Journal of Gerontology: SOCIAL SCIENCES
2008, Vol. 63B, No. 2, S99–S109

Copyright 2008 by The Gerontological Society of America

The Effects of Socioeconomic Status and Health on
Transitions in Living Arrangements and Mortality: A
Longitudinal Analysis of Elderly Finnish Men and
Women From 1997 to 2002
Pekka Martikainen,1,2 Elina Nihtila,2 and Heta Moustgaard1,2
¨
1
Helsinki Collegium for Advanced Studies, Helsinki, Finland.
Population Research Unit, Department of Sociology, University of Helsinki, Finland.

2

Objectives. Objectives were to study the effects of socioeconomic factors on transitions in living arrangements and
mortality for men and women.
Methods. We used a sample of Finns aged 65 years and older living alone or with a partner at the end of 1997 (N ¼
250,787) drawn from population registers, and followed them up for transitions in living arrangements (with partner,
alone, with others, institutionalized) and death at the end of 2002.
Results. Health conditions associated with functional difficulties were major determinants of institutionalization and
death and were associated with transitions between private households. Low income among men and in particular not
owning a home were independently associated with institutionalization and death among those living alone or with
a partner at baseline. Among those living with a partner, the transition to living alone was associated with all
socioeconomic factors but most strongly with a low income and not owning a home. Transitions to living with others were
associated in particular with low occupational social class and education.
Discussion. Variations in the associations of different socioeconomic indicators with living arrangement transitions
imply different social pathways. However, material socioeconomic indicators dominated other measures of socioeconomic
status in determining such transitions, and their effects were only partly mediated by chronic conditions.
Key Words: Living arrangements—Socioeconomic status—Finland.

I

T is widely acknowledged that living arrangements are
important determinants of well-being (Davis, Moritz,
Neuhaus, Barclay, & Gee, 1997; De Jong Gierveld & Van
Tilburg, 1999; Grundy, 2001), informal care (Kemper, 1992;
Lafreniere, Carriere, Martel, & Belanger, 2003), and formal
institutional care (Branch & Jette, 1982; Steinbach, 1992;
Wolinsky, Callahan, Fitzgerald, & Johnson, 1992) in old age.
Rapid secular changes in the distribution of living arrangements
may have significant implications for the well-being of elderly
persons and may have cost implications for the taxpayer
(Chappell, Dlitt, Hollander, Miller, & McWilliam, 2004). In
particular, in developed countries the number of elderly people
living alone has generally increased in the past two decades,
whereas fewer live with their children (United Nations
Population Division, 2005). However, at the same time the
increase in life expectancy—particularly among men—and the
narrowing gap in age at marriage between the sexes
(Lakdawalla & Schoeni, 2003) has led to an increase in the
proportion of older men and women living with a partner.
Many of the trends in living arrangements are most pronounced in North America and northwest Europe (Tomassini,
Glaser, Wolf, Broese van Groenou, & Grundy, 2004; United
Nations Population Division, 2005). However, on the individual level the quantum and determinants of transitions in
living arrangements remain only partly understood, with the
most robust evidence being from the United States (Branch &

Jette, 1982; Davis et al., 1997; Hays, Pieper, & Purser, 2003;
Liang, Brown, Krause, Ofstedal, & Bennett, 2005; Mutchler &
Burr, 1991; Wilmoth, 1998; Wolf & Soldo, 1988; Wolinsky
et al., 1992) but also from some other countries (e.g., Brown
et al., 2002; Evandrou, Falkingham, Rake, Scott, and the ESRC
Research Group Simulating Social Policy for an Ageing
Society, 2001; Glaser, Grundy, & Lynch, 2003; Grundy &
Jitlal, 2007; Nihtila & Martikainen, 2007b; Sarma & Simpson,
¨
2007; Tomiak, Berthelot, Guimond, & Mustard, 2000). We
used a large longitudinal population registration data set to
study the determinants of transitions in living arrangements and
mortality among Finnish elders.

Determinants of Living Arrangements Among Elders
Some of the major determinants of transitions in living arrangements are prior living arrangements, age, and gender (e.g.,
Branch & Jette, 1982; Brown et al., 2002; Glaser et al., 2003;
Nihtila & Martikainen, 2007b; Steinbach, 1992; Wolinsky
¨
et al., 1992). Prior living arrangements partly determine the
transition options that are possible. However, the majority of
previous studies on such transitions have targeted unmarried
populations or people living alone (particularly women),
mainly because variability and change are more extensive in
these population subgroups (Wilmoth, 1998). This is an unnecessary barrier to a broader understanding of living arrangements among elderly persons and limits the policy relevance of

S99

S100

MARTIKAINEN ET AL.

research findings. Our focus was on people living alone or with
a spouse or partner at baseline. These two living arrangements
cover the majority of elders in northwestern Europe and the
United States: More than 80% of men and women older than 65
typically live alone or as a married couple, the proportion of
women living alone being much higher than that of men
(Tomassini et al., 2004).
Many of the studies on the determinants of transitions in
living arrangements have analyzed very broad age groups, often
focusing on the combined population of all persons older than
age 55. The results of such analyses are heavily influenced by
the transitions of the young old, who are the most populous in
such data sets. These analyses show that transitions are rare
events. However, this kind of generalization is likely to best
apply to those younger than the age of 65 or 70, a period of life
before the onset of severe disability, illness, or institutionalization. Given the fact that these events are strongly gendered,
living arrangement transitions will also be very different for
men and women. We took this into account in our study by
stratifying the analyses by gender and living arrangement at
baseline.
Furthermore, our analyses focused on two additional explanatory factors: (a) socioeconomic status and (b) chronic illness.
The emergence of poor health has, of course, major repercussions for elderly persons and may in many cases lead
to transitions in living arrangements. In particular, chronic
illness and functional disabilities (Aguero-Torres, von Strauss,
Viitanen, Winblad, & Fratiglioni, 2001; Branch & Jette, 1982;
Shapiro & Tate, 1988) are major determinants of entry into institutional care. Furthermore, functional status has been associated
with household expansion (e.g., moving in with children or
other relatives and adults; Hays, 2002; Hays et al., 2003).
Our study also emphasized the effects of different socioeconomic circumstances on transitions in living arrangements.
Prior research in the United States in particular has indicated
that low income is associated with a greater risk of entry into
institutional care (Greene, Lovely, Miller, & Ondrich, 1995;
Himes, Wagner, Wolf, Aykan, & Dougherty, 2000) and
coresidence with others (Mutchler & Burr, 1991). Other socioeconomic characteristics are also associated with living arrangement transitions: Homeowners and those with a higher
education are more likely to continue living with a spouse or
partner and have a lower risk of moving into long-term
institutional care (Breeze, Sloggett, & Fletcher, 1999; Liang
et al., 2005; Nihtila & Martikainen, 2007a). These effects are
¨
not always consistent, and their magnitudes vary. Attempts to
explicitly compare the effects of different measures of
socioeconomic status are rare, and it remains unclear whether
these effects are independent of other main determinants of
living arrangement transitions.
Much of the variability in results may relate to the different
measures of socioeconomic position used in the different
studies. Various indicators of socioeconomic position are correlated and are sometimes treated as though they are interchangeable. However, these indicators may instead be partially
independent determinants of living arrangements, and they may
reflect different aspects of social stratification and different
causal pathways (Bartley, Sacker, Firth, & Fitzpatrick, 1999;
Grundy & Holt, 2001; Lahelma, Martikainen, Laaksonen, &
Aittomaki, 2004; Oakes & Rossi, 2003). Education reflects

experiences of early life, when educational qualifications are
usually obtained and when the seeds are sown for many
behavioral patterns, attitudes, and aptitudes to obtain and make
use of health knowledge, which may last until later life.
Occupational social class, in contrast, mirrors experiences and
exposures in adult life. These reflect status and power but may
partly relate to past material conditions and exposure in the
sphere of paid work. Income provides individuals and families
with necessary material resources and determines their
purchasing power. Given the major gender differences in the
educational and labor market histories of those who are
currently aged 65 and older, stratified analysis for both men
and women is justified. Furthermore, many of the studies that
have assessed the association between socioeconomic factors
and living arrangements have been cross-sectional, and it is also
unclear to what extent these effects are independent of or
mediated by higher incidence of illness and disability among
those lower down in the social hierarchy.
The large majority of studies on living arrangements and
their transitions have investigated elderly U.S. populations.
However, national variations in family setting and policy environments of elder care may mean that the transition patterns
vary between countries. For example, in Finland long-term
institutional care of elderly persons is more common than in
southern European countries but about average in comparison
with northwest Europe and the United States (Kinsella &
Velkoff, 2001; Rostgaard, 2002; United Nations Population
Division, 2005). This care is mainly publicly provided in
nursing homes and health centers, and user fees, up to
a maximum of 80%, are related to disposable income.

