Residential Preferences of the Elderly Population: Age, Class, and Geographical Context

Residential Preferences of the Elderly Population:
Age, Class, and Geographical Context
Eva Andersson
Dept of Human Geography, Stockholm University, 106 91 Stockholm, Sweden
[email protected]
Marianne Abramsson
NISAL, Dept of Social and Welfare Studies, Linköping University, 601 74 Norrköping, Sweden
[email protected]
Bo Malmberg
Dept of Human Geography, Stockholm University, 106 91 Stockholm, Sweden
[email protected]
Abstract
In this paper we analyzed how residential preferences in the 55+ population was related to age, socio-economic
background, and geographical context. The share of the 55+ is expected to increase substantially in the coming
decades. With increasing age, the risk of disability increase. Analyzing the residential preferences and residential
strategies of the 55+ population will increase our understanding of how population ageing will affect the housing
market and the extent to which there is a risk for a mis-match between the housing needs and wants of an ageing
population and the ability to find adequate housing. The theoretical perspective used is based on the assumption that
residential preferences are strongly linked to age, socio-economic status and geographical context. The importance
of age will show in changes in health status, physical fitness, and cognitive ability. Socio-economic status changes
like retirement, influences the values of individuals and the opportunities that are available to them. Geographical
context is also assumed to play an important role. Residential preferences often influence where people live (sorting),
and valuations are influenced or re-enforced by the social norms that prevail in the area where one resides
(contextual effects). We used an empirical survey, SHIELD, directed to a stratified sample of the 55+ population in
Sweden with questions about residential status, residential preferences, and residential plans. With respect to
geographical context, our approach was based on a methodological innovation to circumvent potential pitfalls
caused by the modifiable areal unit problem. First results show that decreasing with age are preferences for
closeness to forest and land, that the dwelling has a garden, is owned, that the family can come and stay, that there
are room for social events, and that hobbies can be practiced. Preferences increasing with age are that the dwelling
is located in an area where I feel at home, is designed for independence, and for disability, has an elevator, and that
it is one-storied. Importantly, elderly 55+ is not a homogenous group concerning housing preferences. A challenging
result is however that class and income are not determinants for housing preferences in old age in the same way as
age and geographical context. Finally, geographical context is of importance for elderly’s’ housing preferences, e.g.
garden is preferred in middle class area types but not in elite and marginal area types.
Keywords: elderly, housing preferences, contextual effects, Sweden
WORKSHOP: WS-04: Housing & Living Conditions of Ageing Populations
Residential Preferences of the Elderly Population: Age, Class and Geographical Context
Introduction
In this paper we analyze how residential preferences in the 55+ population are related to age,
socio-economic background, and geographical context. The residential preferences of this
population group are interesting for several reasons. On the one hand, the population share of the
55+ is expected to increase substantially in the coming decades due to increasing life expectancy
and the ageing of the baby-boom cohorts. On the other hand, with increasing age, the risk of
disabilities increases and this will make it difficult for individuals to maintain their quality of life
if their dwellings are not well-adapted for disability. This implies that the need for housing
market adaptation will increase in the coming years. Analyzing the residential preferences and
residential strategies of the 55+ population will increase our understanding of how the challenge
of population ageing will affect the housing market and the extent to which there is a risk for a
mis-match between the housing needs and preferences of the ageing population and their ability
to find adequate housing.
To further add to the research on preferences and supply of housing for elderly, which is mainly
based on qualitative research, we will analyze the preferences linked to age, socio-economic
status and geographical context. Are there differences in preferences between different age
groups supporting the debate on limiting specific housing for certain ages? And most importantly
for different structures of local housing markets; are there differences in preferences between
municipality types and geographical contexts?
The theoretical approach we will use is constructed around a life-course perspective with the
assumption that residential preferences are strongly linked to 1) age, 2) socio-economic status
and 3) geographical context. Firstly the importance of age is based on the fact that people above
55 are experiencing relatively significant changes in their health status, physical fitness, cognitive
ability, and social position (Huber & Skidmore 2003). There is a discussion whether older people
today will be pro-active or not in changing housing to accommodate to these changes and prepare
for old age (Pope and Kang 2010). Socio-economic status, in turn, influences both what values
individuals hold and the opportunities that are available to them. There is a development towards
specific housing for specific groups where housing companies and developers try to find ‘good
costumers’ for their business. Finally, geographical context can be assumed to play an important
role because values are influenced or re-enforced by the social norms that prevail in the area
where one resides. On an even larger scale there is evidence that older people move according to
2
the housing stock on the local housing market (Abramsson & Andersson forthcoming). Therefore
it is of great importance to compare the preferences of elderly and the housing market’s supply of
housing.
Empirically, our analysis of the influence of age and socio-economic status on residential
preferences followed a standard approach. We have used an empirical survey (SHIELD) directed
to a stratified sample of the 55+ population in Sweden, with questions about residential status,
residential preferences, and residential plans, and these answers will be related to the respondents’
age and socio-economic status (compare Ytrehus 2005). With respect to geographical context,
though, our approach will be based on a different methodological design to circumvent potential
pitfalls caused by the modifiable areal unit problem (Openshaw 1984). Essentially this approach
consists of classifying residential areas using multi-scalar context measures that have been
computed for individualized neighborhoods with fixed population size.
For a survey like SHIELD there can be several reasons for finding effects from the above
mentioned age, socio-economic status and geographical context. As such we are in this study
particularly interested in the geographical effects on preferences, since such earlier research for
elderly is largely lacking. Therefore we measure the preferences as formed locally for the main
part of this paper. We do however admit that there are, like in most contextual effects research,
criticism to be made to this declaration. A second explanation for preferences that look like they
are formed in a geographical context, is sorting into locations. That is, preferences govern
mobility decisions. Like most surveys we also anticipate the cognitive dissonance to be at work,
that is, people adjust to where they live and justify their choices as they are already made
(Niedomysl &Malmberg 2009).
With the proliferation of neighborhood-effect studies during the last decade, the question of how
to correctly measure context has become a central concern (Kawachi and Berkman 2003;
Cummins et al. 2007; Spielman and Yoo 2009; Kwan 2012; Sampson 2012; Matthews and Yang
2013). As stated above this paper applies an approach to contextual measurement that uses
individually defined, scalable neighborhoods (Östh, Malmberg, and Andersson forthcoming
2014), and factor analysis to produce a nuanced and composite description of geographical
contexts.
Although contextual-effect studies typically focus on economic, behavioral, or health outcomes,
there are also a number of studies that analyze context effects on attitudes. Studies have analyzed
how context influences attitudes towards health or other social and ethnic groups, civic values,
and political preferences (Stein, Post, and Rinden 2000; Macallister et al. 2001; Cummins et al.
2005; Galster and Santiago 2006; Pettigrew and Tropp 2006; Pettigrew et al. 2007; Bowyer
2009). It is often acknowledged only in passing, however, that the findings of these studies can
be strongly influenced by the way geographical context is defined (Kwan 2012). This has been
3
clearly demonstrated by Spielman and Yoo (2009) who show that a mis-specification of
geographical context can lead to underestimation of the effects.
