuals are thus not treated as inde-
pendent observations.
Predictors of Retirement Timing
Numerous factors influence the retirement timing of indi-
viduals. The most important predictors of retirement are
the economic status of the family in retirement and health
(Henretta and O’Rand 1983). In general, as household
income and wealth increases, individual preferences tend to
shift toward more leisure time (Clarkberg and Moen 2001;
Hatcher 2002), thus resulting in earlier retirement. Given
that, in dual-earner households, both partners contribute to
the household income, the effect of household income is
likely to be related to the effect of personal income on
retirement timing. The household income of dual-earner
couples thus reflects both the resources that are immedi-
ately available and the opportunity costs associated with
early retirement. It therefore has a dual effect on retirement
timing (Pienta and Hayward 2002). People who are affluent
and who feel financially secure have the opportunity to
choose between continuing to work and having increased
leisure time, which could result in early retirement (Schils
2008). In contrast, people with higher incomes might prefer
J Fam Econ Iss
123
to continue earning, due to high opportunity costs, thus
tending to retire later (Parker and Rougier 2007).
Personal health problems push individuals toward
retirement. People who are in poor health tend to retire
earlier than do those who feel that they are in good health.
Having a partner in poor health has a dual effect on the
likelihood to retire. On the one hand, it increases the need
for care, thus accelerating the timing of retirement. On the
other hand, having a partner in poor health may serve to
postpone retirement, given the level of income needed for
medical treatment and meeting the couple’s financial needs
(Gustman and Steinmeier 2004; Pienta and Hayward 2002;
Raymo and Sweeney 2006; Van Solinge and Henkens
2005).
Other important factors include number of working
hours, care tasks, community participation, age and family
composition (Beehr et al. 2000; Dentinger and Clarkberg
2002; Henretta and O’Rand 1983; Kubicek et al. 2010). For
example, Piekkola (2008) examined part-time work but
found no significant results. This finding was likely due to
the existence of two competing theoretical hypotheses
regarding the relationship between total hours worked and
retirement timing. On the one hand, a higher number of
working hours was likely to indicate a stronger commit-
ment to work, whether due to a preference for work or to a
need for higher income (Schils 2008). Along these lines,
therefore, individuals who work more hours should be
more likely to retire later. On the other hand, a lower
number of working hours (e.g., part-time) might be asso-
ciated with a longer working life, given that it facilitates
the work-life balance (Erickson et al. 2010; Schmid 1998).
Contemporary families often struggle in navigating work
and family life (Minnotte 2012). Negative spillovers of
work stress into the family domain (or vice versa) have
been found to play an important role in enhancing the
likelihood of retirement (Kubicek et al. 2010). Given that
working fewer hours creates opportunities to spend more
time on other activities (e.g., interacting with family
members, performing household tasks, or engaging in lei-
sure activities), it is an important factor in understanding
retirement decisions (Voydanoff 2005). Care responsibili-
ties possibly put serious strain on the balance between
work and family demands (Hill et al. 2014). Caring obli-
gations might even prevent older workers from continuing
to work because increasing involvement in care giving
activities may become incompatible with working outside
the house (Dentinger and Clarkberg 2002; Henkens and
Van Solinge 2002).
Moen et al. (2000) indicated that participants in clubs
and volunteer work were more likely to retire early, as they
experienced less uncertainty about their lives after retire-
ment. This is because their social activities provided rou-
tines, rituals and role identity in the same way that their
jobs did, thus helping them to structure their life patterns
and meanings. In contrast, findings from a study by Lancee
and Radl (2012) indicated that engagement in volunteer
work was associated with later retirement, as it indicated
self-selection into active aging. While people who are more
engaged remain attached to their work until later in life,
those who are less engaged withdraw from work much
sooner. The boundary between work and formal partici-
pation might thus be blurred.
Previous research has indicated that, in addition to the
respondent’s age, the age of the partner can be an important
predictor of personal retirement decisions. Given that age is
highly correlated across partners (Hurd 1990; Pienta 2003),
we do not advance any hypotheses with regard to the actual
age of the partner. Instead, we consider the difference
between the respondent’s age and the partner’s age. In
general, as the age gap increases, the older partner tends to
work longer, whereas the younger partner tends to leave the
labour market earlier (Dano et al. 2005). As previous
research of Denaeghel et al. (2011) indicated that both
small and wide educational gaps between partners do not
significantly influence retirement timing, we did not further
investigated measures concerning education. Household
size and family composition also affect retirement timing.
In general, having no dependents tends to accelerate
retirement. Households in which the partners are still in an
earlier stage of the life course are less likely to retire
(Henretta and O’Rand 1983).
Data and Methods
The data addressed in this study were drawn from the first
(2004–2005) and second (2006–2007) waves of the Sur-
vey of Health Ageing and Retirement in Europe
(SHARE). This dataset provided representative, longitu-
dinal, cross-nationally comparable microdata on the
health, economic, and social situations of older people in
Europe. Probability samples were drawn in each partici-
pating country. Given the diversity of institutional con-
ditions regarding sampling in the participating SHARE
countries, the sampling designs varied from the simple
random selection of households to multi-stage designs of
greater complexity. The weighted average household
response rate of the first wave was 62 %, ranging from
38 % in Switzerland to 81 % in France. From the par-
ticipating SHARE countries, we selected those countries
that participated in the first two waves of the survey:
Austria, Germany, Sweden, the Netherlands, Spain, Italy,
France, Denmark, Greece, Switzerland and Belgium
(Bo
¨rsch-Supan et al. 2005).
The current research used a subsample of the SHARE
dataset. In our study, the risk set consisted of men and
J Fam Econ Iss
123
women aged 50 years or older. We imposed several
restrictions with regard to our subsample. First, we inclu-
ded only respondents who had been part of a married or
cohabiting couple at Wave 1 and who were still living
together at Wave 2. We also included only respondents
whose partners had also been interviewed and were living
in the same household. The information on partners
enabled us to construct respondent–partner dyads in which
each participant is both a respondent and partner. It should
be noted that the respondent was always the person whose
retirement timing was being assessed. Second, all of the
respondents included were employed by an employer.