Aims
We used a 40% sample (N ¼ 250,787) of Finnish men and
women aged 65 years and older living alone or with a partner
on December, 31, 1997, drawn from population registration
data, in order to evaluate transitions in living arrangements and
mortality at the end of 2002. These data had several
methodological advantages that have enabled us to fill in the
research gaps identified previously: (a) The data were
longitudinal and thus allowed for a better assessment of causal
direction, particularly in terms of the effects of socioeconomic
status and health conditions on living arrangement transitions;
(b) because these data were based on population registration
there was practically no loss due to follow-up or self-report
bias; (c) we had a uniquely large sample to study and systematically compare many possible transitions, including
transitions into institutions; and (d) the data represented a
non-U.S. population with a low prevalence of elder coresidence
with children and universal coverage of health and social care
for elderly persons; they thus allowed for the testing of previous
U.S.-based results in a different social and policy context.
Furthermore, given that many of the living arrangement transitions as well as the distribution of their determinants—such as
the prevalence of disease or education—vary strongly between
men and women, we carried out separate analyses for both.
Our specific research aims were the following:
1. To quantify transitions in living arrangements between
private households (with a partner, alone, with others),
institutionalization, and mortality among men and women

LIVING ARRANGEMENT TRANSITIONS

living with a partner or living alone at baseline over a 5-year
period.
2. To quantify the strength of the effects of different
socioeconomic indicators on these transitions for men and
women separately.
3. To estimate to what extent the effects of different socioeconomic indicators are overlapping or independent of one
another.
4. To estimate to what extent the independent effects of
different socioeconomic factors on transitions in living
arrangements are mediated by chronic illness among men
and women separately.

METHODS

Data
The data used for these analyses consisted of a 40% random
sample of the Finnish population aged 65 and older at the end
of 1997 drawn from population registration data. For this study
we included those (a) living with a married spouse or a
cohabiting partner or (b) living alone at baseline (101,885 men
and 148,902 women; this composed 89% of the total population
of men and women aged 65 and older living in private
households). We followed this population up for transitions in
living arrangements and mortality at the end of 2002. Statistics
Finland linked these data with the population registration
data on sociodemographic characteristics and information on
chronic health conditions at baseline. Statistics Finland provided data on basic sociodemographic characteristics and dates
of death, the National Research and Development Centre
for Welfare and Health (STAKES) provided information on
long-term institutional care and the principal causes of prior
hospitalization, and the Social Insurance Institution provided
information on purchases of prescribed medication and entitlement to reimbursement for drug costs. Information on income
came from the registers of the Finnish Tax Administration.
Statistics Finland carried out the data linkage using personal
identification codes.

Definition of Living Arrangements and
Survival Status at Follow-Up
The unit of analysis we used for defining living arrangements
was the household in which the person under study lived,
comprising all persons living in the same dwelling unit. We
thus defined subtenants (0.4% of Finns in 1990; Tilastokeskus,
1994) as part of the household, and more than one family may
have lived in the same unit. We used the same criteria to define
living arrangements at baseline and at follow-up.
We defined living arrangements and survival status at followup in the following way: (1) living with a spouse or a cohabiting
partner, (2) living alone, (3) living in other private households
(e.g., with children or other adults), (4) living in institutions
providing long-term 24-hr care, and (5) dead.
We obtained the information on living arrangements and
dates of death from the Statistics Finland population register.
The stock-and-flow data of residents in institutional care
(nursing homes, health centers and hospitals, and service
homes) came from STAKES.

S101

Socioeconomic Status
We used four measures of socioeconomic status: education,
social class, household income, and home ownership. The three
educational categories were based on the highest completed
degree or certificate: tertiary education, intermediate education,
and basic education or less or unknown.
We used four social classes: white collar, manual, farmer,
and other. We classified unemployed and retired persons according to their previous occupations, and housewives according to the occupation of the head of the household.
Household disposable income per consumption unit covered
all taxable income sources for all household members including
wages, capital income, and taxable income transfers but
excluding taxes. In the calculation of household consumption
units the first household member was weighted as 1.0 unit and
any other as 0.7 units. This corresponded to the Organisation
for Economic Co-operation and Development (OECD) equivalence scale (OECD, 1982), with the exception of children
younger than 18 years of age, who we weighted as adults
(having coresident minor children in this kind of study population is very rare). We divided income into quartiles with
cutoff points calculated from the combined data for elderly men
and women.
We categorized home ownership in two classes: owner
occupier and other.

Health at Baseline
In this study we used the following 18 dichotomous
indicators of chronic health conditions 2 years prior to baseline:
cancer, diabetes, dementia, psychosis, depressive symptoms,
other mental health disorder, Parkinson’s disease, other
neurological disease, heart disease, stroke, chronic asthma or
other similar chronic obstructive pulmonary disease, other
respiratory disease, arthritis, osteoarthritis, hip fracture, other
condition related to accident or violence, other hospital
diagnosis, and other chronic disease that qualifies for the right
to reimbursement for drug costs under the Special Refund
Categories, which cover 75% or 100% of the costs of a single
drug purchase. We used these as dichotomous variables in the
multivariate models.
We used two main register sources to assess chronic diseases
and conditions: (a) the principal cause of hospitalization during
1996–1997 and (b) the right to reimbursement for drug costs
under the Special Refund Categories for certain diagnosed
chronic medical conditions during 1997. We categorized the
persons studied as having a chronic condition if they had it
according to at least one of these sources. In addition, the
following six categories of medical conditions were supplemented with information on the purchase of prescription
medication during 1996–1997: cancer, diabetes, depressive
symptoms, other mental health problem, Parkinson’s disease,
and other neurological disease. The principal cause of
hospitalization was based on the 10th revision of the International Classification of Diseases (STAKES, 1999), purchases of prescription medication were based on the Anatomical
Therapeutic Chemical Classification (Laakelaitos, 1997, 1998),
¨¨
and the right to reimbursement for drug costs was based on the
Finnish disease classification of the Social Insurance Institution
(1998). For more details, see Nihila and colleagues (2007).
¨

MARTIKAINEN ET AL.