To address the combined problem of different fields of influence and the possibility that the
relevant geographical context varies between fields of influence, this paper uses factors extracted
from a broad set of contextual variables computed for individually defined, scalable
neighborhoods to estimate contextual influences on older people’s housing preferences and
choice.
A starting point for our approach to contextual measurement is the introduction, in the early
2000s, of bespoke neighborhoods that measure contexts using statistical aggregates for the 100,
500, 2500, or 10,000 nearest neighbors of each individual (Macallister et al. 2001; Bolster et al.
2007). In epidemiological research the same approach, applied to mental health data, has been
called “spatially adaptive filtering” (Chaix et al. 2005). Later, this method has been applied to
segregation measurement (Malmberg, Andersson, and Östh 2011). In addition to being
individually defined, using customized neighborhoods implies that context measures are assigned
an explicit scale based on the number of nearest neighbors included in the aggregate - alternative
ways of constructing scalable neighborhoods are discussed by Reardon (2008) and Spielman and
Logan, (2013). Scalability, however, also creates a problem since each scale level will generate a
distinct contextual value for every socio-economic indicator that is included in the analysis. Thus,
the number of contextual values assigned to each location can become very large.
To handle the number of contextual variables that results from allowing neighborhood scale to
vary we employ factor analysis. Contextual values for different neighborhood scales will be
correlated and factor analysis can be used to compress the original contextual variables to a
smaller set of orthogonal factors that will summarize the spatial structure of the socio-economic
environment. Factor analysis is a well-known method for producing summary measures of the
urban environment but, to our knowledge, it has not been previously applied to a set of contextual
values that are aggregates for differently scaled, bespoke neighborhoods. The advantage of our
approach is that the resulting factors can be seen as representing not only different types of socioeconomic influence but also the geographical structure of the socio-economic environment. We
would, therefore, argue that our approach presents one possible way to address what Kwan
(2012) has called the problem of uncertain geographic context. The idea is that if factors
extracted from scaled contextual variables give a good representation of the geographical context,
then the inclusion of these factors in a regression analysis will generate estimates that indicate
what contextual dimensions are most relevant for the outcome under study.
Following this introduction is a section on housing choice and preferences among older people.
Next data and methods are presented together with the approach of multi-scalar measures of
4
composite contexts. Then there’s the results section on what influences housing preferences and
regression estimates. Finally there is a concluding discussion.
Housing choice and preferences among older people
In general the housing choice of older people reveals the preference of remaining where they are,
as the major proportion of older people are stayers (Abramsson and Andersson 2012; Abramsson
et al. 2012; Ytrehus 2012). Ownership is the most common type of tenure in Sweden, in
particular among older people, but other tenures are not considered subordinate (Andersson et al.
2007) but rather a good housing choice depending on life-stage and life style. Rental housing is
not generally stigmatized and the rights of tenants are comparatively strong as is security of
tenure. As a result, the different types of tenures are attractive to all groups (Andersson 2008).
Among older Canadians, Ostrovsky found a higher transition rate, from single-family housing to
apartments, than that for moves in the reverse direction. This indicates a shifting preference
towards apartments in old age, although limited in scope (Ostrovsky, 2004). Elderly’s satisfaction
with apartment housing was shown by James (2008). Swedish studies similarly indicate a
growing interest among the elderly in moving to more comfortable housing involving less
maintenance (Abramsson et al., 2008). Such moves can be seen also in Norway but only
evidently after the age of 80 (Andersen 2005).
For Norway, Ytrehus (2005) on the other hand, show that most older people express a wish to
remain in a current large dwelling in order to have room for their own activities and family life.
They were prepared to live less comfortably in order to maintain the space as they argued that
they might be unhappier in a smaller dwelling with too little room. A majority preferred to own
their dwelling also in future (Ytrehus 2005). The youngest age groups, in their 50s and 60s,
anticipate that they will have to live in a smaller dwelling in future.
Life course events, such as retirement, the loss of a partner and declining health, influence
mobility rates as well as the destination of moves (Litwak & Longino 1987). Upon retirement the
reasons for moving are most often different to moves conducted later on in life or in relation to
declining health. Among the latter, smaller dwellings are preferred, in rental tenure that are more
easily maintained these dwellings might also be preferred in relation to the loss of a partner
(Abramsson et al 2012; Litwak & Longino 1987; Chevan 1995). Residential mobility among the
young old, on the other hand, can be expected to result from a preference to change housing area,
housing type or housing tenure. In Norway a move to an apartment is more commonly considered
by owner occupiers with higher incomes compared to low income owners (Ytrehus & Fyhn
2006). The former group is more likely to be able to afford a better quality apartment.
5
Recent research show increased residential mobility rates of the 1940s age cohort in Sweden
when compared to older age groups at the same age (Andersson and Abramsson 2012) and may
indicate a change in residential mobility behaviour in which mobility in old age becomes more
frequent. Increased longevity may result in changing plans for life in retirement as more years are
spent in the third age, a time in life that follows family up-bringing and working life while the
individual remains in good health. These years can be planned for new social and recreational
activities (Laslett 1996; Warnes 1992, p 181).
Most moves among the population in general are local moves, most often the same municipality
is preferred and frequently also the same housing area (Abramsson et al. 2012; Fransson and
Borgegård 2002; Ytrehus 2007) and this is particularly true among older movers. Local moves
are most often housing related, and a result of a desire to change housing, such as housing type,
tenure, size or environment (Myers 1990). When moving within a local housing market the
individual’s social networks are not lost.
In a Swedish study mobility rates among older people (65+) in different types of municipalities
were shown to be largely similar with the exception of the metropolitan cities where residential
mobility levels are slightly lower and in the suburbs of major cities where they are higher than in
other types of municipalities (Abramsson et al 2012). The choice of tenure when moving from
owner occupation showed some interesting differences. In the more urbanized municipalities a
larger proportion of older people moved to tenant cooperative apartments whereas in
municipalities with a rural character moves into rental tenure was more frequent (Abramsson et al
2012). This is assumedly a result of the housing market supply (Abramsson & Andersson,
forthcoming) but may not necessarily be a result of the wish of the older individuals themselves.
Data and method
To estimate contextual effects on housing preferences we employ a logistic regression approach
using survey answers as dependent variables, with individual background variables and
contextual variables as explanatory variables.
In SHIELD, Survey of Housing Intentions among the ELDerly in Sweden survey data on housing
preferences were collected in the spring of 2013. The survey was sent out to 4000 individuals
aged 55 years and older. Previous Swedish studies have mostly focused on the young old but here
the oldest respondent was 103. The survey was stratified on age and on municipality type for the
purpose of analyzing age groups as well as geographical differences in attitudes among older
people in Sweden. The response rate was 60.7 per cent or 2400 respondents.
6
To the survey data register data from Statistics Sweden was added. The register data consists of
information on the respondent’s sex, year of birth, country of birth, year of immigration,
educational level, household income, disposable income, municipality and local SAMS-area1.
Table 1 below shows the slightly modified variables used in this study.
Table 1. Composition of respondents.