Those who were working in the household and those who
were seeking employment and self-employed people were
excluded from the risk set. Finally, we limited our study to
heterosexual dual-earner couples in which both the
respondent and the partner were 50 years of age or older.
As the number of homosexual couples in our risk set was
too low, we included only heterosexual couples. This
resulted in a final subsample of 930 respondents, reflecting
465 dual-earner couples.
Measures
Retirement Event
Retirement events were measured according to self-
reported retirement. During each wave of SHARE,
respondents were asked about their activity status, and the
category retired was used to define the retirement event.
During the observation period, 8.3 % of the respondents
retired. Our analysis did not allow for individuals leaving
retirement and re-entering the labour market. The control
variables used were gender and educational level. Gender
compared females (= 1) to males (= 0). As the world’s
educational systems vary widely, it can be difficult to
compare education levels across countries with different
systems of education. The International Standard Classi-
fication of Education (ISCED) provided a framework to
help statisticians classify diverse systems in such a way
that they can be compared across countries. Educational
level, based on the ISCED-97 codes, was measured as low
(ISCED 1–2), medium (ISCED 3–4) or high (ISCED
5–6), and having a low level of education was the refer-
ence category. The category low level of education con-
tained primary or first stage of basic education (ISCED 1)
and lower secondary or second stage of basic education
(ISCED 2). Medium level education consisted of upper
secondary education (ISCED 3) and post-secondary non-
tertiary education (ISCED 4). High level education was
measured as first stage of tertiary education (ISCED 5)
and second stage of tertiary education (ISCED 6) (UNE-
SCO Institute for Statistics 2012).
Respondent and Partner Characteristics
The health status of both the respondent and the partner
was included in the analysis. Nevertheless, the health
measures used posed several problems (Bazzoli 1985). The
literature commonly distinguishes between subjective and
objective health measures. Objective health refers to an
individual’s medically established health condition. Sub-
jective or self-reported health refers to the way in which
individuals feel about their health (Bound 1991; Dwyer and
Mitchell 1999). The use of self-reported health in an
empirical analysis of labour-force participation could be
problematic, however, as it is a potentially endogenous
explanatory variable (Kalwij and Vermeulen 2007). Cases
in which self-reported health reflects a preference for lei-
sure rather than true health capacity might lead to errors in
the estimation of health effects. Previous studies have
indicated that self-reported health measures appear to be
uncorrelated with compensation variables in retirement
equations, and they are therefore not endogenously deter-
mined by the labour supply (Dwyer and Mitchell 1999). In
this study, we investigated the impact of having a long-
term health problem, illness, disability or infirmity, mea-
sured as a dichotomous variable. We dichotomised work
regime as full-time or part-time. Measuring full-time or
part-time employment across the 11 European countries
from our sample was difficult, however, due to the lack of a
standard definition for part-time work (Lemaitre and Van
Bastelaer 1997) as well as to the possibility that respon-
dents who were working part-time might have preferred
full-time employment, or vice versa (Schils 2008). In order
to overcome this difficulty, we applied the uniform OECD
threshold of 30 standard working hours per week as the
criterion for part-time work (Lemaitre and Van Bastelaer
1997). Volunteer work was measured according to whether
the respondent or partner had done volunteer or charity
work in the month before the interview. Having a care task
was a dichotomous variable measuring whether the
respondent regularly helped someone living in the same
household with personal care, such as washing, getting out
of bed, or dressing during the last 12 months.
Household Measures
We distinguished four household factors in our analysis:
age gap between partners, household size, household
income, and country of residence. The age gap between
partners was measured by two dummy variables. The first
measured whether the respondent was older than the part-
ner, and the second measured whether the respondent and
partner were of the same age (i.e., less than 12 months
apart in age). The reference category consisted of respon-
dents whose partners were older. Household size was a
J Fam Econ Iss
123
numerical variable indicating the number of household
members. Household income was measured as the imputed
household-level variable household total income, which
equaled the sum of several income variables (e.g., income
from work, bank accounts, mutual funds, life insurance)
over all household members. Using the Square Root
Equivalence Scale (OECD 2008), we divided household
income by the square root of household size to develop an
idea of household welfare. We also transformed household
income into income deciles, based on the household’s
income position in the country of residence.
Analysis
As stated in the introduction, our aim was to understand
the retirement timing of older workers aged 50?living in
dual-earner households in 11 European countries. More
specifically, we questioned how measures that were sig-
nificantly related to retirement timing differed according
to whether one takes into account the dependence of men
and women living in the same household. In order to
answer this research question, we compared a proportional
hazard (Cox) model to a shared frailty model to observe
possible variation in retirement risks given that non-frailty
models violate the statistical assumption of independence
when clustering men and women in a dual-earner
household.
The Cox (1972) model is semi-parametric, due to the
parametric nature of the covariate term and the non-para-
metric baseline hazard function, which eliminates the
necessity of specifying some particular probability distri-
bution to represent survival times. Given the structure of
the SHARE dataset, we applied the discrete-time propor-
tional hazard method, assuming that events actually
occurred at exactly the same time (Allison 2005). The Cox
model is a marginal method which specifies models for the
effect of covariates on the hazard of individuals under the
working independent assumption (Wienke 2011).
A shared frailty model can be treated as a random
effects model for survival data, with group variation
(frailty) and individual variation described by the hazard
function. In shared frailty models, the failure times for
observations from the same cluster (i.e., the household)
correlate with one another. In other words, shared frailty
models allow the inclusion of correlated observations into
proportional hazard models, thus modelling associations
between event times. The shared frailty model is a condi-
tional independence model. This model assumes that all
event times in a cluster are independent given the frailty
variables. The common risk in each cluster (the frailty) is
assumed to be random. The frailty is common to all indi-
viduals in a cluster and therefore responsible for creating
dependence between event times (Wienke 2011).