S102

Table 1. Distribution (%) of Living Arrangements and Survival at Follow-Up (2002) by Gender, Age, and
Living Arrangement at Baseline (1997)
Men
Age Group

With Partner

Alone

Other

Women

Institution

Dead

Total (n)

With Partner

Alone at baseline
(7,429)
1.2
(6,446)
0.5
(4,331)
0.3
(3,146)
0.1
(1,828)
0.1
(554)
0.0

Alone

Other

Institution

Dead

Total (n)

86.7
79.5
68.2
50.4
31.3
16.1

3.3
3.3
3.6
3.7
3.3
1.8

2.1
4.5
8.0
12.9
15.3
14.3

6.8
12.1
20.0
32.9
50.0
67.9

100
100
100
100
100
100

73.3
88.1

3.4
4.1

5.9
7.1

16.7

100 (86,151)
100 (71,742)

100 (28,166)

65–69
70–74
75–79
80–84
85–89
90þ

3.1
2.2
1.9
1.0
0.5
0.4

70.1
63.0
52.0
39.5
25.0
12.1

3.7
2.9
3.7
2.8
3.0
1.6

2.6
4.1
6.3
8.7
8.9
7.9

20.5
27.9
36.0
48.0
62.6
78.0

100
100
100
100
100
100

Totala
Total of those surviveda

2.2
3.2

58.1
85.0

3.3
4.8

4.8
7.0

31.7

100 (23,734)
100 (16,221)

65–69

82.2

4.0

1.1

0.9

11.8

With partner at baseline
100 (32,005)
78.3

12.8

2.6

1.1

5.2

70–74
75–79
80–84
85–89
90þ

71.1
57.2
37.8
21.5
12.4

5.8
7.3
8.4
7.7
3.5

1.3
1.8
2.1
2.6
0.5

1.9
3.7
6.1
7.8
7.3

19.9
30.0
45.6
60.5
76.3

100
100
100
100
100

(23,818)
(13,447)
(6,351)
(2,158)
(372)

65.4
48.4
30.6
15.8
5.8

18.2
22.3
22.0
16.1
5.8

3.8
4.7
5.4
6.0
4.7

2.8
6.4
10.7
14.4
15.1

9.8
18.2
31.3
47.8
68.6

100
100
100
100
100

Totala
Total of those surviveda

64.9
86.5

5.8
7.8

1.5
2.0

2.8
3.7

25.0

100 (78,151)
100 (58,595)

58.9
69.5

17.3
20.4

3.9
4.6

4.6
5.5

15.3

100 (62,751)
100 (53,159)

0.6
0.8

(18,230)
(21,497)
(20,574)
(15,445)
(8,113)
(2,292)

(19,817)
(10,111)
(3,665)
(906)
(86)

Note: aAge adjusted.

Statistical Methods
We based the descriptive analyses of transitions in living
arrangements on direct age-standardized percentages. We
carried out age adjustment in single-year age groups, with the
combined male and female populations for 1997 being the
standard, and assessed tests of the associations of explanatory
factors with transitions by means of chi-square tests. We used
multinomial logistic regression models with similar age adjustment for the explanatory analyses of the transition determinants.
We present the results of the multinomial logistic regression
models as odds ratios, the first category of each explanatory
variable being the reference group with an odds ratio of 1. We
used STATA (StataCorp, 2003) for all of the calculations.
In evaluating the study aims we adopted the following
modeling strategy. In order to establish the magnitude of the
associations between different socioeconomic indicators and
living arrangement transitions we fitted models that included
each socioeconomic indicator and age (Model 1). In the
following stage we aimed to determine the effects of each
indicator net of the other, and thus we fitted models with all
socioeconomic indicators (Model 2). In the final stage we
assessed the possible mediating effects of health by including
18 dichotomous variables describing chronic conditions (Model
3). We carried out all of the analyses separately for men and
women in order to quantify the differences in the magnitude
and determinants of the living arrangement transitions.

RESULTS

Transitions in Living Arrangements and Mortality by
Gender, Age, and Living Arrangements at Baseline
The proportions of men and women living with a partner at
baseline were 77% and 42%, respectively. These figures
declined with age, particularly among women. In all, 58% of
the men and 73% of the women living alone at baseline

continued to do so at follow-up (see Table 1). Among those
living alone, death was the most common transition, followed
by transition to an institution. Both of these transitions were
more likely among those living alone than among those living
with a partner, particularly at younger ages. Repartnering was
very rare, except in the youngest age groups.
At the end of the follow-up period, about 65% of the men
living with a partner at baseline remained in the same living
arrangement, although this proportion varied strongly with age.
Among these men, the most common transition was to death
(see Table 1). Among younger men, the second most common
transition was to living alone, and among the older men to
institutions. Transitions to other living arrangements (e.g.,
living with children or other adults) were relatively infrequent.
A somewhat different picture emerged among the women:
The most common transition of women aged 65–79 years was
to living alone, whereas among the older women it was to
death. Transitions to institutions were almost twice as common
among the women as among the men. However, women were
less likely to live with a partner and die than men and were
much more likely to live alone or with others at follow-up.

Age-Adjusted Transitions in Living Arrangements by
Socioeconomic Status and Health Conditions
Overall, socioeconomic variables were rather consistently
associated with the transitions (see Tables 2 and 3). Having
a higher education, being in a white-collar occupational social
class, having a higher income, and living in owner-occupied
housing were associated with a lower risk of institutionalization
and death regardless of baseline living arrangement. Moreover,
moving to live with others was associated with a lower social
position. Among those living alone at baseline, the effects of
socioeconomic factors appeared stronger among men than
women, whereas there were few gender differences in these
effects among those living with a partner at baseline.

LIVING ARRANGEMENT TRANSITIONS

S103

Table 2. Age-Adjusted Transitions (%) in Living Arrangements and Survival Between Baseline (1997) and
Follow-Up (2002) Among Men and Women Living Alone at Baseline
Men
Characteristic

With Partner Other Institution Dead

Women
n

p for difference With Partner Other Institution Dead

n

p for difference

Socioeconomic characteristics
Education
Tertiary
Intermediate
Basic

.000
4.9
2.2
1.9

2.9
2.8
3.4

3.7
4.3
5.0

24.3 2,345
28.8 2,621
32.9 18,768

3.3
1.7
1.6
3.8

2.9
3.0
3.6
4.9

4.3
5.2
4.6
4.0

26.6 4,939
33.9 12,650
29.5 4,103
32.5 2,042

4.0
1.9
1.6
1.6

3.2
3.1
2.9
3.8

3.8
4.8
5.1
5.4

26.4
30.1
32.7
35.9

2.4
1.7

3.1
3.6

4.2
6.0

28.3 16,083
38.5 7,651

1.6
1.8
0.0
1.5
2.1
1.5
0.4
2.1
2.1
1.3
2.0
1.4
1.6
2.3
5.6
1.3
2.3
2.1

2.2
2.8
2.0
5.2
3.8
4.1
3.3
3.7
2.6
3.2
3.1
3.1
2.9
5.5
0.0
3.5
3.3
2.8

3.5
5.9
30.8
7.9
7.8
7.7
8.3
7.3
4.3
12.4
4.2
6.7
3.6
6.7
11.3
8.0
5.4
4.6

2.2

3.3

4.8

Social class
White collar
Manual
Farmer
Other

5.2
6.3
5.9

14.1 7,178
15.9 11,417
17.1 67,556

0.7
0.5
0.4
0.9

2.9
3.4
4.2
4.4

5.8
6.1
6.1
6.1

15.5 29,783
17.4 37,830
17.0 12,317
17.4 6,221

0.7
0.5
0.5
0.8

3.2
3.0
3.1
3.8

5.5
5.5
6.0
6.4

14.7
15.5
16.6
18.5

0.6
0.6

3.2
3.7

5.6
6.9

15.2 60,174
20.5 25,977

0.4
0.6
0.1
0.3
0.5
0.5
0.2
0.5
0.7
0.4
0.5
0.2
0.4
0.9
0.2
0.5
0.7
0.6

2.4
3.5
3.9
4.9
3.8
3.7
4.2
3.5
3.2
2.5
3.3
3.4
3.0
3.4
4.1
3.4
3.5
3.3

5.1
7.3
32.5
14.0
10.4
11.7
17.7
9.6
6.0
13.1
5.6
7.6
6.5
6.5
13.5
9.2
7.3
5.8

0.6

3.4

5.9

.000

.000
5,113
4,899
6,733
6,989

Home ownership
Owner
Other

2.9
2.9
3.5

.000

Household disposable income
1st quartile (high)
2nd quartile
3rd quartile
4th quartile (low)

.000
0.7
0.7
0.6

.000
12,944
15,412
24,548
33,247

.000

.000

Presence of chronic conditionsa
Cancer
Diabetes
Dementia
Psychosis
Depressive symptoms
Other mental health disease
Parkinson’s disease
Other neurological disease
Heart disease
Stroke
Chronic asthma
Other respiratory disease
Arthritis
Osteoarthritis
Hip fracture
Accident or violence
Other hospital diagnosis
Other disease
Total

54.3
45.6
51.0
38.0
40.5
45.9
51.4
43.7
38.9
45.7
46.3
51.4
40.0
27.7
46.8
42.5
39.0
34.7

1,264
2,466
137
735
1,983
1,381
460
1,110
7,089
588
1,845
1,165
556
420
165
1,136
6,723
7,680

.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.018
.000
.000
.000
.000

31.7 23,734

36.7
28.5
55.0
25.0
23.8
26.3
31.4
23.9
23.1
30.8
22.8
33.2
25.3
15.3
31.6
25.3
22.7
19.0

3,604
8,787
552
2,758
10,352
5,334
1,549
3,573
24,401
1,451
5,664
2,800
4,117
2,564
983
4,076
24,334
36,789

.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.126
.000
.000
.000
.000

16.7 86,151

Note: adichotomous variables for each condition. Reference category of no disease not shown.