Respondents' background n= 2400 Sex Share respondents in SHIELD Share within variable Women 1326 55.3 59.9 Men Age, 10-­‐years 1074 44.8 61.8 55-­‐64 564 23.5 56.5 65-­‐74 685 28.5 68.4 75-­‐84 628 26.2 63.9 85 and older 523 21.8 53.8 1. Metropolitan cities 403 16.8 2. Suburbs of metropolitan cities 410 17.1 3. Major cities 415 17.3 4. Suburbs of major cities 426 17.8 5. Industrial and commuting 107 16.0 6. Low density and tourism 141 15.0 2143 89.3 61.8 257 10.7 52.6 853 Lower secondary school 35 52.6 Secondary school 956 40 63.6 University degree 591 25 71.2 2396 Municipality type Country of birth Swedish born Non Swedish born Educational level Income Disposable percentiles income Marital status 1
per person, Sweden is divided into 9000 areas of statistics, so called SAMS, or Small Area Market Statistics.
7
Married 1269 52.9 66.8 Unmarried 241 10.0 51.7 Divorced 399 16.6 58.2 Widow/Widower 491 20.5 54.5 The classification of municipalities, made by the Swedish Association of Local Authorities and
Regions (SALAR) label the Swedish municipalities according to type of municipality, that is 1)
metropolitan cities; 2) suburbs of metropolitan cities; 3) major cities; 4) suburbs of major cities;
5) commuter municipalities; 6) tourism municipalities; 7) goods producing municipalities, 8)
rural municipalities, 9) municipalities in densely populated regions and 10) municipalities in
sparsely populated regions. In this study we joined commuter, goods producing and densely
populated region municipalities into one type (no. 5). Also tourism, rural and sparsely populated
region municipalities were joined (no. 6). The new names to municipality type 5, industrial and
commuting and 6, low density and tourism are found in Table 1.
In the regression analysis we use the responses to Question 32 in the survey, that is 21 different
alternatives, as outcome variables, see Table 2.
8
Table 2. Question and response alternatives used as dependent variables.
Survey question Response alternatives % Q32 What is most important in the dwelling for you? Choice of seven most important properties in the dwelling. That the dwelling is one floor 1176 49,0 That the dwelling has an elevator if more than a second floor That the dwelling is design for disability That the dwelling is easily maintained That the dwelling is design in a way that I/we can manage ourselves That the dwelling makes it possible for me/us to practice hobbies That there is room for social events like parties, dinners and meetings etc. That the family can come and stay in the dwelling That one can have pets in the dwelling That I/we own the dwelling That the dwelling has a nice view That the dwelling has a balcony/terrace That the dwelling has a private garden That there are good possibilities for parking close to the dwelling That the dwelling is located in an area where I feel at home That the dwelling is located close to the family That the dwelling is located close to forest and land That the dwelling is located close to city life/environment That the dwelling is located close to one or more grocery shops That the dwelling is situated in an area with large supply of service and culture That the dwelling is located close to public transport 868 407 866 1090 36,2 17,0 36,1 45,4 397 233 16,5 9,7 1098 448 817 712 1513 726 805 938 452 669 236 1073 383 45,8 18,7 34,0 29,7 63,0 30,3 33,5 39,1 18,8 27,9 9,8 44,7 16,0 1078 44,9 Geographical context
In addition to individual level background variables (age, sex, education, income) and
municipality type the regression analyze will also estimate the effect on housing preferences of
the type of areas that the respondent live in. The area types used for this analysis are based on a
cluster analysis of variables describing the socio-economic composition of the geographic
environment of residential locations in Sweden. Ten area types have been identified: Prime Elite
Suburban elite, Major city elite, Single mother neighborhoods, Middle class, Low middle class,
White unemployed, White marginal and Marginal non-elite, see Figure 6. The approach used to
construct these clusters is described in more detail below.
9
Using multi-scalar measures of context to construct geographical clusters
A common approach for measuring the effect of geographic context is to analyze how individual
-level attitudes and behavior varies between residential areas that have been classified on the
basis of their population composition. The classification can be done in different ways but
multivariate cluster analysis tends to be the preferred tool. What is not always acknowledged is
that variables that have been computed as aggregates for fixed geographical sub-divisions such as
census tracts or output areas can be strongly influenced by the way the boundaries of these areas
have been drawn. Moreover, using aggregates for fixed geographical sub-divisions to classify
residential areas implies an assumption that only conditions within the sub-division matters for
individual level outcomes. This is a very restrictive assumption for how geographical context can
play a role, and it has recently been demonstrated that it can lead to seriously biased estimates.
Using aggregates for individualized scalable neighborhoods to measure context has been
proposed as method for circumventing this problem. Such individualized neighborhoods can be
constructed by expanding a circular buffer around different residential locations until the
population encircled by the buffer corresponds to a selected population threshold. When this
threshold is reached, one can compute aggregate statistics on selected socio-economic variables
for the encircled population. By varying the population threshold, contextual measures computed
in this way can be designed to focus only the closest neighbors or on larger number of neighbors.
In the present study, we allow for multiple scales of ecological influence by varying k, the
number of nearest neighbors that are included in the computation of population, from 12 to
12 800 in successive doublings of the population thresholds. The computation was carried out
using a new software, Equipop, developed by John Östh in order to address the modifiable areal
unit problem (MAUP) in segregation measurement (Östh, Malmberg, and Andersson 2011).
Equipop requires that the input data are geo-coded to a high level of detail. In this case data from
the Population Labor Market Chorology database (PLACE) of Uppsala University has been used.
PLACE contains register-based, individual level information for the population in Sweden from
1990 to 2010 with geocodes of the residential location in 100 meter squares. 8 different sociodemographic indicators were extracted to use as input for EquiPop. On the individual level these
indicators are: (1) Being unemployed, (2) Having a tertiary education (3) Being a single mother
(4) Belonging to the top-ten percent of the income earners (5) Arrived in Sweden during the last
five years (6) Having received social allowance during the year (7) Without employment during
the entire year, and (8) Country of birth in Asia, Africa, or Latin America. Before being imputed
these data were aggregated to 100 meter squares based on the geo-coordinates.
8 socio-demographic indicators and 11 different scale levels (k = 12, 25, 50, 100, 200, 400, 800,
1600, 3200, 6400, and 12800) result in 88 different measures of neighborhood context that can be
used to classify residential areas using cluster analysis. However, since many of these variables
will be highly correlated we have subjected the contextual indicators to a factor analysis that
10
compresses the 88 original indicators to 15 orthogonal factors before proceeding with the cluster
analysis. The factor analysis was based on correlations and the number of factors was selected
based on them having eigenvalues higher than one. The factor analysis was based on correlations
and the number of factors was selected based on them having eigenvalues higher than one. The
factors were rotated using the varimax method.
Figure 1 illustrates the results of the factor analysis. The panels in this figure show the loading of
the different factors (columns) for each indicator (rows). The loadings are indicated by a graph
showing the values for different k-levels. Every panel also has a reference line indicating the
zero-level. The first factor in Figure 1 (left most column) represents an elite context with high
values for tertiary education and top ten percent income. The second factor represents top-ten
near because of high values for top ten percent income for lower k-levels (closer neighbors). The
fourth factor indicates a high share of newly arrived immigrants and visible minorities (VM). The
fifth factor signals marginal groups in adjacent areas with high values for many marginal groups
but only for relatively large k-values (distant neighbors). The sixth factor highlights a high share
of visible minorities but not newly arrived. The tenth factor represent white non-elite context with
high levels of unemployment and low values for top ten percent income. The seventh factor has
high loadings for social benefits. The ninth factor has high loadings for single mothers. The
eleventh factor has high loadings for non-employed. This can indicate either low employment
shares in the working age population or a high share retirees (65+). The interpretation of the
remaining factors can be done from the graphs.