Besides the methodological issue of correlated retire-
ment risks of partners, another aspect of this approach is to
focus on the frailty model as a way of dealing with possible
heterogeneity due to unobserved covariates. Basic survival
models assume that the study population is homogeneous
up to some observed covariates. If covariates are observed,
they can be included in the analysis by using, for example,
a proportional hazard model. It was, however, impossible
to include all important risk factors at the individual and
familial level influencing retirement timing. It is therefore
useful to consider two sources of variability in duration
data: theoretically predictable variability accounted for by
observable risk factors included in the model and hetero-
geneity caused by unknown covariates which is theoreti-
cally unpredictable (Wienke 2011). Taking into account all
relevant determinants of retirement timing would cause
frailty in any group of families to disappear. Frailty effects
reflect a diversity of factors such as behavioural or envi-
ronmental factors occurring at different levels. This study
limits itself to the individual and family level.
It is important to bear in mind the interpretation of the
various regression parameters. In the Cox model, the
parameters described relative risk at the population level.
The hazards ratio referred to comparisons between indi-
viduals randomly drawn from the total study population,
independent of the cluster to which an individual respon-
dent belongs. In the shared frailty model, the parameters
should be interpreted in terms of relative risk at the cluster
level. The hazard ratio referred to comparisons within
clusters where individuals share the same frailty. Conse-
quently, estimates were not expected to be the same in both
models, unless the within-cluster correlation was zero
which indicated that clustering did not play any role. If this
was the case, the shared frailty model complied with the
Cox model for independent data (Wienke 2011). All
analyses were carried out using SAS 9.3.
Results
Descriptive statistics are presented in Table 1. The sample
of analysis (Table 1) consisted of 930 respondents, 77 of
whom retired between the two waves. The respondents
were on average 54 years old. Thirty-seven percent
(37.31 %) were in the medium educated bracket, whereas
30.75 % were in the high educated bracket. Thirty-eight
percent (38.39 %) suffered from long-term illnesses,
17.85 % reported performing volunteer work, 1.37 %
reported taking care of dependents in the household, and
78.39 % worked full-time. Forty-three percent (43.44 %)
were older than their partner, whereas 13.12 % were the
same age as their partner. The mean household size of the
respondents in the sample equalled 2.62. Respondents were
J Fam Econ Iss
123
on average equally distributed among the income deciles
with slightly lower percentages of respondents in the first
deciles. Slightly more than one-quarter (26.12 %) reported
looking after grandchildren, and 80.03 % of the respon-
dents lived in a partnership.
The results of the Cox proportional hazard model and
the shared frailty model, including hazard ratios and 95 %
confidence intervals, are presented in Table 2. Model 1
examined the influence of individual, partner, and house-
hold characteristics on retirement timing, without taking
into account the clustering of partners within dual-earner
households. The findings suggested that educational level,
doing volunteer work, and having a partner performing
care tasks significantly influenced the retirement timing of
individuals. Medium-educated workers were likely to
remain on the labour market longer than lower-educated
workers (HR =0.558, p\0.05). Participation in volun-
teer work was positively associated with retirement timing
(HR =1.926, p\0.05). Our findings refuted the previous
findings of Lancee and Radl (2012), which suggested that
people who were more engaged tended to retire later as
they remain attached to their work until later in life through
self-selection into active ageing. Older workers having a
partner performing care tasks tended to time their retire-
ment earlier (HR =3.183, p\0.05). Informal care giving
responsibilities of one partner seemed to engage the other
partner to exit the labour market, possibly to facilitate the
balance between work and life within the household (Er-
ickson et al. 2010). Our analyses provided no evidence of
significant effects for the number of working hours, health
status, personal care tasks, and community engagement of
the partner.
An older worker who was the older partner
(HR =2.093, p\0.05) in a dual-earner household or who
was the same age (HR =2.120, p\005) as the partner
was significantly more likely to retire earlier, as compared
to an older worker who was the younger partner in the
household. Household size was negatively related to
retirement timing (HR =0.508, p\001). Our findings
indicated that a number of household members delayed
retirement. Having dependent children in the household
suggested that the household was still in an earlier stage of
the life course, thus resulting in delayed retirement (Hen-
retta and O’Rand 1983). Finally, household income did not
significantly influence retirement timing of older workers.
In Model 2, we used a shared frailty model that
explicitly controlled for the correlation of partners’ retire-
ment risks and unobserved heterogeneity. The presence of
a significant frailty suggested that retirement events were
likely to be clustered in certain families as partners’
retirement timing was correlated and that unmeasured (and/
or unmeasurable) factors shared by partners living in a
dual-earner household affected their retirement timing.
Unlike Model 1, Model 2 did not estimate a unique
retirement risk depending on the characteristics we selected
based on theory. As the retirement risk in Model 2
depended on both the measured factors and the unmeasured
family effect, its hazard ratios were family-specific and
measured the effect of a particular variable on the retire-
ment risk within a particular family (Cleves et al. 2008).
In order to answer our research question how measures
that are significantly related to retirement timing differ
according to whether one takes into account the depen-
dence of men and women living in the same household, we
compared the results of Model 1 and Model 2. In com-
parison to Model 1, the influence of educational level of the
respondent on retirement timing was similar, as was that of
the age gap when the respondent was older, and the size of
the household. Whereas Model 1 showed that older
workers who were medium educated were less likely to
retire later than lower-educated workers, Model 2 showed
that older workers who were medium educated had a lower
risk of retiring (HR =0.560, p\0.05) in comparison with
Table 1 Characteristics of older workers aged 50?in Europe
Variables and categories N%
Respondent
Mean age (STD) 54.38 3.72 (STD)
Gender (men) 465 50.0
Medium education level (ISCED 3–4) 347 37.31
High education level (ISCED 5–6) 283 30.75
Long-term illness 357 38.39
Volunteer work 166 17.85
Care task in the household 16 1.72
Working fulltime 729 78.39
Household
Age gap (Ref. Partner is older)
Respondent is older 404 43.44
Equal age 122 13.12
Mean household size (STD) 2.617 0.904 (STD)
Household income
Decile 1 48 5.16
Decile 2 44 4.73
Decile 3 64 6.88
Decile 4 86 9.25
Decile 5 106 11.40
Decile 6 124 13.33
Decile 7 112 12.04
Decile 8 118 12.69
Decile 9 140 15.05
Decile 10 88 9.14
Sample size 930
Events 77 8.3
Source of data SHARE, Wave 1 (2004/2005)
J Fam Econ Iss
123
lower-educated older workers, taking into account their
specific unobserved familial situations. Educational level
and household size (HR =0.601, p\0.01) had a slightly
stronger influence, while having an age gap between
partners where the respondent was older (HR =1.914,
p\0.05) had slightly less influence. In this model, how-
ever, doing volunteer work, having a partner with care
duties, and being member of a household where one’s
partner was of equal age, no longer had a significant effect
on the retirement timing of older workers belonging to
dual-earner households. In Model 2, the estimates of doing
volunteer work and having a partner of equal age both were
significant at trend level (i.e., respectively p\0.06,
p\0.09) indicating that the discrepancy of findings with
Model 1 was only moderate.
