Trends and gender associations in social exclusion in older adults in Sweden over two decades

Contents lists available atScienceDirect

Archives of Gerontology and Geriatrics

Trends and gender associations in social exclusion in older adults in Sweden

over two decades

Lena Dahlberg

_{*, Kevin J McKee}

_{, Johan Fritzell}

_{, Josephine Heap}

_{, Carin Lennartsson}

a_{Aging Research Center, Karolinska Institutet & Stockholm University, Tomtebodavägen 18A, 171 65 Solna, Sweden} b_{School of Education, Health and Social Studies, Dalarna University, 791 88 Falun, Sweden}

A R T I C L E I N F O Keywords: Social inclusion Poverty Social integration Social engagement Political engagement Older people A B S T R A C T

Background: Social exclusion in older adults is associated with lower well-being and poorer health. To date there has been little research on whether the level of social exclusion in older adults changes over time, and its association with gender.

Aim: To examine trends and gender associations in social exclusion indicators in older adults for the years 1992, 2002 and 2011.

Methods: Three waves of data from the Swedish Panel Study of Living Conditions of the Oldest Old (SWEOLD), a national survey of adults aged 77 years or older, were analysed: 1992 (n = 537), 2002 (n = 621), and 2011 (n = 904). Summative scales were created for four domains of social exclusion: material resources, social re-lations and leisure activities, civic participation, and services. Associations between gender and social exclusion within waves were examined as were trends in social exclusion across years.

Results: The analyses of trends found significant reductions in exclusion in the domains of material resources and services. Higher levels of exclusion from material resources and civic participation were found in women than men. Within domains, significant trends and gender associations in exclusion were found on several indicators, with indicators showing opposing trends.

Conclusion: Although levels of social exclusion have reduced in certain domains during the years examined, our results reflect the persistence of social exclusion in the population of older adults. This underlines the continuing importance of a well-developed welfare and social security system to ensure the social inclusion of vulnerable groups such as older adults.

1. Background

Social exclusion in old age has negative effects including reduced well-being and health and an increased risk of mortality (seeDahlberg & McKee, 2018;Lee & Cagle, 2018;Sacker, Ross, MacLeod, Netuveli, & Windle, 2017;Saito, Kondo, Kondo, Ojima, & Hirai, 2012). Combatting social exclusion is high on the European Union (EU) policy agenda, most recently in the European Pillar of Social Rights (European Commission, 2017). The concept has also been adopted by the United Nations (United Nations, 2010) and the World Health Organization (Popay et al., 2008). Despite this high profile, there has been little re-search on whether the degree of social exclusion in older adults changes over time, and how the level and change may differ between women and men. This article presents analyses of trends and gender associa-tions in social exclusion among older adults in Sweden between 1992 and 2011.

1.1. Theoretical framework: social exclusion

While there are many different definitions of social exclusion (see e.g. Burchardt, Le Grand, & Piachaud, 2002; Council of European Union, 2004; Levitas et al., 2007; Walsh, Scharf, & Keating, 2017), these definitions have some common features (for an overview, see Walsh et al., 2017). These include social exclusion being seen as a process rather than a static condition, and as normative and relative in that it relates to activities that are seen as standard within a given so-ciety (Percy-Smith, 2000;Silver & Miller, 2003). Another key feature is the multidimensionality of social exclusion, that is, individuals can be excluded from various life domains (Burchardt et al., 2002). Research on social exclusion in older adults commonly covers the domains of material resources, social relations, civic participation and access to services (Van Regenmortel et al., 2016;Walsh et al., 2017).

Living conditions in later life differ between women and men. For

https://doi.org/10.1016/j.archger.2020.104032

Received 11 October 2019; Received in revised form 4 February 2020; Accepted 17 February 2020

⁎_{Corresponding author at: School of Education, Health and Social Studies, Dalarna University, 791 88 Falun, Sweden.}

E-mail address:ldh@du.se(L. Dahlberg).

Available online 05 March 2020

0167-4943/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

example, health problems and mobility limitations are more common in older women than older men (Dahlberg, Agahi, & Lennartsson, 2018; Lennartsson & Lundberg, 2007; Schön & Parker, 2009), although women tend to live longer than men and also marry men who are older than themselves such that women are more likely to enter widowhood than men and at a younger age (Lennartsson & Lundberg, 2007). His-torically, women have been less active on the labour market than men (Esping-Andersen, 2016) and so tend to have lower pension incomes than men and a higher risk of poverty (European Commission, 2018). Living conditions such as health status and widowhood have been found to be determinants of social exclusion (e.g.Becker & Boreham, 2009; Kneale, 2012; Sacker et al., 2017). It could therefore be hy-pothesised that social exclusion is likely to manifest itself in different ways and to differing extents in older women and men.