For the men living alone at baseline, the transition to living
with a partner was associated with a higher social status,
whereas the associations were weaker and less consistent
among the women. For those living with a partner at baseline,
however, all socioeconomic measures were consistently inversely associated with the transition to living alone.
As expected, most of the health conditions were associated
with death and entry into institutional care; dementia, mental
health problems, and stroke were particularly strongly associated with both outcomes. The effects of health on mortality
were stronger for the men than for the women, whereas the
opposite was true for entry into institutional care.
The transition to living alone or with others was usually
associated with somewhat better health among the men and
women living with a partner at baseline, whereas for those

living alone at baseline the transition to living with others was
often associated with poorer health.

Multivariate Analyses of Determinants of
Transitions in Living Arrangements
Living alone at baseline. —Because repartnering was relatively rare among those living alone at baseline, our focus in
this section is on transitions to living with others, transitions
into institutional care, and death. The age-adjusted Model 1 of
Table 4 mainly confirms the associations presented in Table 2.
Of the various socioeconomic indicators, having a low income
and not owning a home were most consistently associated with
institutionalization and death. Having a lower educational level
and being in the blue-collar class were associated with

MARTIKAINEN ET AL.

S104

Table 3. Age-Adjusted Transitions (%) in Living Arrangements and Survival Between Baseline (1997) and
Follow-Up (2002) Among Men and Women Living With Partner at Baseline
Men

Women

Alone

Other

Institution

Dead

n

5.4
5.7
6.0

0.9
1.4
1.6

2.5
3.0
2.8

21.18
22.86
26.13

12,024
10,095
56,032

5.7
6.4
4.8
5.7

1.0
1.4
2.3
1.7

2.5
3.2
2.5
2.6

22.57
26.48
24.69
25.22

22,402
34,013
14,773
6,963

5.1
5.8
6.5
6.3

1.5
1.4
1.4
2.0

2.3
2.8
3.1
2.9

22.57
25.08
26.61
26.81

24,760
23,961
17,216
12,214

5.6
7.5

1.5
1.7

2.7
3.5

24.45
29.39

69,333
8,818

3.2
4.9
1.3
4.1
4.7
4.3
3.3
4.4
5.1
3.1
5.0
4.2
4.5
5.3
3.5
4.7
4.9
5.5

1.0
1.3
1.1
1.4
1.3
1.3
1.0
1.2
1.3
1.0
1.2
1.6
1.2
1.5
1.6
1.3
1.1
1.3

2.1
3.5
14.7
7.1
5.7
6.4
7.8
4.4
2.7
6.1
2.5
2.8
2.6
2.9
7.3
4.7
3.2
2.8

50.6
36.9
69.8
36.4
37.9
44.5
46.6
36.8
31.6
42.4
35.6
48.0
33.7
20.0
46.5
32.3
32.4
28.3

3,955
7,800
409
1,042
4,638
2,915
1,429
3,660
23,284
1,933
6,043
2,884
1,972
1,467
289
2,347
19,118
27,121

5.8

Characteristic

1.5

2.8

25.0

p for difference

78,151

Alone

Other

Institution

Dead

n

16.8
17.3
17.4

2.4
3.1
4.1

4.0
4.2
4.8

11.4
14.0
15.8

5,420
9,445
47,886

17.2
18.0
15.7
17.8

2.7
4.0
5.7
3.8

4.5
4.9
4.4
4.7

13.5
16.2
16.6
14.5

22,838
25,697
10,140
4,076

15.3
16.6
19.2
19.1

4.1
3.8
3.6
4.4

4.1
4.8
4.9
4.9

13.3
15.5
16.2
16.0

18,052
19,561
14,623
10,515

16.9
20.6

4.0
3.4

4.4
6.0

14.9
18.1

55,238
7,513

13.2
13.9
1.3
13.8
15.4
14.4
11.0
13.8
16.0
9.9
15.3
13.6
14.9
16.0
11.8
14.7
15.2
16.6

2.3
3.7
2.0
3.5
3.3
3.5
4.4
3.5
3.8
4.2
3.9
2.3
3.8
4.3
2.0
4.0
3.5
3.8

3.6
6.0
32.0
9.7
8.8
9.4
10.5
7.0
4.8
9.0
4.4
6.1
5.0
3.9
9.0
6.6
5.4
4.8

34.5
27.1
50.6
23.5
22.4
24.4
32.3
26.4
21.0
33.9
19.6
31.7
21.6
15.0
30.3
24.5
20.5
17.8

2,499
5,887
271
1,386
5,726
2,467
905
2,328
13,599
803
4,254
1,271
3,033
1,773
368
1,808
13,981
25,515

17.3

3.9

4.6

15.3

p for difference

62,751

Socioeconomic characteristics
Education
Tertiary
Intermediate
Basic

.000

Social class
White collar
Manual
Farmer
Other

.000

Household disposable income
1st quartile (high)
2nd quartile
3rd quartile
4th quartile (low)
Home ownership
Owner
Other

.000

.000

.000

.000

.000

.000

Presence of chronic conditionsa
Cancer
Diabetes
Dementia
Psychosis
Depressive symptoms
Other mental health disease
Parkinson’s disease
Other neurological disease
Heart disease
Stroke
Chronic asthma
Other respiratory disease
Arthritis
Osteoarthritis
Hip fracture
Accident or violence
Other hospital diagnosis
Other disease
Total

.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000

.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.238
.000
.000
.000
.000

a

Note: dichotomous variables for each condition, Reference category of no disease not shown.

institutionalization, but these effects were mostly due to having
lower income and not owning a home (see Model 2): Their
associations with institutionalization and death were often fully
explained or attenuated by about three fourths when we added
income and home ownership to the model. Income and home
ownership had the strongest and most consistent effects on
mortality. Some of the effects of being in the lowest income
bracket and not owning a home on institutionalization and
mortality were mediated by the existence of various chronic
health conditions (Model 3 adjusted for 18 dichotomous
indicators of chronic conditions). In the fully adjusted model,
the effects of socioeconomic status on these transitions
continued to be stronger among men.
However, social class was more strongly associated with the
transition to living with others. At least in the case of farmers,
this could have reflected a tendency to live with children,
a living arrangement that may still be more strongly valued and

made possible given the predominance of detached housing in
the countryside. The adjusted effects of home ownership, and
particularly income and education, on the transition to living
with others were weaker and were little influenced by further
adjustment for chronic conditions (see Model 3).
Multivariate results not shown here confirmed the finding
that the effects of health conditions on the transitions were
strong. Because of relatively strong comorbidity these effects
were attenuated after we adjusted for all chronic conditions
simultaneously. Further adjustment for socioeconomic factors
had little effect on most of these associations.
Living with a partner at baseline. —Model 1 of Table 5 also
mainly confirms the associations presented in Table 3 for those
living with a partner at baseline. When we adjusted all of the
socioeconomic indicators in Model 2 simultaneously, household
income, and home ownership in particular, remained the