Discussion
Integrating the life-course approach and family systems
theory, our analysis advanced understanding with regard to
retirement trajectories by considering them not as indi-
vidual decisions, but as decisions taken within family
system thus depending upon the specific household con-
texts of specific families (Cox and Paley 1997,2003; Or-
range et al. 2003). Social science research tended to focus
on individuals, however. The attention given to the indi-
vidual is cultural. As the United States has dominated
social and behaviour research and is the most individual-
istic country in the world, the prevalence of individual
research is not surprising. Moreover, psychology which
emphasizes the individual before higher levels of analysis
Table 2 Retirement timing of
older workers from dual-earner
couples in Europe
Source of data SHARE
*p\0.05; ** p\0.01;
*** p\0.001
Measures Model 1 Model 2
Cox model Shared frailty model
Hazard ratio 95 % CI Hazard
ratio
95 % CI
Respondent
Gender (men) 0.726 0.397–1.327 0.704 0.387–1.281
Medium education level (ISCED 3–4) 0.558* 0.316–0.984 0.560* 0.319–0.982
High education level (ISCED 5–6) 0.594 0.309–1.141 0.583 0.305–1.117
Long-term illness 0.845 0.517–1.380 0.872 0.535–1.421
Volunteer work 1.926* 1.067–3.472 1.759 0.980–3.159
Care task in household 1.650 0.388–7.006 1.229 0.232–6.503
Working fulltime 1.183 0.559–2.502 1.201 0.568–2.536
Partner
Long-term illness 1.159 0.723–1.856 1.167 0.729–1.868
Volunteer work 0.884 0.460–1.700 0.919 0.484–1.746
Care task in household 3.183* 1.207–8.392 2.268 0.747–6.888
Working fulltime 1.311 0.705–2.439 1.252 0.675–2.321
Household
Age gap
Respondent is older 2.093* 1.130–3.876 1.914* 1.048–3.497
Equal age 2.120* 1.001–4.492 1.879 0.884–3.994
Household size 0.508** 0.333–0.776 0.601** 0.405–0.893
Household income 0.957 0.871–1.051 0.956 0.870–1.050
Country (Ref. =Switzerland)
Austria 5.305** 1.772–15.884 3.655* 1.300–10.275
Belgium 1.200 0.357–4.032 0.916 0.295–2.837
Denmark 0.543 0.163–1.805 0.444 0.146–1.352
France 2.504 0.795–7.883 1.804 0.615–5.291
Germany 1.147 0.347–3.788 0.848 0.276–2.605
Greece 0.802 0.089–7.248 0.833 0.128–5.414
Italy 3.341 0.527–21.157 1.245 0.163–9.498
The Netherlands 0.913 0.153–1.728 0.446 0.147–1.349
Spain 8.953** 1.844–43.198 1.970 0.242–16.068
Sweden 0.923 0.331–2.575 0.720 0.281–1.842
J Fam Econ Iss
123
has dominated research in the social and behaviour sci-
ences. Besides, the reliance on standard statistical methods
also contributed to this individualistic orientation. These
statistical techniques usually make the independence
assumption requiring the data from each individual in a
study to be unrelated to the data from every other indi-
vidual in the study (Kenny et al. 2006). Two partners living
in the same household are not simply two independent
individuals. Previous research (Pienta 2003; Szinovacz and
Deviney 2000) investigated spousal characteristics and
household factors influencing older workers’ retirement
decisions. These studies, however, overlooked the corre-
lation between partners. Consequently, our aim was to
understand the retirement timing of older workers aged
50?living in dual-earner households in 11 European
countries. We questioned how measures that are signifi-
cantly related to retirement timing differed according to
whether one takes into account the household context in the
retirement analysis. In order to answer this research ques-
tion, this study compared a proportional hazard (Cox)
model to a shared frailty model.
First, the influence of individual, partner, and household
characteristics on retirement timing was tested, without
taking into account the clustering of partners within dual-
earner couples. The findings suggested that medium-edu-
cated workers were likely to remain in the labour market
longer than lower-educated workers. The number of years
of education decreased job exit (Meghir and Whitehouse
1997). In contrast, older workers doing volunteer work
and older workers having a partner performing care tasks
tended to retire earlier. Older workers, who had a partner
who was the same age or was younger, tended to leave the
labour market earlier. Finally, older workers who were part
of a large household tended to retire later due to family
responsibilities.
Second, the influence of retirement predictors on
retirement timing was tested taking into account the cor-
relation in survival experiences of couples using a shared
frailty model. The significant frailty parameter in the
shared frailty model indicated that correlation exists. This
correlation illustrated that unobserved factors at the
household level influenced older workers’ retirement
decision. The findings suggested that older workers who
were medium educated and were part of large families had
a lower risk of retiring in comparison with lower-educated
workers and older workers from small families, taking into
account their specific unobserved familial situations. In
comparison to the Cox model, performing volunteer work
and having a partner with care responsibilities no longer
influenced individual’s retirement timing.