11
1
Elite
3
4
5
6
7
8
9
g10
11
12
13
14
15
Singl
Top Top VM, Margi
Social Margi Singl White
Terti
Non- e
Misc
Misc
ten ten new, nal VM benef nal
e nonary
emp
2
1
adjac
near adjac near adjac
it medi near elite
near
Top Tertia Non Singel Unem Social New Visibe
10
ry
Emp Moth
p
Ben Arriv l Min
0.7
0.3
-0.1
2
0.7
0.3
-0.1
0.7
0.3
-0.1
0.7
0.3
-0.1
0.7
0.3
-0.1
0.7
0.3
-0.1
0.7
0.3
-0.1
0.7
0.3
-0.1
10
1000 10
1000 10
1000 10
1000 10
1000 10
1000 10
1000 10
1000 10
K
1000 10
1000 10
1000 10
1000 10
1000 10
1000 10
1000
Figure 1. Description of 15 factors and their loadings.
As a result of the factor analysis every populated 100 meter square in Sweden can be assigned
factor scores for each of the 15 contextual factors. Together these factor scores can be seen as
determining the position of different geographical location in a socio-demographic contextual
space. Geographic context in this representation does not only consist of the demographic
composition of a residential area. Instead, context is a composite feature that includes
characteristics of both the closest environment and of more extended local contexts.
What we propose in this paper is that these factors can be used to classify residential areas into
categories that capture important dimensions in the spatial variation of geographical context. This
is achieved by computing residential-area averages for the 15 factors and, then, to use cluster
analysis of these averages to group residential areas into classes with varying contextual
characteristics. Although this can appear to be almost equivalent to a cluster analysis based on
characteristics of the residential area population it is not. Consider for example a residential area
where no one has a tertiary education but with neighboring areas that have a high percentage of
highly educated. The population composition of these neighboring areas will affect the averages
of Equipop-based contextual measures but not measures that are based only on the population
12
composition of the residential area in focus. This implies that our proposed approach will not be
as strongly influenced by MAUP considerations as the traditional approach. This increases the
probability that contextual effects on residential preferences will be successfully captured by a
regression design that uses residential-area categories in an explanatory role.
Using Ward's clustering method we initially grouped the SAMS areas into 20 different clusters.
However, in order to avoid having cluster with comparatively few SAMS-areas, the number of
clusters was reduced to 10 by merging smaller cluster into larger ones. The principle used was
that clusters identified as being too small were merged with the cluster having the most similar
mean factor scores (minimum Euclidian distance between cluster centroids).
In Figure 2, the mean factor scores for the resulting ten contextual clusters are presented. The
names given to the clusters are based on the profile of these mean factor scores and on the
municipality type where the cluster is found, see Figure 3.
p
Cluster means
Prime Elite
1
Sub urban elite
1
Major city elite
1
Single mother
1
Middle class
1
Low middle
class
1
White unemp
1
White marginal
1
White older
1
Marginal nonelite
Soci Mar Sing Whit Non Sing
Top Top VM, Mar
Terti
Misc
Misc
e
le
Elite ten ten new, gina VM al gina le
ary
2
1
ben l near non emp adja
near adja near l
near
dj
# Factor
d
New Cluster formula
1
Figure 2. Clusters explained by factors.
The Prime Elite clusters have high values for the Elite, the Top Ten Near, and Top Ten Adjacent
factor. The Sub Urban Elite cluster has positive but smaller values for the same factors and a
13
negative value for the White Non-Elite factor. The Major City Elite cluster has a high value for
the Elite factor but negative for the Top Ten Near and Top Ten adjacent. This cluster also has a
high value for the Marginal Group Nearby factor. The Single Mother cluster has high values for
the Single Mother factor, the Marginal Nearby and VM factor, as well as for the Non-Employed
factor. The characteristic of the Middle-Class cluster is factor scores close to the mean on all the
factors. This cluster, thus, represents areas in which the composition is close to the Swedish
average. The Low Middle-Class cluster is similar but it has a higher value on the Non-Employed
factor. The White Unemployed cluster is also similar to the Middle-Class cluster but has high
values for the White Non-Elite factor. The White Older cluster has high values for the NonEmployed and Single Mother Factor. The Marginal Non-Elite has high values for factors with
high loadings for visible minorities, social allowances, and new arrivals. The White Marginal
cluster has positive but lower values for the same factors. See Figure 2.
1.00
Marginal non-elite
White older
White marginal
White unemp
0.75
Low middle class
0.50
Middle class
Single mother
0.25
2011
Industrial and
commuting
Major cites suburban
K
Low density and
tourism
Major cities
Metropolitan
suburbs
Metropolitan
0.00
Major city elite
Sub urban elite
Prime Elite
Figure 3 Municipality type and clusters.
14
Figure 4 presents a map of the geographical location of the different clusters which helps explain
the results of the preferences in different clusters/area types. A close up of mid-Sweden and six
major Swedish cities are provided in the appendix.
Stockholm area
Sweden
Gothenburg area
Scania area
Samsclusters
Low middle class
Middle class
No data
Cluster
Single mother
Marginal non-elite
Major city elite
White older
Sub urban elite
White marginal
Prime Elite
White unemp
0
0
25
50 Km
110 220 Km
Figure 4. Map of clusters, Sweden SAMS-areas with enlargement of Stockholm, Malmö and Gothenburg.
Results: Influences on Housing Preferences
The estimation results for the 21 logit models are presented in Table 3, Figure 5, and Figure 6.
Table 3, first, shows the likelihood ratio Chi-Square values for the explanatory variables for each
response to the question What is most important in the dwelling for you? (Q32). The Chi-Square
values indicate the strength of the effect of the explanatory variable and, can be used to calculate
significance levels. In the table significance levels below 0.001% are marked in yellow and levels
below 1% in pink. Figure 5 presents graphically the point estimates for the different levels of the
explanatory variables, and Figure 6 presents a close up for the effects of different clusters on
responses.
15
The first results concerning age groups show that questions concerning the interior of the
dwelling, alternative answers 1 to 9, all have significant and important ChiSquare estimates,
Table 3. The necessity to divide elderly into age groups with different preferences is strengthened.
Secondly the contextual effects on housing preferences among elderly cannot be ignored. There
are surprisingly many significant survey responses in the Area type column in Table 3 The
significant response results concern the location and environment of the dwelling (Q32,
alternative 10-13, 17-21). This is in a way inherent in that responses reflect present housing
(cognitive dissonance) among the elderly but nevertheless show significant results.
Table 3. L-R ChiSquare, significant <0.001 in pink and yellow.