LIVING ARRANGEMENT TRANSITIONS

S105

Table 4. Odds Ratios From Multinomial Logistic Regression Models of Living Arrangements and Survival at
Follow-Up (2002) by Gender for Those Living Alone at Baseline (1997)
With Partner vs. Alone
Socioeconomic Characteristic

M1

M2

Other vs. Alone

M3

M1

M2

Institution vs. Alone
M3

M1

M2

M3

Dead vs. Alone
M1

M2

M3

Living arrangement at follow-up for men living alone at baseline (n ¼ 23,734)
Education (tertiary as reference)
Intermediate
Basic

0.48*
0.44*

0.64*
0.66*

0.64*
0.66*

1.02
1.30*

0.96
1.16

0.97
1.18

1.23
1.61*

1.05
1.26

1.06
1.29

1.24*
1.59*

1.06
1.22*

1.06
1.21*

0.95
0.92
1.97*

0.96
0.92
1.97*

1.18
1.33*
1.91*

1.06
1.19
1.67*

1.06
1.18
1.65*

1.41*
1.06
1.11

1.05
0.78*
0.83

1.06
0.78*
0.84

1.48*
1.19*
1.45*

1.17*
0.93
1.14*

1.20*
0.90
1.17*

0.61*
0.53*
0.53*

0.62*
0.53*
0.53*

1.01
0.98
1.42*

0.94
0.85
1.11

0.94
0.85
1.06

1.34*
1.49*
1.68*

1.20
1.25*
1.46*

1.20
1.27*
1.41*

1.21*
1.39*
1.64*

1.05
1.12*
1.33*

1.03
1.12*
1.32*

0.99

0.97

1.50*

1.46*

1.38*

1.80*

1.65*

1.48*

1.76*

1.62*

1.47*

Social class (white collar as reference)
Manual
Farmer
Other

0.58*
0.54*
1.34*

Household income (highest quartile as reference)
2nd quartile
3rd quartile
4th quartile

0.51*
0.43*
0.46*

Home ownership (home owner as reference)
Other

0.87

Living arrangement at follow-up for women living alone at baseline (n ¼ 86,151)
Education (tertiary as reference)
Intermediate
Basic

1.11
1.03

1.23
1.16

1.22
1.14

1.03
1.28*

0.95
1.09

0.95
1.09

1.26*
1.24*

1.16*
1.07

1.17*
1.06

1.20*
1.34*

1.07
1.11*

1.06
1.07

0.73*
0.61*
1.13

0.72*
0.60*
1.13

1.23*
1.59*
1.48*

1.13*
1.40*
1.36*

1.13*
1.39*
1.35*

1.14*
1.06
1.11

1.03
0.93
1.02

1.03
0.91
1.00

1.20*
1.20*
1.20*

1.05*
1.02
1.07

1.05
0.94
1.06

0.78
0.80
1.38

0.78
0.80
1.41*

0.97
1.03
1.35*

0.93
0.91
1.06

0.92
0.91
1.03

1.03
1.17*
1.25*

0.99
1.08
1.16*

0.99
1.07
1.09

1.07*
1.20*
1.38*

1.02
1.07
1.18*

1.00
1.04
1.09*

1.04

1.06

1.31*

1.25*

1.21*

1.39*

1.35*

1.26*

1.53*

1.47*

1.37*

Social class (white collar as reference)
Manual
Farmer
Other

0.81
0.85
1.39

Household income (highest quartile as reference)
2nd quartile
3rd quartile
4th quartile

0.76
0.74*
1.21

Home ownership (home owner as reference)
Other

1.09

2

2

Notes: Model 3 for men alone at baseline: likelihood ratio v (244) ¼ 5,359.44, pseudo R ¼ .1091, p , .000. Model 3 for women alone at baseline: likelihood ratio
v2(244) ¼ 24,245.69, pseudo R2 ¼ .1503, p , .000. M1 ¼ each socioeconomic variable adjusted for age; M2 ¼ adjusted for age and all socioeconomic variables
simultaneously; M3 ¼ M2 þ chronic health conditions (18 dichotomous variables).
*p , .05.

strongest determinants of both entry into institutional care and
mortality. The effects of both education and social class were
more modest. These results were similar to those observed for
men and women living alone at baseline. However, the transition
to living with others was much more strongly associated with
education and social class, and these effects remained strong
after we adjusted for income and home ownership.
The transition to living alone was associated with all of the
socioeconomic indicators, the effects being strongest for
income and home ownership (see Models 1 and 2) for both
men and women. Further adjustment for the presence of chronic
conditions at baseline (Model 3 adjusted for 18 dichotomous
indicators of chronic conditions) did not influence the effects of
socioeconomic indicators on transitions to living alone or with
others, but their effects on institutionalization and mortality
were attenuated by about 10% to 20%, particularly for women.
Multivariate results not shown confirmed the results
presented in Table 3 that the effects of chronic conditions
were strong for institutional entry and mortality. These effects
were attenuated after we adjusted for other chronic conditions,
but adjustment for socioeconomic factors had little effect.

DISCUSSION

Transitions in Living Arrangements
We used a 40% sample (N ¼ 250,787) of Finnish men and
women aged 65 years and older at the end of 1997 and followed
this sample up for transitions in living arrangements and
mortality. At the end of the 5-year follow-up period, about 65%
of the men and 59% of the women living with a partner at
baseline remained in the same living arrangement. Among those
living alone at baseline, the corresponding figures were 58% and
73%. Death was the most common transition in all groups
except for the 65- to 79-year-old women living with a partner at
baseline, for whom transition to living alone was more common.
Moving to an institution was generally the second most common
transition. Transitions to living with others (e.g., children and
other relatives) were less common, with only 3.9% of all women
and 1.5% of all men making this transition in the 5-year period.
The quantum of these transitions is difficult to compare
across countries because of large differences in data sources
(e.g., the definition of the study population), data quality (e.g.,
nonresponse), the definition of living arrangements (e.g.,

MARTIKAINEN ET AL.

S106

Table 5. Odds Ratios From Multinomial Logistic Regression Models of Living Arrangements and Survival at
Follow-Up (2002) by Gender for Those Living With Partner at Baseline (1997)
Alone vs. With Partner
Socioeconomic Characteristic

M1

M2

M3

Other vs. With Partner
M1

M2

Institution vs. With Partner

M3

M1

M2

M3

Dead vs. With Partner
M1

M2

M3

Living arrangement at follow-up for men living with partner at baseline (n ¼ 78,151)
Education (tertiary as reference)
Intermediate
Basic

1.13
1.27*

1.04
1.10

1.04
1.11

1.39*
1.80*

1.19
1.47*

1.19
1.47*

1.33*
1.35*

1.10
1.03

1.09
1.01

1.18*
1.46*

1.06
1.24*

1.01
1.17*

1.04
0.71*
0.90

1.03
0.71*
0.90

1.43*
2.39*
1.79*

1.32*
2.19*
1.65*

1.32*
2.19*
1.64*

1.44*
1.05
1.15

1.24*
0.89
1.02

1.25*
0.87
1.03

1.34*
1.15*
1.23*

1.11*
0.95
1.05

1.13*
0.93*
1.08

1.16*
1.37*
1.44*

1.16*
1.37*
1.44*

0.91
1.02
1.41*

0.76*
0.73*
0.84

0.76*
0.73*
0.84

1.31*
1.49*
1.43*

1.20*
1.36*
1.37*

1.19*
1.38*
1.38*

1.23*
1.39*
1.38*

1.11*
1.22*
1.22*

1.10*
1.23*
1.24*

1.47*

1.47*

1.43*

1.42*

1.40*

1.62*

1.53*

1.45*

1.48*

1.41*

1.33*

Social class (white collar as reference)
Manual
Farmer
Other

1.25*
0.88*
1.06

Household income (highest quartile as reference)
2nd quartile
3rd quartile
4th quartile

1.22*
1.42*
1.40*

Home ownership (home owner as reference)
Other

1.57*

Living arrangement at follow-up for women living with partner at baseline (n ¼ 62,751)
Education (tertiary as reference)
Intermediate
Basic