When comparing the Cox model to the frailty model,
our findings indicated that ignoring the clustering factor in
the survival of couples led to small standard errors and
therefore falsely significant estimates. When studying
retirement decisions in a household context, the shared
frailty model was more precise than the traditional Cox
model. The results showed that covariates that are depen-
dent among partners (e.g., household size, educational
level) changed the least between the standard model and
the shared frailty model. The effects of controlling for
unobserved family heterogeneity on parameter values and
standard errors, although fairly modest, appeared, however,
to be important, because they revealed a systematic pattern
in the direction of bias in estimates that assume an inde-
pendent sample. Since people tend to marry within their
social group or to marry a person who is close to them in
status, homogamy by education is widespread among dual-
earner couples. These homogamous tendencies are strong
also according to age, cognitive ability, personal traits,
family background, and work preferences (Ha
¨rko
¨nen
2007). Previous research indicated, however, that cohab-
iting couples were less homogamous than married couples
(Blackwell and Lichter 2000). It would therefore be
interesting to see how measures that are significantly
related to retirement timing differ when taking into account
the clustering of men and women in cohabiting dual-earner
couples. As cohabiting couples are less homogamous,
indicating less dependency of covariates among partners,
one might expect more variation in hazard ratios when
comparing Cox and shared frailty models.
Limitations
Life course theory asserts that linked lives are embedded
within socio-historical contexts (Elder 1999). The shift
from domestic activities to paid work affected the labour
market participation of women and increased the preva-
lence of dual-earner households (Hall and MacDermid
2009), in addition to changing the modal structure of
families from extended to nuclear. Increasing divorce rates
and an increasing heterogeneity of family forms transform
the primary function of the family from that of a social
institution to that of an emotional supportive environment
(Bengtson 2001). Families’ lives are linked far beyond the
nuclear unit (Kohli et al. 2005). Bengtson (2001) empha-
sised the increasing importance of multigenerational bonds,
defining the family as a generation system that goes beyond
the nuclear household and intensifies various types of
solidarity including personal and instrumental support. In
the light of an ageing society, these multigenerational
bonds place new demands on families in terms of support
and care for the elderly and grandchildren (Kohli et al.
2005). Despite the results of previous research concerning
the influence of informal caregiving on the retirement
transition (Dentinger and Clarkberg 2002) indicating that
the political goal to raise pension ages increases the
J Fam Econ Iss
123
potential for a work-family imbalance for informal care-
givers (Zuba and Schneider 2013), the present study
focused solely on dual-earner nuclear families, thus
ignoring generational lineage.
Several other limitations of the present analysis should
be noted. In the present study, we did not consider differ-
ences between dual-earner households, even though dif-
ferent dual-earner families experience different work and
family circumstances and use a variety of strategies for
arranging work and family tasks. Facing both employment
and domestic obligations, couples allocate their collective
tasks in a variety of ways. (Hall and MacDermid 2009; Hill
et al. 2004). Previous research has contributed various ty-
pologies for dual-earner couples (Hall and MacDermid
2009; Johnson et al. 1992). Following Duxbury et al.
(2007), the modern dual-income family can best be con-
ceptualised according to gender, with the occupational
status of the male and of the female partner resulting in
four dual-income types: dual-career couples (both partners
are engaged in managerial or professional work), dual-
earner couples (both partners have jobs rather than careers),
status-reversal couples (the female partner has a career, the
male partner has a job) and new-traditional couples (the
male partner has a career, the female partner has a job).
Differences between types of dual-worker couples might
mediate the retirement timing of older workers in dual-
earner couples. It would therefore be interesting to examine
the influence of the household context for various types of
dual-worker households. Furthermore, the nature of the
sample population should be considered. The older work-
ing couples in the sample were a select group of survivors
where neither spouse was retired, a homemaker, divorced,
died or left the labour market. This resulted in a small
subsample of 465 dual-earner couples from the large
SHARE dataset. Consequently, caution was needed when
interpreting these results.
Although this study demonstrated that retirement risks
varied due to unobserved factors at the family level,
unobserved heterogeneity at the national level might not be
neglected. National differences might also be an important
factor to consider when studying the influence of the family
context on retirement timing. For example, families in
Europe share many features that are common to all coun-
tries (e.g. weakening marriages with increasing age), in
addition to displaying major differences between countries
(Kohli et al. 2005). At the European level, southern
countries are often grouped as strong-family countries,
whereas those in the centre and north of Europe are char-
acterised as weak-family countries (Reher 1998). The
weak–strong dichotomy refers to cultural patterns of family
centrality and loyalties, as well as to co-residence with
adult children and other family members and the organi-
sation of support for the elderly. In general, Scandinavian
countries have the least traditional family structures,
whereas southern countries have the most traditional
structures, with other continental countries lying in
between. Cultural factors (e.g., the priority of family in
life) might pose important obstacles that keep older
workers in the labour market longer. Cultural factors might
be particularly important predictors of retirement timing
for older employees, who typically have greater financial
sustainability (Phillipson 2002).
Given the small number of dual-earner households
across the 11 European countries included in our dataset,
we were not able to consider possible national differences.
The dichotomy between weak-family and strong-family
countries might not be an oversimplification (Glaser et al.
2004). The North–South gradient is especially noticeable
with respect to rates of co-residence and frequency of
contact among adult family generations (Kohli et al. 2005).
Given that our analysis focused on the extent to which
retirement predictors were affected by the clustering of
individuals into dual-earner nuclear households, and given
that we did not aim to capture the influence of the extended
family and the family context as such, we believe that
possible country differences are only of minor importance.
Nevertheless, further research is needed.
Conclusion
With this research we aimed to counterbalance the indi-
vidualistic line of thought in retirement analysis in social
sciences. Two partners living in the same household are not
simply two independent individuals. Popular in epidemio-
logical work but less so in the sociological literature, we
used shared frailty models instead of standard proportional
hazard models to predict hazard ratios for retirement timing
of dual-earners since they explicitly acknowledge and
adjust for the notion that individuals within the same
household have a similar likelihood of retirement (Denney
2010). Results obtained from the comparison between the
Cox and shared frailty model are of both methodological
and policy interest. As research results are used to inform
and evaluate policy measures such as those aimed at
extending working life, it is important that research-based
policy measures acknowledge the impact of family-level
influence on labour market exit. Despite the shortcomings
mentioned above, the results of the analyses presented in
this article clearly indicated the importance of clustering
male and female partners into households when studying
retirement timing. Our findings might drive future
researchers and policy makers to acknowledge the impor-
tance of taking into account the influence of household
constellation on life events which are at first sight indi-
vidual, such as retirement timing. Besides, given increasing
J Fam Econ Iss
123
variation in household constellation, emergence of one-
income families and reconstituted families, researchers and
policy makers should question how findings of social
analyses should be interpreted correctly given the general
under-acknowledgement of household context when
studying life events.