Less pronounced in Table 3 (fewer significant, colored ChiSquare estimates), nonetheless our
results also show significant results in differences in housing preferences among elderly men and
women. Gender, thus, is another important aspect to consider in the planning of housing for
elderly. Also some significant results are found from municipality type for the preferences of
elderly. Finally and somewhat surprising when compared to earlier research is the little
significance found for housing preferences for differently educated groups and income groups.
Normally this is referred to in the ‘right to choose’ debate of care and planning for elderly; that
16
preferences are different and that the supply of housing must differ in both profile and quality to
meet demands of e.g. well-educated middle class in the future.
As stated in the introduction we will analyze the preferences linked to age, socio-economic status
and geographical context. Are there differences in preferences between different age groups
supporting the debate on limiting specific housing for certain ages? And most importantly,
differences in preferences for different structures of local housing markets; are there differences
in preferences between municipality types and geographical contexts/area types?
The above overall description of variables (age, geographical context, sex, municipality type,
education and income) of importance for housing preferences will be elaborated on in further
detail below. We will show for instance the difference between age groups’ preferences etc. Age
groups will start the section of results to be continued with geographical context, sex,
municipality type, education and income as important determinates of housing preferences
among the elderly.
17
0.3
That$the$dwelling$is$one$floor$
-0.3
0
-0.3
-1
n
Area$Type$
$
Municipality$
$type$$
0.4
0
-0.3
-0.5
0.3
0.4
0.2
0
0.5
0
-0.5
-0.4
0.4
0.2
0.1
0
-0.3
That$the$dwelling$has$a$nice$view$
0
0
-0.5
-0.5
-1
-1
-0.4
-1.5
-0.2
-0.3
0.3
0.5
-0.3
-0.4
That$there$are$good$possibilities$for$
parking$close$to$the$dwelling$
0.4
0.2
0
-0.3
0.4
0
-0.1
-0.3
0
-0.3
-1
0
0
-0.2
-0.3
0.4
0.2
0
0
-0.1
-0.3
0.4
0
0
-0.2
-0.2
-0.3
!
!
Primary!
!
0
-1
-1.5
Female!
!
Male!
1
-0.5
-1
-0.4
!
!
0.5
0
-0.4
75#84!
85#105!
!
0
-0.5
!
!
-1
0.5
0.2
0.1
-1
-1.5
0.3
-0.1
-0.5
-0.5
-1
-0.4
0.2
0
!
!
55#64!
65#74!
$
1
0.5
0
-0.4
0
-0.5
-0.2
-0.2
0.5
-1
0.5
0.2
0.1
-1
-1.5
0.3
-0.5
-0.5
-1
-0.4
-0.4
0
1
0.5
0
-0.5
-0.2
0
0.5
-1
0.5
0.2
0.1
-0.1
-1
That$the$dwelling$is$located$close$to$
public$transport$
0.4
0.2
0
-1.5
0.3
0.5
-0.5
-0.5
-1
-0.4
0
-0.5
-0.2
-0.2
-0.3
Data
1
0.5
0
0
-0.1
-0.4
That$the$dwelling$is$located$close$to$
one$or$more$grocery$shops$
That$the$dwelling$is$situated$in$an$
area$with$large$supply$of$service$and$
culture$
-1
0.5
0.2
0.1
-1
-1.5
0.4
0.2
-0.5
-0.5
-1
-0.4
0.3
0
0
-0.5
-0.2
-0.2
0.5
1
0.5
0
0
-0.4
That$the$dwelling$is$located$close$to$
city$life/environment$
-1
0.5
0.2
0.1
-0.1
-0.5
-1.5
0.4
0.2
0
-0.5
-1
-0.4
0.3
0.5
0
-0.5
-0.2
-0.2
1
0.5
0
0
-0.4
That$the$dwelling$is$located$close$to$
forest$and$land$
-1
0.5
0.2
0
-1
-0.5
-1
0.4
0.2
0.1
-0.5
-1.5
0.3
0
0
-0.5
-0.4
-0.4
1
0.5
0
0
-0.2
-0.2
-0.3
0.5
That$the$dwelling$is$located$close$to$
the$family$
-1
0.5
0.2
0.1
-0.1
-1
-1.5
0.2
0
-0.5
-0.5
-1
-0.4
0.3
0.5
0
-0.5
-0.2
-0.2
-0.4
That$the$dwelling$is$located$in$an$
area$where$I$feel$at$home$
1
0.5
0
0
-0.1
-1
-1
0.5
0.2
0.1
0
-0.5
-1.5
0.3
0.5
-0.5
-1
-0.4
0
-0.5
-0.2
-0.2
-1
1
0.5
0
0
-0.1
-0.5
0.5
0.2
0
-1.5
0.4
0.2
0.1
0
!
!
Prime!elite!
Sub!urb!
Maj!citty!!
Sing!moth!
Mid!class!
Low!middle!
W!unemp!
Wh!marg!
W!older!
Marg!non!
!!!
!
Data
0
-1
Metro!
Metro!Sub!
Major!cit!
Major!Sub!
Low!dens.!
Ind!!&!Com!
!
10!
25!
50!
75!
90!
0.5
-0.5
-1
-0.4
-0.4
That$the$dwelling$has$a$private$
garden$
1
0
0
-0.2
0
-1
0.5
-0.5
0.2
0.1
-0.1
0.4
0.2
0
-0.5
Data
Data
Tertiary!
Secondary!
0.3
0.5
That$the$dwelling$has$a$
balcony/terrace$
-0.4
0.5
0
0
-0.2
-0.2
-0.3
0
-0.5
1
0.5
0.2
0.1
-0.1
-1
-1.5
0.4
0.2
0
-0.5
-1
-1
-0.4
0.3
0.5
0.5
0
-0.5
-0.2
-0.2
-0.4
1
0.5
0
-0.1
-1
-1.5
0.2
0
-0.5
-1.5
-1
-0.5
-1
0
0
-0.2
-0.2
-0.3
1
0.5
0.2
0.1
-0.1
-0.4
-1.5
-1
-1
-0.4
0.3
0
-0.5
-0.2
-0.2
0.5
1
0.5
0
0
-0.4
That$I/we$own$the$dwelling$
-1
0.5
0.2
0.1
-1
-1.5
0.3
0.2
-0.5
-0.5
-1
-0.4
-0.1
0
0
-0.5
-0.4
1
0.5
0
0
-0.2
-0.3
-1
-1
0.5
0.2
0
0.5
0.4
0.2
0.1
-0.5
-1.5
0.3
0
-0.5
-1
-0.4
0
-0.5
-0.4
-0.1
0.5
That$one$can$have$pets$in$the$
dwelling$
0.5
0
0
-0.2
-0.2
-1
That$the$family$can$come$and$stay$in$
the$dwelling$
0
0
1
0.5
0.2
0.1
0.5
-1
-1
0.4
0.2
-0.2
-0.3
-0.5
-0.5
-0.4
-1.5
-0.1
-1
0
-0.5
-0.4
0.3
0
1
0.5
0
0
-0.2
-0.3
-0.5
0
0.5
0.2
0
-0.2
0.5
-1
-0.4
-0.5
-1
0.4
0.1
-0.1
-1
-1.5
0.2
0
1
0.5
0
-0.5
-0.4
0.3
0.5
That$the$dwelling$makes$it$possible$
for$me/us$to$practice$hobbies$
That$there$is$room$for$social$events$
like$parties,$dinners$and$meetings$
etc.$
0
-0.2
-0.3
-0.5
0
-0.4
Q
-1
0.5
0.2
0
-0.2
-1
That$the$dwelling$is$design$in$a$way$
that$I/we$can$manage$ourselves$
0.4
0.2
0.1
-0.1
-0.5
-1.5
0.3
0
-0.5
-1
-0.4
Data
1
0.5
0
-0.5
-0.4
That$the$dwelling$is$easily$maintained$
0
-0.2
-0.3
0.5
-1
0.5
0.2
0
-0.1
-0.2
-1
0
-0.5
-1
0.4
0.2
0.1
0
-0.5
-1.5
0.3
0.5
0.5
0
-0.5
-0.4
-0.4
1
0.5
0
-0.2
-0.2
-1
-1.5
1.5
0.2
0.1
-0.1
-0.5
-1
0.4
0.2
0
-0.5
That$the$dwelling$is$design$for$
disability$
0.3
0.5
Data
0
-0.5
-0.4
-0.4
0.5
0
0
-0.2
-0.2
Data
0
1
0.5
0.2
0.1
-0.1
-1
That$the$dwelling$has$an$elevator$if$
more$than$a$second$floor$
0.4
0.2
0
-0.5
Data
Data
Data
Data
0.5
Disp$Income$
percentile$
Education$
Sex$
Age$
Q32$What$is$most$important$
in$the$dwelling$for$you?$$
Figure 5. Effect of age, sex education, income, municipality type, and cluster type on responses to the question "What is
the most important in the dwelling for you".