1.11*
1.23*

0.95
0.97

0.96
0.97

1.53*
2.13*

1.35*
1.81*

1.36*
1.79*

1.11
1.46*

0.96
1.17

1.00
1.18

1.36*
1.70*

1.16*
1.35*

1.14
1.29*

1.04
0.84*
1.01

1.04
0.85*
1.01

1.65*
2.26*
1.48*

1.60*
2.26*
1.46*

1.59*
2.24*
1.46*

1.26*
1.16*
1.23*

1.05
0.97
1.08

1.04
0.94
1.11

1.39*
1.40*
1.21*

1.21*
1.24*
1.11

1.18*
1.17*
1.13*

1.21*
1.50*
1.52*

1.21*
1.50*
1.52*

1.01
1.05
1.27*

0.79*
0.72*
0.77*

0.79*
0.72*
0.76*

1.27*
1.50*
1.52*

1.18*
1.36*
1.34*

1.14
1.31*
1.24*

1.26*
1.43*
1.46*

1.10*
1.18*
1.14*

1.06
1.13*
1.04

1.41*

1.41*

1.10

1.07

1.05

1.73*

1.64*

1.55*

1.63*

1.55*

1.46*

Social class (white collar as reference)
Manual
Farmer
Other

1.21*
1.01
1.14*

Household income (highest quartile as reference)
2nd quartile
3rd quartile
4th quartile

1.23*
1.52*
1.51*

Home ownership (home owner as reference)
Other

1.50*

2

2

Notes: Model 3 for men with partner at baseline: likelihood ratio v (244) ¼ 17,236.05, pseudo R ¼ .1229, p , .000. Model 3 for women with partner at baseline:
likelihood ratio v2(228) ¼ 13,569.58, pseudo R2 ¼ .1033, p , 0.000. M1 ¼ each socioeconomic variable adjusted for age; M2 ¼ adjusted for age and all socioeconomic
variables simultaneously; M3 ¼ M2 þ chronic health conditions (18 dichotomous variables).
*p , .05.

institutional residence), and analytical choices. However, if one
takes these methodological differences into account, the
evidence indicates that the stability in living arrangements in
Finland is similar to that in the United States (Liang et al.,
2005), the United Kingdom (Evandrou et al., 2001), and Japan
(Brown et al., 2002). However, when transitions occur, the
Japanese are more likely to move into coresidence with their
children, a pattern that is fully consistent with more familybased care of elderly people (Brown et al., 2002).
In addition to family care, the provision of and access to
public or private home-based services may also go some way in
explaining national differences in transitions in living arrangements. Elderly people living in countries with comprehensive
home-based services may be able to maintain independent
living longer than those residing in other countries. Overall,
however, evidence of national differences in home-based
services is quite difficult to assess (Kinsella & Velkoff, 2001;
OECD Health Project, 2005; Rostgaard, 2002). In particular,
definitions of home services differ between countries and in the
relevant studies, and the most commonly used measure—the

proportion of elders receiving at least some care—does not
capture the intensity of home services satisfactorily. Furthermore, in many countries the observed increases in home-based
care may reflect a policy change away from institutional care
(OECD Health Project, 2005). However, in some other
countries—as in the Finnish case—home-based care appears
to have declined since the early 1990s, which may be due to
a policy shift from less intensive help (e.g., with cleaning) to
more intensive personal care (e.g., with bathing, daily getting
into and out of bed; Rostgaard, 2002). More research is needed
before researchers can establish the links between home-based
services and national differences in living arrangement
transitions with any certainty.

Effects of Socioeconomic Factors on Transitions in
Living Arrangements
Our specific aim was to study and contrast the effects of
various socioeconomic factors on living arrangement transitions and to assess the independent effects of each factor after
adjusting for the other factors and health status. By doing so we

LIVING ARRANGEMENT TRANSITIONS

hoped to be able to more clearly identify the social pathways by
which socioeconomic factors influence the transitions.
Our results showed that income and home ownership were
independently associated with institutionalization and mortality
among those living alone or with a partner at baseline. The
effects of education and social class were weaker, particularly
after mutual adjustment for all socioeconomic indicators.
Previous evidence corroborates our findings that the more
material aspects of social position are strongly associated with
health and mortality at older ages (Grundy & Holt, 2001;
Martelin, 1994; Martikainen, Makela, Koskinen, & Valkonen,
¨ ¨
2001) as well as in middle age (Martikainen, Adda, Ferrie,
Davey Smith, & Marmot, 2003; Martikainen et al., 2001).
In particular, the independent association between home
ownership—a measure of social position earned over decades
and possibly across generations—and institutionalization and
mortality indicates that the causes of these associations are
likely to be related to a set of both early and current material
advantages. In the case of entry into institutional care, these
effects of higher income may also be associated with better
quality housing and the ability to purchase home care. However,
as we had no direct information on home care and informal
support, particularly given by living children (Frankenberg,
Chan, & Ofstedal, 2002; Grundy, 2000; Wolf & Soldo, 1988),
it remains difficult to accurately estimate to what extent formal
and informal support mediate the effects of these socioeconomic
factors.
Although death and institutionalization are in many cases
preceded by a period of ill health, the higher risk of these
transitions in the lower socioeconomic groups was only partially (about 20% attenuation) explained by the chronic health
conditions that we measured. There are at least two reasons
why these independent effects remained. The first of these
relates to our health measurement. Full and comprehensive
health measurement is probably impossible in a populationbased study (for a more detailed discussion, see ‘‘Methodological
Issues’’), and thus the unexplained effects of socioeconomic
indicators may be due to unmeasured health differences between
the socioeconomic groups. For example, we underestimated or
lacked information on conditions that are associated with less
serious functional difficulties, but we believe that the way in
which we measured health covered serious life-threatening
chronic diseases very well. Second, over the 5-year follow-up
period health problems that were not present at baseline were
likely to emerge and to contribute to the transitions that we were
studying. The implication of this is that socioeconomic differences in transitions in old age are not necessarily driven by
differences in long periods of disability and frailty but are often
more acute (e.g., accidents at home).
The transition to living with others was associated with social
class in particular. Among those living with a partner at
baseline, the transition to living alone was also associated with
all socioeconomic factors, but most strongly with income and
home ownership. The effects of socioeconomic factors on
transitions between private households were, to a large extent,
independent of chronic conditions, although their effects on
institutionalization and mortality were somewhat attenuated.
Overall, it appears likely that the effects of socioeconomic
factors on transitions in living arrangements can be only
partially explained by socioeconomic differences in chronic

S107

health conditions. Consequently, more direct material factors—
such as having the money to maintain a single household and
having the ability to purchase home care, and the values and
preferences of elderly persons regarding their living arrangements—could also play an important role. This latter
explanation may be particularly potent for understanding the
effects of education and social class net of income and home
ownership: Those in higher socioeconomic positions may value
independent living more than those in lower positions.
Although attitudes toward elder care are more family oriented
in southern European countries than in northern Europe
(Tomassini et al., 2004), evidence concerning socioeconomic
differences in values, attitudes, and preferences in living
arrangements among elders is scarce. Furthermore, the
applicability of findings between countries is complicated
because national differences in the perceived quality, availability, and cost of long-term care may influence responses to
attitudinal questions about living arrangement preferences.

Effects of Chronic Conditions on Transitions in
Living Arrangements
As expected, several chronic health conditions—particularly
dementia, mental health problems, stroke, and hip fracture—
were strongly associated with transitions to institutions and
mortality. In the case of transitions to institutions, these findings
reflect the strong impact that these chronic conditions have, with
cognitive and physical functional difficulties that make independent living impossible. Because of comorbidity, these effects
were attenuated after we adjusted for other health indicators.
Health also determined transitions between private households, but these effects were modest compared with the effects
on the transition to institutions and on mortality. In particular,
among those living with a partner at baseline only dementia
was strongly associated with the transition to living with others.
Consistent with the results of Liang and colleagues (2005) on
U.S. adults, somewhat larger effects of chronic conditions were
evident among those living alone at baseline. Our findings
provide support for the idea that the effects of severely
disabling chronic conditions are buffered by the presence of,
and support provided by, a coresident partner.