Acknowledgments This paper uses data from SHARE release
2.5.0, as of May 24 2011. The SHARE data collection was primarily
funded by the European Commission through the 5th framework
programme (project QLK6-CT-2001-00360 in the thematic pro-
gramme Quality of Life), through the 6th framework programme
(projects SHARE-I3, RII-CT- 2006-062193, COMPARE, CIT5-CT-
2005-028857, and SHARELIFE, CIT4-CT-2006-028812) and
through the 7th framework programme (SHARE-PREP, 211909 and
SHARE-LEAP, 227822). Additional funding from the U.S. National
Institute on Aging (U01 AG09740-13S2, P01 AG005842, P01
AG08291, P30 AG12815, Y1-AG-4553-01 and OGHA 04-064, IAG
BSR06-11, R21 AG025169) as well as from various national sources
is gratefully acknowledged (see www.share-project.org for a full list
of funding institutions).
References
Allison, P. D. (2005). Survival analysis using SAS. A practical guide.
Cary, NC: SAS Institute.
Bazzoli, G. J. (1985). The early retirement decision: new empirical
evidence on the influence of health. The Journal of Human
Resources, 20(2), 214–234. http://www.jstor.org/stable/146009.
Beehr, T. A., Glazer, S., Nielson,N. L., & Farmer, S. J. (2000).Work and
nonwork predictors of employees’ retirement ages. Journal of
VocationalBehavior, 57(2),206–225. doi:10.1006/jvbe.1999.1736.
Bengtson, V. L. (2001). Beyond the nuclear family: The increasing
importance of multigenerational bonds. Journal of Marriage and
Family, 63(1), 1–16. doi:10.1111/j.1741-3737.2001.00001.x.
Bingley, P., & Lanot, G. (2007). Public pension programmes and the
retirement of married couples in Denmark. Journal of Public
Economics, 91(10), 1878–1901. doi:10.1016/j.jpubeco.2007.09.006.
Blackwell, D.L., & Lichter, D. T. (2000). Mate selection among married
and cohabiting couples. Journal of Family Issues, 21(3), 275–302.
Blossfeld, H. P., & Drobnivc, S. (2001). Careers of couples in
contemporary society: From male breadwinner to dual-earner
families. Oxford: Oxford University Press.
Bo
¨rsch-Supan, A., Brugiavini, A., Ju
¨rges, H., Mackenbach, J.,
Siegrist, J., & Weber, G. (2005). Health, ageing and retirement
in Europe: First results from the Survey of Health, Ageing and
Retirement in Europe. Mannheim: Mannheim Research Institute
for the Economics of Aging.
Bound, J. (1991). Self-reported versus objective measures of health in
retirement models. [Article]. Journal of Human Resources,
26(1), 106–138. http://www.jstor.org/stable/145718.
Bronfenbrenner, U. (1977). Toward an experimental ecology of
human development. American Psychologist, 32(7), 513–531.
Clarkberg, M., & Moen, P. (2001). Understanding the time-squeeze
married couples’ preferred and actual work-hour strategies.
American Behavioral Scientist, 44(7), 1115–1136. doi:10.1177/
0002764201044007005.
Cleves, M. A., Gould, W. W., & Gutierrez, R. G. (2008). An
introduction to survival analysis using Stata. College Station,
TX: Stata Press.
Cox, D. R. (1972). Regression models and life-tables. Journal of the
Royal Statistical Society. Series B (Methodological), 34(2),
187–220. http://www.jstor.org/stable/2985181.
Cox, M. J., & Paley, B. (1997). Families as systems. Annual Review
of Psychology, 48(1), 243–267. doi:10.1146/annurev.psych.48.1.
243.
Cox, M. J., & Paley, B. (2003). Understanding families as systems.
Current Directions in Psychological Science, 12(5), 193–196.
http://www.jstor.org/stable/20182875.
Dahl, S. A., Nilsen, O. A., & Vaage, K. (2003). Gender differences in
early retirement behaviour. European Sociological Review,
19(2), 179–198. doi:10.1093/esr/19.2.179.
Dano, A. M., Ejrnaes, M., & Husted, L. (2005). Do single women
value early retirement more than single men? Labour Econom-
ics, 12(1), 47–71. doi:10.1016/j.labeco.2004.03.002.
Denaeghel, K., Mortelmans, D., & Borghgraef, A. (2011). Spousal
influence on the retirement decisions of single-earner and dual-
earner couples. Advances in life Course Research, 16(3),
112–123.
Denney, J. T. (2010). Family and household formations and suicide in
the United States. Journal of Marriage and Family, 72(1),
202–213. doi:10.1111/j.1741-3737.2009.00692.x.
Dentinger, E., & Clarkberg, M. (2002). Informal caregiving and
retirement timing among men and women. Journal of Family
Issues, 23(7), 857–879. doi:10.1177/019251302236598.
Duxbury, L., Lyons, S., & Higgins, C. (2007). Dual-income families
in the new millennium: Reconceptualizing family type.
Advances in Developing Human Resources, 9(4), 472–486.
doi:10.1177/1523422307305488.
Dwyer, D. S., & Mitchell, O. S. (1999). Health problems as
determinants of retirement: Are self-rated measures endogenous?
Journal of Health Economics, 18(2), 173–193. doi:10.1016/
S0167-6296(98)00034-4.
Elder, G. H. (1999) The life course and aging: Some reflections.
Distinguished Scholar Lecture Section on Aging American
Sociological Association (pp. 26).
Erickson, J. J., Martinengo, G., & Hill, E. J. (2010). Putting work and
family experiences in context: Differences by family life stage.
Human Relations, 63(7), 955–979. doi:10.1177/0018726709
353138.