18
Age effects on preferences
The results for outcomes about location and environment show age effects in several alternatives
Figure 5. Public transport in the vicinity is less important with increasing age except for the
youngest 65 to 75 years, also decreasing with age is preferences for closeness to forest and land,
that the dwelling has a garden, that we/I own the dwelling, that one can have pets, that the family
can come and stay, that there are room for social events like parties, dinners and meetings, that
we/I can practice hobbies and that the dwelling is easily maintained (first column, Age, Figure 5).
Preferences increasing with age are, that the dwelling is located in an area where I feel at home,
that the dwelling is designed in a way that we can manage ourselves (except for the oldest, 85+),
that the dwelling is designed for disability, that the dwelling (if higher than on the second floor)
has an elevator, and finally increasingly important with age is that the dwelling has one floor
(first column, Age, Figure 5).
Geographical context, area types
In this section we will refer to the map in Figure 4 and its colored pattern but also to Figure 6 for
a quicker overview of how respondents in different geographical contexts/clusters responded.
Prime elite (202 respondents): The cluster type of Prime elite areas is visible as purple areas. The
areas are characterized by having a higher proportion of people with tertiary education and high
shares of people with top incomes. These elite concentrations are found in the metropolitan cities
and in major cities, as well as in attractive coastal locations. These are areas that have typically
experienced favorable development in the last 20 years with population gains and employment
growth. Our logistic regression on housing preferences as outcomes shows interesting results for
elderly in Prime elite areas, Figure 6. The respondents stated that the most important in the
dwelling was the location close to a large supply of service and culture, that the dwelling should
be located close to city life/city environment and that it should have a nice view. If compared to
the map of Prime elite cluster locations these preferences seemed to be able to be fulfilled there.
Elderly living in Prime elite areas also preferred to have room for social events, hobbies and that
the dwelling should have an elevator. These preferences were also by large possible to fulfill in
Prime elite areas. The preferences with positive estimates for Prime elite clusters can be
considered indicators for higher classes in the Swedish society. Also the negative estimates, that
is, the preferences not important for elderly in Prime elite clusters can be seen as indicators of
higher classes. Negative estimates were found for importance of a garden, and that the dwelling is
located close to family and that the family can come and stay in the dwelling.
Weak elite suburban (205 respondents): This cluster is mainly found in the sub-urban parts of
metropolitan Stockholm and metropolitan Göteborg, but there are also some such areas in
Western Skane and in some parts of the major cities. Individuals in this cluster value having a
garden and having a dwelling that include room for hobbies. Less stress is given to having and
19
urban environment and good communications. A dwelling designed for disabilities and having an
elevator is also of less importance. Instead people living in these areas think that having a
dwelling on one floor only is a plus.
Weak elite, major city (255 respondents): This cluster is well represented in Göteborg and Malmö
but less well represented in Stockholm. It is also very well-represented in major cities such as
Uppsala, Västerås, Örebro, Lund, Linköping etc. Individuals in this cluster put great stress on
living in an urban environment but little stress on living near nature, owning their own dwelling,
having a dwelling suited for hobbies and pets. Instead they stress the importance of having an
elevator if more than a second floor and they also prefer an area with large supply of service and
culture.
Middle class (427 respondents): Areas belonging to the middle class cluster tend to be found in
rural areas under urban influence. This cluster is, for example, well represented in the non-urban
areas north and west of Stockholm, east of Göteborg, and in western Skane. It is also found in the
rural areas around Linköping and Jönköping, in central Gotland, around Falun and Borlänge;
around Örebro, north of Östersund, and in the coastal areas of north Sweden. Individuals in these
areas express a strong preference for having a garden and a preference for having access to forest
and nature. They also have a tendency to stress owning their dwelling, having a dwelling good
for pets and where family can stay. Little emphasis is given to living in an urban environment,
good communications, availability of grocery stores, or having a terrace or elevator.
Low middle class (380 respondents): Low middle class areas tend to be located at larger distances
from major urban centers, in border areas between more central regions and in areas strongly
specialized in agriculture. Examples include the most peripheral parts of the Stockholm region,
the most peripheral parts of the Skane region, and the most peripheral part of the Göteborg region
between Lake Vänern and the Norwegian border. Examples also can be found in the inland parts
of northern Sweden between the costal and mountain areas. The residential preferences of the low
middle class areas are similar to those of the middle class cluster.
Single Mother (247 respondents): The Single mother cluster is mostly represented in the
metropolitan areas and in the major cities. Individuals living in these areas tend strongly tend not
to stress the importance of having a garden. Neither is it important to have room for social events
or room for having family members staying. Likewise, being close to forest and owning one's
own dwelling is not emphasized. What is important is instead to have a balcony or terrace, having
an elevator, good communications, an urban environment and, to some extent, having a great
view.
20
Figure 6. Important traits (preferences) for elderly living in different clusters.
White older (287 respondents): Areas belonging to this cluster tend to be located in smaller
Swedish towns in different parts of Sweden. Individuals living in these areas value an urban
environment and availability of grocery stores but not closeness to nature and forest.
White marginal (152 respondents): Areas in this cluster tend to lie close to white unemployment
cluster areas. To live in these areas is not associated with any significant patterns in residential
preferences.
The white unemployment (140 respondents) cluster demonstrate high levels of unemployment in
the more sparsely populated, forested areas of northern and north-western Sweden, as well as in
areas that are distant from the metropolitan areas and other major cities. Preferences with
negative estimates for elderly in these contexts are that the dwelling should be located close to
public transportation, service and culture, grocery shops and close to city life/ city environment.