1.2. Previous research on trends and gender associations in social exclusion among older adults

Research on trends in social exclusion among older adults has pri-marily been based on data from the United Kingdom (UK). An English national study of people aged 50 years or older compared the years 2002 and 2008 (Kneale, 2012) and found that there was little change in exclusion from the domains of financial products, social relations, cultural/leisure activities, or civic activities. Regarding specific in-dicators of social exclusion, the study found a reduction in exclusion regarding relations with friends but increases in difficulties accessing the dentist and the general practitioner, while trends in indicators of financial exclusion were mixed.

Another English national study found a reduction in relative poverty among pensioners from the beginning of the 1990s to 2014 (Tinson et al., 2016). A different UK study found, however, that although fewer older adults lived under the poverty line, there had been an increase in older adults living in severe and persistent poverty, that is, an in-creasing inequality within the older age group (McKee, 2010).

At a European level (European Commission, 2018), data from 2016 showed that among people aged 75 years or older, 22.8 percent of women and 15.4 percent of men were at risk of poverty or social ex-clusion, a measure combining risk of poverty and severe material de-privation (AROPE). These figures represent a decrease from 2008. In Sweden, the corresponding figures are considerably higher for women (31.7 %) and slightly lower for men (14.7 %), in both cases showing an increase from 2008. As far as we are aware, these are the only data on gender associations in trends in social exclusion among older adults. It should be noted, however, that although social exclusion is a key target in the EU 2020 strategy, the annual monitoring of social exclusion is of the material domain only.

While gender associations in older adults have rarely been the main focus of social exclusion research, as noted in two recent reviews (Van Regenmortel et al., 2016;Walsh et al., 2017), gender has been con-sidered as a determinant or correlate of social exclusion. Both an Eng-lish national study (Becker & Boreham, 2009) and a European study (Ogg, 2005) found that women were more likely than men to experi-ence social exclusion. There is however contrasting evidexperi-ence: an Aus-tralian study and an English community study found no association between gender and social exclusion (Miranti & Yu, 2015; Scharf, Phillipson, & Smith, 2005).

Regarding social exclusion from specific domains, research has found that women compared to men are at higher risk of exclusion from the domains of: material resources (Becker & Boreham, 2009;Heap, Fors, & Lennartsson, 2017;Kneale, 2012); civic participation (Del Bono, Sala, Hancock, Gunell, & Parisi, 2007;Kneale, 2012); cultural activities (Kneale, 2012); and access to services such as consulting a doctor (Becker & Boreham, 2009;Del Bono et al., 2007), and a lower risk of exclusion from social relations (Del Bono et al., 2007;Kneale, 2012).

In summary, there are only a few studies on trends in social ex-clusion among older adults, mostly from the UK, and a lack of research

on gender associations in social exclusion among older adults, espe-cially when considering trends in social exclusion.

2. Aim

The aim of this study is to examine trends and gender associations in social exclusion indicators in older adults living in Sweden for the years 1992, 2002 and 2011.

3. Methods

3.1. Design and participants

The analyses reported in this article are based on data from the Swedish Panel Study of Living Conditions of the Oldest Old (SWEOLD), a national longitudinal survey of people aged 77 years or older, which started in 1992 (Lennartsson et al., 2014). SWEOLD recruits those in-dividuals randomly sampled to participate in the Swedish Level of Living Survey (LNU) who have reached the upper age limit of that study. In this article, data from three data collection waves of SWEOLD, undertaken in 1992 (n = 537), 2002 (n = 621), and 2011 (n = 904), were analysed, as these used the same data collection procedures and comprehensive questionnaires. In the 2011 wave of data collection, the SWEOLD sample was complemented with an additional sample of the oldest old (85+ years), in particular men.

3.2. Material

SWEOLD uses an extensive questionnaire that covers a wide range of topics. Below, we describe those questions used as indicators of four domains of social exclusion. The operationalisations of these indicators are presented inTable 1.

The domain Exclusion from material resources contained two in-dicators: cash margin and financial resources. The domain Exclusion from social relations and leisure activities contained three indicators: so-cial contacts with friends, soso-cial contacts with children, and leisure activities. The domain Exclusion from civic participation contained four indicators: voting in public elections, reading newspapers, ability to deal with public authorities, and participation in organisations. The domain Exclusion from services contained two indicators: access to health care and access to dental care (waves 2002 and 2011 only).