Ginn, J., & Arber, S. (1996). Gender, age and attitudes to retirement
in mid-life. Ageing and Society, 16(1), 27–55. doi:10.1017/
S0144686X00003123.
Glaser, K., Tomassini, C., & Grundy, E. (2004). Revisiting conver-
gence and divergence: Support for older people in Europe.
European Journal of Ageing, 1(1), 64–72. doi:10.1007/s10433-
004-0006-1.
Gustman, A., & Steinmeier, T. (2000). Retirement in dual-career
families: A structural model. Journal of Labor Economics, 18(3),
503–545. doi:10.1086/209968.
Gustman, A., & Steinmeier, T. (2004). Personal accounts and family
retirement (no. w10305). National Bureau of Economic
Research, NBER Working Papers, Cambridge.
Hall, S. S., & MacDermid, S. M. (2009). A typology of dual earner
marriages based on work and family arrangements. Journal of
Family and Economic Issues, 30(3), 215–225. doi:10.1007/
s10834-009-9156-9.
Ha
¨rko
¨nen, J. (2007). Jobless couples in Europe. Comparative studies
with longitudinal data. Florence: European University Institute.
Hatcher, C. B. (2002). Wealth, reservation wealth, and the decision to
retire. Journal of Family and Economic Issues, 23(2), 167–187.
Henkens, K., & Van Solinge, H. (2002). Spousal influences on the
decision to retire. International Journal of Sociology, 32(2),
55–74. http://www.jstor.org/stable/20628649.
Henretta, J. C., & O’Rand, A. M. (1983). Joint retirement in the dual
worker family. Social Forces, 62(2), 504–520. http://www.jstor.
org/stable/2578319.
Hill, E. J., Erickson, J. J., Fellows, K. J., Martinengo, G., & Allen, S.
M. (2014). Work and family over the life course: Do older
J Fam Econ Iss
123
workers differ? Journal of Family and Economic Issues, 35(1),
1–13. doi:10.1007/s10834-012-9346-8.
Hill, E. J., Ma
¨rtinson, V. K., Ferris, M., & Baker, R. (2004). Beyond
the mommy track: The influence of new-concept part-time work
for professional women on work and family. Journal of Family
and Economic Issues, 25(1), 121–136. doi:10.1023/B:JEEI.
0000016726.06264.91.
Ho, J.-H., & Raymo, J. M. (2009). Expectations and realizations of
joint retirement among dual-worker couples. Research on Aging,
31(2), 153–179. doi:10.1177/0164027508328308.
Hurd, M. D. (1990). The joint retirement decision of husbands and
wives. In D. A. Wise (Ed.), Issues in the Economics of Aging.
Chicago: The University of Chicago Press.
Johnson, M. P., Huston, T. L., Gaines, S. O., & Levinger, G. (1992).
Patterns of married life among young couples. Journal of Social
and Personal Relationships, 9(3), 343–364. doi:10.1177/026540
7592093002.
Kalwij, A., & Vermeulen, F. (2007). Health and labour force
participation of older people in Europe: What do objective health
indicators add to the analysis? Health Economics, 17(5),
619–638. doi:10.1002/hec.1285.
Kapur, K., & Rogowski, J. (2007). The role of health insurance in
joint retirement among married couples. Industrial and Labor
Relations Review, 60(3), 397–408. http://www.jstor.org/stable/
25249093.
Kenny, D. A., Kashy, D., & Cook, W. L. (2006). Dyadic data
analysis. New York: Guilford Press.
Kohli, M., Ku
¨nemund, H., & Lu
¨dicke, J. (2005). Family structure,
proximity and contact. In A. Bo
¨rsch-Supan, A. Brugiavini, H.
Ju
¨rges, J. Mackenbach, J. Siegrist, & G. Weber (Eds.), Health,
ageing and retirement-first results from the survey of health,
ageing and retirement in Europe (pp. 164–170). Mannheim:
Mannheim Research Institute fu
¨r the Economics of Ageing.
Kreppner, K., & Lerner, R. M. (1989). Family systems and life-span
development: Issues and perspectives. Hillsdale: Lawrence
Erlbaum Associates Inc.
Kubicek, B., Korunka, C., Hoonakker, P., & Raymo, J. M. (2010).
Work and family characteristics as predictors of early retirement
in married men and women. Research on Aging, 32(4), 467–498.
doi:10.1177/0164027510364120.
Lancee, B., & Radl, J. (2012). Social connectedness and the transition
from work to retirement. The Journals of Gerontology Series B:
Psychological Sciences and Social Sciences, 67(4), 481–490.
doi:10.1093/geronb/gbs049.
Lemaitre, G. M. P., & Van Bastelaer, A. (1997). International compar-
isons of part-time work. OECD Economic Studies, 29(II), 139–151.
Loretto, W., & Vickerstaff, S. (2012). The domestic and gendered
context for retirement. Human Relations, 66(1), 65–86. doi:10.
1177/0018726712455832.
Meghir, C., & Whitehouse, E. (1997). Labour market transitions and
retirement of men in the UK. Journal of Econometrics, 79(2),
327–354. doi:10.1016/S0304-4076(97)00026-2.
Minnotte, K. L. (2012). Family structure, gender, and the work–
family interface: Work-to-family conflict among single and
partnered parents. Journal of Family and Economic Issues,
33(1), 95–107. doi:10.1007/s10834-011-9261-4.
Moen, P. (1996). A life course perspective on retirement, gender, and
well-being. Journal of Occupational Health Psychology, 1(2),
131–144. doi:10.1037/1076-8998.1.2.131.
Moen, P., Fields, V., Quick, H., & Hofmeister, H. (2000). A life-
course approach to retirement and social integration. In K.
Pillemer, P. Moen, E. Wethington, & N. Glasgow (Eds.), Social
integration in the second half of life (pp. 75–107). Baltimore:
The Jopkins University Press.
Moen, P., Huang, Q., Plassmann, V., & Dentinger, E. (2006).
Deciding the future. Do dual-earner couples plan together for
retirement? American Behavioral Scientist, 49(9), 1–22. doi:10.
1177/0002764206286563.