According to our mapping these preferences correspondingly seem difficult to fulfill at least in
some of the white unemployment areas. The less importance of the very same preferences is
contrasting those by elderly in Prime elite areas, see Figure 6. However, positive estimates were
21
found for preferences for closeness to forest and land, that I feel at home in the area, that the
dwelling has a garden, that we/I own the dwelling and that the dwelling makes it possible for
me/us to practice hobbies.
Marginal non-elite (104 respondents): Figure 5 shows that the marginal non-elite cluster is
associated with metropolitan areas and with Sweden’s larger cities but also with some major
cities. This macro-level segregation is associated with the establishment of large housing estates
during the 1960s and 1970s that today have an overrepresentation of visible minorities, newly
arrived immigrants, and social allowance recipients. Preferences found in these areas are that the
dwelling should be located close to public transportation, grocery shops and close to city life/ city
environment, see Figure 6.
Most importantly marginal non-elite areas are the only ones with high positive estimates for the
dwelling to be located close to the family. Clearly not important in this geographical context is to
feel at home in the area, possibilities of parking, having a garden, or that the dwelling makes it
possible to practice hobbies. These preferences are rather much in accordance with the living
opportunities in the brownish/darkest colored areas in Figure 5. Effect of age, sex education,
income, municipality type, and cluster type on responses to the question "What is the most
important in the dwelling for you".
Gender differences
Eight housing preferences significant for gender were found, see Table 3. The preferences
primarily concerned the whereabouts of the dwelling. The most important one was that the
dwelling according to women should be located close to the family. Secondly the dwelling
should have a balcony or terrace. Women also preferred if the dwelling had an elevator should it
be located higher than on the second floor.
Men, on the other hand answered that owning the dwelling was important as well as a location
close to forest and land. Also important was having a private garden as well as parking facilities
and the possibilities to practice hobbies in the dwelling.
Municipality type
Six housing preferences differed significantly with municipality type, Table 3. An association
between the preference of the dwelling to be located close to public transport and the
municipality type according to population density/urbanity could be observed. The metropolitan
cities, suburbs of metropolitan cities, and major cities were home to respondents answering that
public transport nearby was important. Public transport was however not important to those
living in municipality types of suburbs of metropolitan cities, industrial and commuting and low
density and tourism type municipalities. In the same way elevator if the dwelling was situated
22
higher than second floor was only important to those living in the three most urban municipality
types but not in the three less population-dense/ rural municipality types. Of course public
transportation and elevator is inherent in urban areas whereas the same preferences would not
naturally show among respondents living in rural/sparsely populated areas. Here we can observe
the above mentioned sorting according to preferences and/or adjusted preferences probably
visible in the municipality significances, that is, a selection effect. An additional sorting
according to preferences is that those living in the more rural and sparsely populated municipality
types preferred having a garden of one’s own to a higher degree than others. Also connected to
the urban – rural divide but with mixed results, were respondents in metropolitan cities preferring
the dwelling to be located close to city life or city environment whereas they didn’t prefer forest
and land to be close to the dwelling.
Education and Income
The overall analysis of housing preferences showed, as stated above, that education and income
did not have significance to a very large degree compared to age and geographical context. We
could however observe some associations in that the higher educated the more preferences were
found for large supplies of service and cultural facilities in the area, that the dwelling should have
room for social events like parties, dinners and meetings, and provide possibilities to practice
hobbies. The two latter preferences were also associated with higher incomes. That the dwelling
was designed in a way that one can manage without help from others was preferred by those with
lower education as well as those with lower incomes. Significant preferences among lower
income groups were also the fact that the dwelling is located in an area where the respondent feel
at home.
Concluding Discussion
In this paper we have added to the research on preferences and considerations of the supply of
housing for elderly. We have analyzed the preferences linked to age, socio-economic status and
geographical context. The main questions were if there are differences in preferences between
different age groups? and if there are differences in preferences between municipality types and
geographical contexts? The findings indicate precisely that age and area type together with
municipality type is highly important for preferences held by elderly in our SHIELD survey.
Concerning these questions this paper makes three important contributions. First, we have been
able to show that neighborhood context plays a fundamental role in housing preferences among
elderly Swedes. The strongest effects of geographical context are found for preferences that are
related to the type and the location of dwellings rather than the dwelling’s physical design. One
suggestion that explains this is that age is the determining factor for the interior of the dwelling in
23
terms of needs and preferences regarding for instance disability. This is also in line with the
finding in our study that age is rather everywhere putting demand on the interior of the dwelling
such as the dwelling in one floor, elevator, design for disability and a dwelling designed in a way
that we/I can manage myself/ourselves.
Second, we have demonstrated that a classification of areas based on multi-scalar measures of
geographical context provides a powerful tool both for analyzing spatial variation in socioeconomic structure and for analyzing how residential preferences vary geographically. Even if
the multi-scalar approach is more cumbersome compared to methods that build on aggregate
fixed geographical subdivision it leads to strong and useful results. One particular strength of our
cluster based approach is that survey data can be used to identify spatial patterns in residential
preferences not only for a few selected locations but in a way that allows a generalization to the
whole of Sweden.
Third, our analysis shows that variation in residential preferences among the elderly does not
follow a simple urban rural dichotomy. For sure, there are contrasts between metropolitan and
non-metropolitan areas. But also within these two broad groups there is substantial variation, for
example with respect to the preference of access to service and centrality.
Despite the strong and systematic results pointing in the same direction of context being
important there are certainly limitations to the result that most preferences are formed locally. As
mentioned in the introduction we also admit that preferences govern choice of housing in a way
that people can already be sorted according to preferences. However using a method where scale
is inherent is in some ways protective to that. Another limitation to the idea off effects from the
geographical context is that respondents most likely form justifications for living in a certain
dwelling/location and for moves already made.
What we have not, for reasons of data availability, been able to do in this paper is to
systematically explore how variations in socio-economic context relate to the difference in
housing market structure. Such differences are clearly of large importance for the local sociodemographic composition and access to such information would have provided additional
insights into how residential preferences influence and are influenced by local conditions.
Acknowledgement: The Equipop software used for computing contextual variables based on individualized neighborhoods was
developed by John Östh, Uppsala University. This research was supported by the Linnaeus Center on Social Policy and Family
Dynamics in Europe, SPaDE, Stockholm University and by the Swedish Research Council grant 2009-6299.
24
25
References
Abramsson, M., and E. Andersson. (forthcoming). Changing locations –Central or peripheral
moves of seniors? Resubmitted.
Abramsson, M. & E. K. Andersson. 2012. Residential mobility patterns of older people - leaving
the house for an apartment. Housing Studies 27(5): 582-604.
Abramsson, M. 2012. Housing careers. In S. J. Smith, M. Elsinga, L. Fox O'Mahony, Ong Seow
Eng, S. Wachter & D. Clapham (eds.), International Encyclopedia of Housing and
Home, 385-389. Oxford: Elsevier.
Abramsson, M., A. Elmqvist, A. & L. Magnusson Turner. 2012. Äldres flyttningar och motiv till att
flytta eller bo kvar. Report. Hjälpmedelsintsitutet, Stockholm.