The focus in this article is on gender associations in social exclusion. Information on gender and age was obtained from the national register as part of the sampling process.

3.3. Procedure

The SWEOLD questionnaire was primarily administered by means of a face-to-face structured interview conducted in the participant’s home, with some of the participants interviewed via telephone. In the 2011 survey, there was also an option for participant self-completion of the questionnaire. If a participant was unable to answer questions, for ex-ample, due to cognitive impairment or physical frailty, proxy interviews were carried out with the participant’s spouse/partner or another close person. The decision to undertake an indirect interview was made by the interviewer in consultation with a relative or professional carer. The interviewers were also instructed to terminate the interview if they suspected that the responses were not reliable and to ask for permission to undertake an indirect interview instead.

Informed consent was obtained from the participant prior to each interview. In order to ensure high quality interviews and reduce in-terviewer bias, all inin-terviewers were trained before the data collection started. The samples achieved in 1992, 2002 and 2011 represent re-sponse rates of 95.4 %, 84.4 %, and 86.4 %, respectively. Ethical ap-provals for the SWEOLD study have been provided by Uppsala University Hospital (reg.no. 247/91), Karolinska Institutet Regional

Research Ethics Committee (reg.no. 03-413) and the Regional Ethical Review Board in Stockholm (reg.no. 2010/403-31/4).

3.4. Analysis

Analyses explored trends and gender associations in social exclusion indicators and domains. For consistency, and with recoding of variables as appropriate, higher values for all social exclusion indicators re-present higher levels of exclusion.

A summative scale was created for each of the four domains of so-cial exclusion. As the indicators’ scoring ranges varied, the indicators within each domain were transformed into standardized z-scores (mean of 0 and standard deviation of 1) before they were added together to create the summative scales. The standardisation of indicators means that all indicators are of equal weight for the domain scale.

As mentioned above, an additional sample was drawn in 2011, and in 1992 there was a lower probability of 77 year olds to be included in the study than for people above that age. Data for these waves had sampling weights applied to adjust for the sampling probability (for further details, seeLennartsson et al., 2014).

In 1992, some questions were not given to people living in an in-stitution and/or in proxy interviews. Excluding cases that are missing not-at-random can lead biased results (van der Heijden, Donders, Stijnen, & Moons, 2006; see also Kelfve, Thorslund, & Lennartsson, 2013). For example, women are more likely to have poor health and to enter institutions (Schön & Parker, 2009; NBHW, 2019), which has implications when analysing gender effects (Peeters, Debels, & Verpoorten, 2013). Therefore, values were imputed for these in-dividuals (see Appendix A for details on the method of imputation).

For tests of statistical significance in gender associations, the phi

coefficient was calculated for dichotomous variables and Pearson’s correlation coefficient for interval variables. For tests of statistical sig-nificance in trends in indicators and domains over time, linear and lo-gistic regression analyses were used. These tests concern changes be-tween 1992 and 2011, unless otherwise stated.

SWEOLD has a longitudinal panel, which means that some partici-pants have been included in more than one wave of data collection. To correct for intra-individual correlations, trend analyses were carried out using Huber-White sandwich estimator of variance through which ro-bust standard errors are obtained. Trend analyses were conducted in Stata (15.0) All other analyses were conducted in SPSS 25.0 for Windows.

For all analyses, the criterion for test significance was p < .05. No adjustment to experimental alpha was made for multiple testing, and so note should be taken of the potential for inflated Type I error rate. Given also the discrepancy in size between male and female samples, the results of significance tests should be interpreted with reference to their respective effect sizes.

4. Results

The average age of the participants across the data collection waves was 82–83 years and approximately 60 percent were women.

Tables 2–5present findings regarding exclusion from different do-mains, with the first row presenting findings at domain level (as stan-dardised z-scores) and the following rows presenting findings regarding specific indicators.

Table 1

Domain indicator/items and response options.

Social exclusion domain Indicators/items Response options1

Exclusion from material resources Cash margin: ‘If a situation suddenly arose where you had to raise SEK 14 000 (approximately € 1 560; indexed value, here provided for 2011) in a week, would you be able to manage it?’

Withdrawal from their own bank account or similar (0); Loan from a family member (1); Loan from other relatives or friends (1); Bank loan or equivalent (1); Other way (1).