Methodological Issues
One problem with these data is that we had no direct
information on living children (Frankenberg et al., 2002;
Grundy, 2000; Wolf & Soldo, 1988) and limited information on
complex households. A more specific problem was that one of
our living arrangement categories at follow-up—living with
others—included those who were living with their children as
well as those who were living with other relatives or other
adults. Confirmatory analyses on another data set of these
participants showed that of the households classified as
‘‘other,’’ 62% were two-person households. In about 56% of
these households the elder was living with a child who had no
family of his or her own, and in the rest of the households the
coresident was usually of approximately the same age as the
index person. A further 19% of ‘‘other’’ households were threeperson households. In about 30% of these households the elder
was living with children with no family of their own, whereas
a further 50% were living with a younger family (one of the
members of this younger family was likely to be a child of the

S108

MARTIKAINEN ET AL.

elderly person). This proportion was even higher in households
with more than three members. We thus feel that we were
justified in considering these ‘‘other’’ households as mainly
consisting of close relatives.
Our measure of living arrangement transition may have been
problematic because it only considered states at baseline and at
the end of the follow-up; thus, all transitions occurring between
these two remained unobserved. Analyses of first transitions
would have overcome this problem. However, although our
data allowed us to date institutional entry and death on a daily
basis, they allowed us to assess transitions between private
households only on an annual basis. Although such data did not
allow us to carry out full analyses of first transitions, we did
carry out confirmatory analyses. As might have been expected,
these results showed that in the case of first transitions, those
between private households and those to institutional care
became more common, and death correspondingly became less
common (in comparison with transitions measured only between
baseline and the end of the follow-up). However, the effects
of socioeconomic factors on first transitions were very similar.
We based our data on disease prevalence on registration data
on hospital and medication use. In general, the prevalence rates
we found were quite close to those derived from populationbased clinical examinations and other sources (e.g., Aromaa &
Koskinen, 2004). The two most notable exceptions were for
dementia and osteoarthritis, both of which were underestimated
in our study. Information on both conditions was based on
hospital-use data and was thus likely to represent the most
severe cases. Similarly, our study could not capture cancers that
were not being actively treated or that had been cured. The
chronic disease data were thus likely to underestimate physical
and cognitive functional difficulties in the population (for
a more thorough discussion, see Nihtila et al., 2007).
¨
Large population-based datasets linking different administrative registers carry several major advantages. In the analyses
of the association between health and living arrangements it is
often uncertain which of the two factors precedes the other.
Using longitudinal registration data with detailed information
on transitions in living arrangements and various chronic health
conditions and socioeconomic indicators at baseline allowed us
to assess these effects reliably. Furthermore, the data did not
suffer from loss to follow-up, missing values, misreporting, or
lack of power, which may be serious problems with surveybased data.

Conclusions
We have shown that socioeconomic conditions and chronic
health problems are important determinants of transitions in
living arrangements and mortality among Finnish men and
women aged 65 years and older. Health conditions—particularly
those that can be associated with both physical and mental
functional difficulties—are, of course, major determinants of
institutionalization and death, but they are also associated with
transitions in living arrangements between private households.
Whereas material socioeconomic indicators are more closely
associated with transitions to institutions and mortality, education, and particularly social class, seem to be major determinants of transitions to living with others, possibly reflecting
the values and preferences of elders regarding their living
arrangement choices. Among those living with a partner at

baseline, the transition to living alone is associated with all of
the socioeconomic indicators.
Our results show that the associations between different
socioeconomic factors and living arrangement transition are
only partly mediated by differences in chronic health conditions.
Although socioeconomic indicators reflecting the material
conditions of life appear to dominate over other measures of
socioeconomic status as determinants of elder transitions,
variations in these associations imply that different transitions
are determined by different social pathways. Because for most
individuals home ownership more than any other measure of
socioeconomic position reflects wealth accumulated over the
life course, it is likely that the causes of the associations with
living arrangement transition are also related to the accumulation of early and current material benefits and, for example, to
the security of returning back to owner-occupied housing after
extended stays in hospital or long-term care. However, to the
extent that the socioeconomic circumstances associated with
household income—and direct material factors, such as the
ability to purchase home care and better quality housing—are
modifiable, these results may provide opportunities for reducing
the risk of premature institutionalization and death in the elderly
population.
ACKNOWLEDGMENTS
We are grateful to the National Research and Development Centre for
Welfare and Health (STAKES), the Social Insurance Institution, and
Statistics Finland (Permission TK 53-576-04 and TK 53-499-05) for
making the data available to us. This study was supported by the Academy
of Finland (Grants 210752 and 205631). The research was part of the
European Union-funded project Major Ageing and Gender Issues in
Europe.
P. Martikainen planned the study, supervised the data analysis, and
wrote the paper. E. Nihtila helped plan the study and analyses and revised
¨
the paper. H. Moustgaard performed the statistical analyses and contributed
to the writing of the paper.
CORRESPONDENCE
Address correspondence to Pekka Martikainen, Helsinki Collegium for
Advanced Studies & Population Research Unit, Department of Sociology,
University of Helsinki, P.O. Box 18, FIN-00014. E-mail: pekka.martikainen@
helsinki.fi
REFERENCES
Aguero-Torres, H., von Strauss, E., Viitanen, M., Winblad, B., &
Fratiglioni, L. (2001). Institutionalization in the elderly: The role of
chronic diseases and dementia. Cross-sectional and longitudinal data
from a population-based study. Journal of Clinical Epidemiology, 54,
795–801.
Aromaa, A., & Koskinen, S. (2004). Health and functional capacity in
Finland: Baseline results of the Health 2000 health examination survey
(No. B12). Helsinki, Finland: National Public Health Institute.
Bartley, M., Sacker, A., Firth, D., & Fitzpatrick, R. (1999). Understanding
social variation in cardiovascular risk factors in women and men: The
advantage of theoretically based measures. Social Science & Medicine,
49, 831–845.
Branch, L. G., & Jette, A. M. (1982). A prospective study of long-term care
institutionalization among the aged. American Journal of Public Health,
72, 1373–1379.
Breeze, E., Sloggett, A., & Fletcher, A. (1999). Socioeconomic and
demographic predictors of mortality and institutional residence among
middle aged and older people: Results from the Longitudinal Study.
Journal of Epidemiology and Community Health, 53, 765–774.
Brown, J. W., Liang, J., Krause, N., Akiyama, H., Sugisawa, H., & Fukaya,
T. (2002). Transitions in living arrangements among elders in Japan:

LIVING ARRANGEMENT TRANSITIONS

Does health make a difference? Journal of Gerontology: Social
Sciences, 57B, S209–S220.
Chappell, N. L., Dlitt, B. H., Hollander, M. J., Miller, J. A., & McWilliam,
C. (2004). Comparative costs of home care and residential care. The
Gerontologist, 44, 389–400.
Davis, M. A., Moritz, D. J., Neuhaus, J. M., Barclay, J. D., & Gee, L.
(1997). Living arrangements, changes in living arrangements, and
survival among community dwelling older adults. American Journal of
Public Health, 87, 371–377.
De Jong Gierveld, J., & Van Tilburg, T. (1999). Living arrangements of
older adults in the Netherlands and Italy: Coresidence values and
behaviour and their consequences for loneliness. Journal of CrossCultural Gerontology, 14, 1–24.
Evandrou, M., Falkingham, J., Rake, K., Scott, A., and the ESRC Research
Group Simulating Social Policy for an Ageing Society. (2001). The
dynamics of living arrangements in later life: Evidence from the British
Household Panel Survey. Population Trends, 105, 37–44.
Frankenberg, E., Chan, A., & Ofstedal, M. B. (2002). Stability and change
in living arrangements in Indonesia, Singapore, and Taiwan, 1993–99.
Population Studies, 56, 201–213.
Glaser, K., Grundy, E., & Lynch, K. (2003). Transitions to supported
environments in England and Wales among elderly widowed and
divorced women: The changing balance between co-residence with
family and institutional care. Journal of Women & Aging, 15(2/3), 107–
126; discussion 185–187.
Greene, V. L., Lovely, M. E., Miller, M. D., & Ondrich, J. I. (1995).
Reducing nursing home use through community long-term care: An
optimization analysis. Journal of Gerontology: Social Sciences, 50B,
S259–S268.
Grundy, E. (2000). Co-residence of mid-life children with their elderly
parents in England and Wales: Changes between 1981 and 1991.
Population Studies, 54, 193–206.
Grundy, E. (2001). Living arrangements and the health of older persons in
developed countries. Population Bulletin of the United Nations, 42/
43(Special issue), 311–329.
Grundy, E., & Holt, G. (2001). The socioeconomic status of older adults:
How should we measure it in studies of health inequalities? Journal of
Epidemiology and Community Health, 55, 895–904.
Grundy, E., & Jitlal, M. (2007). Socio-demographic variations in moves to
institutional care 1991–2001: A record linkage study from England and
Wales. Age and Ageing, 36, 424–430.
Hays, J. C. (2002). Living arrangements and health status in later life: A
review of recent literature. Public Health Nursing, 19(2), 136–151.
Hays, J. C., Pieper, C. F., & Purser, J. L. (2003). Competing risk of
household expansion or institutionalization in late life. Journal of
Gerontology: Social Sciences, 58B, S11–S20.
Himes, C. L., Wagner, G. G., Wolf, D. A., Aykan, H., & Dougherty, D. D.
(2000). Nursing home entry in Germany and the United States. Journal
of Cross-Cultural Gerontology, 15, 99–118.
Kemper, P. (1992). The use of formal and informal home care by the
disabled elderly. Health Services Research, 27, 421–451.
Kinsella, K., & Velkoff, V. A. (2001). An aging world: 2001 (Series P95/
01-1). Washington, DC: U.S. Government Printing Office.
Laakelaitos. (1997). Laakkeiden luokitus (ATC) ja maaritellyt vuorokau¨¨
¨¨
¨¨
siannokset (DDD) [Classification of medicines and defined daily
portions]. Helsinki, Finland: National Agency for Medicines.
Laakelaitos. (1998). Laakkeiden luokitus (ATC) ja maaritellyt vuorokau¨¨
¨¨
¨¨
siannokset (DDD) [Classification of medicines and defined daily
portions]. Helsinki, Finland: National Agency for Medicines.
Lafreniere, S. A., Carriere, Y., Martel, L., & Belanger, A. (2003).
Dependent seniors at home—formal and informal help. Health Reports,
14(4), 31–40.
Lahelma, E., Martikainen, P., Laaksonen, M., & Aittomaki, A. (2004).
Pathways between socioeconomic determinants of health. Journal of
Epidemiology and Community Health, 58, 327–332.
Lakdawalla, D. N., & Schoeni, R. F. (2003). Is nursing home demand
affected by the decline in age difference between spouses? Demographic Research, 8, 279–304.
Liang, J., Brown, J. W., Krause, N. M., Ofstedal, M. B., & Bennett, J.
(2005). Health and living arrangements among older Americans: Does
marriage matter? Journal of Aging and Health, 17, 305–335.
Martelin, T. (1994). Mortality by indicators of socioeconomic status among
the Finnish elderly. Social Science & Medicine, 38, 1257–1278.

S109

Martikainen, P., Adda, J., Ferrie, J. E., Davey Smith, G., & Marmot, M.
(2003). Effects of income and wealth on GHQ depression and poor self
rated health in white collar women and men in the Whitehall II Study.
Journal of Epidemiology and Community Health, 57, 718–723.
Martikainen, P., Makela, P., Koskinen, S., & Valkonen, T. (2001). Income
¨ ¨
differences in mortality: A register-based follow-up study of three
million men and women. International Journal of Epidemiology, 30,
1397–1405.
Mutchler, J. E., & Burr, J. A. (1991). A longitudinal analysis of household
and nonhousehold living arrangements in later life. Demography, 28,
375–390.
National Research and Development Centre for Welfare and Health
(STAKES). (1999). Second edition of Finnish version of the
International Statistical Classification of Diseases and Related Health
Problems (10th rev., Vol. 1). Helsinki, Finland: Author.
Nihtila, E., & Martikainen, P. (2007a). Household income and other socio¨
economic determinants of long-term institutional care among older
adults in Finland. Population Studies, 61, 299–314.
Nihtila, E., & Martikainen, P. (2007b). Why older people living with
¨
a spouse are less likely to be institutionalised: The role of socioeconomic factors and health characteristics. Scandinavian Journal of
Public Health, 36(1), 35–43.
Nihtila, E. K., Martikainen, P. T., Koskinen, S. V., Reunanen, A. R., Noro,
¨
A. M., & Hakkinen, U. T. (2007). Chronic conditions and the risk of
long-term institutionalization among older people. European Journal of
Public Health, 18(1), 77–84.
Oakes, J. M., & Rossi, P. H. (2003). The measurement of SES in health
research: Current practice and steps toward a new approach. Social
Science & Medicine, 56, 769–784.
Organisation for Economic Co-operation and Development. (1982). The
OECD list of social indicators. Paris: Author.
Organisation for Economic Co-operation and Development Health Project.
(2005). Long-term care for older people. Paris: Organisation for
Economic Co-operation and Development.
Rostgaard, T. (2002). Caring for children and older people in Europe: A
comparison of European policies and practice. Policy Studies, 23(1),
51–68.
Sarma, S., & Simpson, W. (2007). A panel multinomial logit analysis of
elderly living arrangements: Evidence from Aging in Manitoba
longitudinal data, Canada. Social Science & Medicine, 65, 2539–2552.
Shapiro, E., & Tate, R. (1988). Who is really at risk of institutionalization?
The Gerontologist, 28, 237–245.
Social Insurance Institution. (1998). Statistical yearbook of the Social
Insurance Institution, Finland, 1997 (No. T1:33). Helsinki, Finland:
Author.
StataCorp. (2003). Stata statistical software: Release 8.0. College Station,
TX: Author.
Steinbach, U. (1992). Social networks, institutionalization, and mortality
among elderly people in the United States. Journal of Gerontology:
Social Sciences, 47, S183–S190.
Tilastokeskus. (1994). Vaestolaskenta 1990:9. Vuoden 1990 vaestolas¨ ¨
¨ ¨
kennan luotettavuustutkimus [Reliability study on the 1990 census].
Helsinki, Finland: Statistics Finland.
Tomassini, C., Glaser, K., Wolf, D. A., Broese van Groenou, M. I., & Grundy,
E. (2004). Living arrangements among older people: An overview
of trends in Europe and the USA. Population Trends, 115, 24–34.
Tomiak, M., Berthelot, J. M., Guimond, E., & Mustard, C. A. (2000).
Factors associated with nursing-home entry for elders in Manitoba,
Canada. Journal of Gerontology: Medical Sciences, 55A, M279–M287.
United Nations Population Division. (2005). Living arrangements of older
persons around the world (No. ST/ESA/SER.A/240). New York:
United Nations.
Wilmoth, J. M. (1998). Living arrangement transitions among America’s
older adults. The Gerontologist, 38, 434–444.
Wolf, D. A., & Soldo, B. J. (1988). Household composition choices of
older unmarried women. Demography, 25, 387–403.
Wolinsky, F. D., Callahan, C. M., Fitzgerald, J. F., & Johnson, R. J. (1992).
The risk of nursing home placement and subsequent death among older
adults. Journal of Gerontology: Social Sciences, 47, S173–S182.
Received July 24, 2007
Accepted January 10, 2008
Decision Editor: Kenneth F. Ferraro, PhD