Andersen, A. S. (2005) Boforhold og nærmiljø. I E. Ugreninov (red.) Seniorer i Norge. Statistisk
sentralbyrå. Rapport 72 Oslo–Kongsvinger
Andersson, E. 2007. Sweden: To Own or to Rent. In Home ownership beyond asset and security.
Perceptions of housing related security and insecurity in eight European countries,
eds. M. Elsinga, P. Decker, D, N. Teller & J. Toussaint, 225-257. Amsterdam: IOS
Press.
Andersson, E. 2008. Risk and security in home owning and renting. An interview study in Sweden.
Gävle: Institute for housing and urban research, Uppsala University.
Bolster, A., S. Burgess, R. Johnston, K. Jones, C. Propper, and R. Sarker. 2007. Neighbourhoods,
households and income dynamics: a semi-parametric investigation of neighbourhood
effects. Journal of Economic Geography 7 (1):1-38.
Bowyer, B. T. 2009. The contextual determinants of whites’ racial attitudes in England. British
Journal of Political Science 39 (3):559-586.
Chaix, B., J. Merlo, S. V. Subramanian, J. Lynch, and P. Chauvin. 2005. Comparison of a Spatial
Perspective with the Multilevel Analytical Approach in Neighborhood Studies: The
Case of Mental and Behavioral Disorders due to Psychoactive Substance Use in
Malmö, Sweden, 2001. American Journal of Epidemiology 162 (2):171-182.
Chevan, A. 1995. Holding on and letting go: Residential mobility during widowhood. Research on
Ageing, 17, 278–302.
Cummins, S., M. Stafford, S. Macintyre, M. Marmot, and A. Ellaway. 2005. Neighbourhood
environment and its association with self rated health: evidence from Scotland and
England. Journal of epidemiology and community health 59 (3):207-213.
Cummins, S., S. Curtis, A. V. Diez-Roux, and S. Macintyre. 2007. Understanding and representing
‘place’in health research: a relational approach. Social Science & Medicine 65
(9):1825-1838.
26
Fransson, U. & Borgegård, L.-E. 2002. Lokala och regionala flyttningar och boendekarriärer
bland äldre. Bilagedel D till SOU 2002:29 Riv ålderstrappan.
Galster, G. C., and A. M. Santiago. 2006. What's the 'hood got to do with it? Parental perceptions
about how neighborhood mechanisms affect their children. Journal of Urban Affairs
28 (3):201-226.
Huber, J & Skidmore, P 2003. Why baby boomers won’t be pensioned off. London: Demos
Kawachi, I., and L. F. Berkman eds. 2003. Neighborhoods and health. Oxford, New York: Oxford
University Press.
Kwan, M.-P. 2012. The Uncertain Geographic Context Problem. Annals of the Association of
American Geographers 102 (5):958-968.
Laslett, P. 1996. A fresh map of life. The emergence of the third age. Basingstoke: Macmillan.
Litwak, E. & C. F. J. Longino 1987. Migration Patterns among the Elderly: A Developmental
Perspective. The Gerontologist, 27, 266–272.
Macallister, I., R. J. Johnston, C. J. Pattie, H. Tunstall, D. F. L. Dorling, and D. J. Rossiter. 2001.
Class Dealignment and the Neighbourhood Effect: Miller Revisited. British Journal
of Political Science 31 (01):41-59.
Malmberg, B., E. Andersson, and J. Östh. 2011. To what extent does the level of segregation vary
between different urban areas? Introducing a scalable measure of segregation. In
European
Network
for
Housing
Research
Conference.
Toulouse:
http://www.enhr2011.com/sites/default/files/Paper-BoMalmberg-WS16.pdf.
Matthews, S. A., and T.-C. Yang. 2013. Spatial polygamy and contextual exposures (spaces):
Promoting activity space approaches in research on place and health. American
Behavioral Scientist.
Myers, D. 1990. Filtering in time: Rethinking the longitudinal behaviour of neighbourhood housing
markets. In Myers, D. (ed) Housing Demography. Madison: University of Wisconsin
Press.
Niedomysl, T., and B. Malmberg. 2009. Do open-ended survey questions on migration motives
create coder variability problems? Population, Space and Place 15 (1):79-87.
Openshaw, S. 1984. The modifiable areal unit problem, CATMOG (Concepts and Techniques in
Modern Geography). Geo Abstracts:40.
Pettigrew, T. F., and L. R. Tropp. 2006. A meta-analytic test of intergroup contact theory. Journal
of personality and social psychology 90 (5):751.
27
Pettigrew, T. F., O. Christ, U. Wagner, and J. Stellmacher. 2007. Direct and indirect intergroup
contact effects on prejudice: A normative interpretation. International Journal of
Intercultural Relations 31 (4):411-425.
Pope, N. D., and B. Kang. 2010. Residential Relocation in Later Life: A Comparison of Proactive
and Reactive Moves. Journal of Housing For the Elderly 24 (2):193-207.
Reardon, S. F., S. A. Matthews, D. O’Sullivan, B. A. Lee, G. Firebaugh, C. R. Farrell, and K.
Bischoff. 2008. The Geographic Scale of Metropolitan Racial Segregation.
Demography 45 (3):489-514.
Sampson, R. J. 2012. Great American City: Chicago and the Enduring Neighborhood Effect.
Chicago: University of Chicago Press.
Spielman, S. E., and E.-h. Yoo. 2009. The spatial dimensions of neighborhood effects. Social
Science & Medicine 68 (6):1098-1105.
Spielman, S. E., and J. R. Logan. 2013. Using High-Resolution Population Data to Identify
Neighborhoods and Establish Their Boundaries. Annals of the Association of
American Geographers 103 (1):67-84.
Stein, R. M., S. S. Post, and A. L. Rinden. 2000. Reconciling context and contact effects on racial
attitudes. Political Research Quarterly 53 (2):285-303.
Warnes, T. 1992. Migration and the life course. In Champion, T. & Fielding, T. (eds.), Migration
processes and patterns (pp. 175-187). London: Belhaven Press.
Ytrehus, S. & L. Gulbrandsen. 2012. Äldres boendepreferenser i Norge (Housing preferences of
older people in Norway. In Äldres boende - forskningsperspektiv i Norden (Older
people's housing - Nordic research perspectives), eds. M. Abramsson & C. Nord,
139-152. Lund: Studentlitteratur.
Ytrehus, S. 2005. Mellom modernitet og tradisjon. Unge eldres syn på flytting. Avhandling for dr
polit-graden, Universitetet i Bergen 2004.
Ytrehus, S. and A. B. Fyhn. 2006. Bufast, Bjørvika og Benidorm. Boligplaner og boligønsker for
eldre år. FAFO-rapport 511.
Ytrehus,S. 2007. Elderly’s mobility and migration in Norway. Paper Nordic Congress of
Gerontology. Oslo.
Östh, J., B. Malmberg, and E. Andersson. (forthcoming 2014). Analysing segregation with
individualized neighbourhoods defined by population size. In Social-Spatial
Segregation: Concepts, Processes and Outcomes, eds. C. D. Lloyd, I. Shuttleworth
and D. Wong: Policy Press.
Östh, J., B. Malmberg, and E. K. Andersson. 2011. Introducing Equipop. 6th international
conference on population geographies, June 15, 2011.
28