Financial difficulties: ‘Have you at any time over the last 12 months had difficulties managing your current expenses for food, rent, bills etc.?’ Yes; no. Exclusion from social relations and

leisure activities A scale for social contacts with friends based on two items: visitingfriends and having friends over to visit For each form of contact: no (0); yes, sometimes (1); yes, often (2).Scale range: from 0 (no social contacts) to 4 (many social contacts). Social contacts with children: ‘How often do you meet and spend time

with any of your children?’ Six response options anchored by seldom/never (1) and daily (6).Having no children coded as 0. Leisure activities: ‘Which of the following leisure activities do you

usually do?’ followed by a list of activities, of which ten were considered relevant to this domain: gardening; cultural activities (going to the movies, theatre, concerts, museums, exhibitions); eating out at restaurants; dancing; reading books; walks/Nordic walking; helping family members (with babysitting or other small favours); study circles or courses; hobby activities (e.g. knitting, sewing, carpentry, painting, stamp collecting); and solving crosswords.

For each activity: no (0); yes, sometimes (1); yes, often (2). Summative index range: 0–20.

Exclusion from civic participation Voting: ‘Did you vote in the [year of previous] elections?’ Yes; no. Reading newspapers: measured as one of the activities on the list

described above but used as an indicator of this domain. No (0); yes, sometimes (1); yes, often (2). Ability to deal with public authorities: ‘Would you be able to write a

letter yourself to appeal a decision made by a public authority?’ Yes; no. Participation in organisations: ‘How often do you participate in

activities with this or these organisations?’, described as pensioner organisations, political parties, and specified other organisations. Attendance in religious services did not count as a form of participation in organisations.

Five response options anchored by never or almost never (1) and several times a week (5). Not members of any organisation coded as 0.

Exclusion from services Access to health care: ‘Have you at any time during the last 12 months refrained from visiting a doctor even though you needed to?’ Yes; no. Access to dental care: ‘Have you at any time during the last 12 months refrained from visiting a dentist even though you needed to?’ Yes; no.

4.1. Exclusion from material resources

Between 1992 and 2011, there was a decrease in exclusion from material resources, as shown inTable 2. When looking at women and men separately, this trend was statistically significant for women only. Despite this reduction in exclusion, there was a consistent gender as-sociation with women being at higher levels of exclusion from material resources than men at all survey years.

When the individual indicators were analysed, 19.2 percent of the total sample did not have cash margin in 1992. There was a steady decrease in this proportion over the study period, with 13.2 percent of the participants lacking cash margin in 2011. At each point of mea-surement, men were better off than women, although the gender gap decreased over the period. Separate analyses for women and men also showed that the trend in cash margin was only significant for women. In 2011, 15.8 percent of the women lacked cash margin compared to 8.90 percent of the men.

Most of the participants had no financial difficulties. In 1992, only 1.50 percent of the participants had experienced difficulties. However, this proportion grew over the study period, with 4.45 percent having such difficulties in 2011, again this trend was significant for women only. At the beginning of the study period there was no significant gender association, but by the end of the period more than twice as many women as men experienced financial difficulties.

4.2. Exclusion from social relations and leisure activities

At domain level, there were no statistically significant changes in the level of exclusion from social relations and leisure activities and no statistically significant gender association (seeTable 3).

When analysing individual indicators in this domain, the mean score for the indicator low contacts with friends was 1.85 out of 4 in 1992. There was a significant increase in the mean score between 1992 and 2002 and a decrease between 2002 and 2011. This means that a lower proportion of older adults was at risk of exclusion from contacts with friends in the beginning than towards the end of the study period. There were no significant gender associations regarding social contacts with friends.

The mean score for low contacts with children was 2.73 out of 6 in 1992 and 2.49 in 2011, albeit the change over time was not significant. Men scored significantly higher than women on low contacts with children in 2011.

On the scale measuring low leisure activities, the participants had a mean of 14.65 out of 20 in 1992, 14.52 in 2002 and 13.95 in 2011. Women were less active in leisure activities than men, with this gender association significant in 2011 only.

4.3. Exclusion from civic participation

The summative scale for exclusion from civic participation indicated a reduction in exclusion from 1992 to 2011, although not statistically significant (seeTable 4). Women were significantly more excluded than men at each data collection wave.

When considering the individual indicators of exclusion from civic participation, in 1992, 22.6 percent of the participants had not voted in the previous general election (held in 1991). There was no significant change in non-voting by 2002, but between 2002 and 2011 the pro-portion of participants who had not voted in the previous election dropped to 16.1 percent. A larger proportion of women than men did not vote in the general elections, with the association being statistically significant in 1992 and 2002. Separate analysis for men and women showed that the overall trend of reduced non-voting was only sig-nificant for women.