OECD. (2008). Growing unequal? Income distribution and poverty in
OECD countries. Paris: OECD.
Omariba, D. W. R., Beaujot, R., & Rajulton, F. (2007). Determinants
of infant and child mortality in Kenya: An analysis controlling
for frailty effects. Population Research and Policy Review,
26(3), 299–321. doi:10.1007/s11113-007-9031-z.
O’Rand, A. M., & Farkas, J. I. (2002). Couples’ retirement timing in
the United States in the 1990s: The impact of market and family
role demands on joint work exits. International Journal of
Sociology, 32(Summer), 11–29. http://www.jstor.org/stable/2062
8647.
Orrange, R. M., Firebaugh, F. M., & Heck, R. K. (2003). Managing
households. In P. Moen (Ed.), It’s about time: Couples and
careers (pp. 153–167). New York: Cornell University Press.
Parker, S. C., & Rougier, J. C. (2007). The retirement behaviour of
the self-employed in Britain. Applied Economics, 39(4–6),
697–713. doi:10.1080/00036840500447807.
Pedersen, D. E., & Minnotte, K. L. (2012). Dual earner husbands and
wives: Marital satisfaction and the workplace culture of each
spouse. Journal of Family and Economic Issues, 33(3), 272–282.
doi:10.1007/s10834-012-9294-3.
Phillipson, C. (2002). Transitions from work to retirement: Develop-
ing a new social contract. Bristol: The Policy Press.
Piekkola, H. (2008). Flexible pension systems: Postponed retirement
and distributional fairness. European Network of Economic
Policy Research Institutes, Social Welfare Policies: ENEPRI
Research Reports (pp. 1–36): ENEPRI.
Pienta, A. M. (2003). Partners in marriage: An analysis of husbands’
and wives’ retirement behavior. The Journal of Applied Geron-
tology, 22(3), 340–358. doi:10.1177/0733464803253587.
Pienta, A. M., & Hayward, M. D. (2002). Who expects to continue
working after age 62? The retirement plans of couples. Journals
of Gerontology Series B: Psychological Sciences and Social
Sciences, 57(4), S199–S208. doi:10.1093/geronb/57.4.S199.
Raymo, J. M., & Sweeney, M. M. (2006). Work-family conflict and
retirement preferences. Journals of Gerontology Series B:
Psychological Sciences and Social Sciences, 61(3), S161–S169.
Reher, D. S. (1998). Family ties in Western Europe: Persistent
contrasts. Population and Development Review, 24(2), 203–234.
http://www.jstor.org/stable/2807972.
Ruhm, C. J. (1996). Gender differences in employment behavior
during late middle age. Journals of Gerontology Series B:
Psychological Sciences and Social Sciences, 51(1), S11–S17.
doi:10.1093/geronb/51B.1.S11.
Schils, T. (2008). Early retirement in Germany, the Netherlands, and
the United Kingdom: A longitudinal analysis of individual
factors and institutional regimes. European Sociological Review,
24(3), 315–329. doi:10.1093/esr/jcn009.
Schmid, G. (1998). Transitional labour markets: A new European
employment strategy. In W. B. f. Sozialforschung (Ed.),
Discussion Paper (Vol. FS I 98–206). Berlin.
Smith, D. B., & Moen, P. (1998). Spousal influence on retirement:
His, her, and their perceptions. Journal of Marriage and Family,
60(3), 734–744. http://www.jstor.org/stable/353542.
Szinovacz, M. E., & Deviney, S. (2000). Marital characteristics and
retirement decisions. Research on Aging, 22(5), 470–498.
UNESCO Institute for Statistics. (2012). ISCED: International
standard classification of education. Accessed 25 May, 2013
from http://www.uis.unesco.org/EDUCATION/Pages/interna
tional-standard-classification-of-education.aspx.
Van Gils, W., & Kraaykamp, G. (2008). The emergence of dual-
earner couples. A longitudinal study of the Netherlands.
International Sociology, 23(3), 345–366. doi:10.1177/
0268580908088894.
J Fam Econ Iss
123
Van Solinge, H., & Henkens, K. (2005). Couples’ adjustment to
retirement: A multi-actor panel study. The Journals of Geron-
tology Series B: Psychological Sciences and Social Sciences,
60(1), S11–S20. doi:10.1093/geronb/60.1.S11.
Voydanoff, P. (2005). Work demands and work-to-family and family-
to-work conflict. Direct and indirect relationships. Journal of
Family Issues, 26(6), 707–726. doi:10.1177/0192513X05277516.
Wienke, A. (2011). Frailty models in survival analysis. Boca Ration:
Taylor & Francis Group.
Zuba, M., & Schneider, U. (2013). What helps working informal
caregivers? The role of workplace characteristics in balancing work
and adult-care responsibilities. Journal of Family and Economic
Issues, 34(4), 460–469. doi:10.1007/s10834-012-9347-7.
Hanne De Preter is a Ph.D. student at the Department of Sociology
at the University of Antwerp’s Research Centre for Longitudinal and
Life Course Studies. Currently, she is working on a research project
financed by the Flemish government, studying the pension protection
of migrants in Belgium. Her Ph.D. focuses on retirement timing after
the age of 50 in Europe. She has authored a number of national and
international peer-review publications
Dorien Van Looy is a research and teaching assistant at the Research
Centre for Longitudinal and Life Course Studies (CELLO) at the
University of Antwerp (Department of Sociology). She is currently
preparing a Ph.D. on the relationship between end-career working
hours reduction, work-life balance and retirement timing. She is also
involved in different projects on retirement timing and working career
transitions financed by the Flemish government
Dimitri Mortelmans is senior full professor in Sociology at the
Faculty of Political and Social Sciences of the University of Antwerp
(Belgium). He obtained his Ph.D. at the University of Antwerp in
1997. He is head of the Centre for Longitudinal and Life Course
Studies (CELLO). His research concentrates on family sociology and
sociology of labour. He has published on divorce, new constituted
families, gendered labour careers, early retirement and work-life
balance. He is also the main author of the Step in Statistics book
series and the Handbook of Qualitative Research Methods and
Qualitative Analysis with Nvivo
J Fam Econ Iss
123