The proportion of older adults who did not read newspapers in-creased during the study period: from 8.08 percent of the participants in 1992 via 15.4 percent in 2002 to 13.0 percent in 2011. This increase

Table 2 Exclusion from material resources domain and indicators for women and men 1992, 2002 and 2011. Domain/Indicator 1992 2002 2011 Trend Total (n= 537) Men (n = 210) Women (n= 327) Gender association Total (n= 621) Men (n = 253) Women (n= 368) Gender association Total (n= 904) Men (n = 403) Women (n= 501) Gender association Total sample Men (M) Women (W) Domain (z-score), M (SD) 0.24 (2.21) −0.10 (1.90) 0.46 (2.37) r= .122** 0.14 (2.13) −0.26 (1.71) 0.41 (2.33) r= .152*** −0.09 (1.90) −0.34 (1.60) 0.05 (2.05) r= .099** β=-0.36* M:β=−0.27 W:β=−0.42* Indicators No cash margin (%) 19.2 13.1 23.2 φ = .125** 17.4 10.4 22.2 φ = .152*** 13.2 8.90 15.8 φ = .099** OR = 0.62** M: OR = 0.61 W: OR = 0.62* Financial difficulties (%) 1.50 1.44 1.55 φ=.005 2.47 1.22 3.32 φ=0.67 4.45 2.42 5.68 φ = .077* OR = 3.17** M: OR = 1.68 W: OR = 4.05** Note 1: Due to missing data on individual variables, n will vary across analyses. Note 2: * p < .05; ** p < .01; *** p < .001. Note 3: p for trend is a test for significant difference between 1992 and 2011.

Table 3 Exclusion from social relations and leisure activities domain and indicators for women and men 1992, 2002 and 2011. Domain/Indicator 1992 (n = 537) 2002 (n = 621) 2011 (n = 904) Trend Total (n= 537) Men (n= 210) Women (n= 327) Gender association Total (n= 621) Men (n= 253) Women (n= 368) Gender association Total (n= 904) Men (n= 403) Women (n= 501) Gender association Total sample Men (M) Women (W) Domain (z-score), M (SD) −0.17 (2.15) −0.08 (2.16) −0.23 (2.14) r=−.033 0.03 (2.05) −0.08 (2.11) 0.11 (2.00) r= .045 −0.37 (2.06) −0.38 (2.14) −0.37 (2.01) r= .004 β=-0.25 M:β=−0.34 W: β=−0.20 Indicators Low contact with friends (0–4), M (SD) 1.85 (1.15) 1.90 (1.06) 1.82 (1.20) r=−.035 2.18 (1.19) 2.14 (1.23) 2.21 (1.17) r= .029 2.01 (1.29) 2.00 (1.26) 2.02 (1.31) r= .009 β = 0.14 M: β = 0.07 W: β = 0.19 Low contact with children (0–6), M (SD) 2.73 (2.10) 2.94 (2.14) 2.59 (2.07) r= −.081 2.64 (1.93) 2.69 (1.88) 2.61 (1.96) r=-.021 2.49 (1.82) 2.69 (1.88) 2.37 (1.77) r=-.084* β=−0.25 M:β=−0.26 W: β=−0.24 Low leisure activity (0–20), M (SD) 14.65 (3.72) 14.45 (3.79) 14.78 (3.68) r=.044 14.52 (3.65) 14.17 (3.68) 14.76 (3.62) r= .079 13.95 (3.91) 13.59 (4.03) 14.17 (3.82) r= .072* β=−0.82** M:β=−0.94* W: β=−0.75* Note 1: Due to missing data on individual variables, n will vary across analyses. Note 2: * p < .05; ** p < .01; *** p < .001. Note 3: p for trend is a test for significant difference between 1992 and 2011. Table 4 Exclusion from civic participation domain and indicators for women and men 1992, 2002 and 2011. Domain/Indicator 1992 (n = 537) 2002 (n = 621) 2011 (n = 904) Trend Total (n= 537) Men (n= 210) Women (n= 327) Gender association Total (n= 621) Men (n= 253) Women (n= 368) Gender association Total (n= 904) Men (n= 403) Women (n= 501) Gender association Total sample Men (M) Women (W) Domain (z-score), M (SD) 0.18 (2.31) −0.59 (2.38) 0.09 (2.23) r= .144** −0.16 (2.70) −0.82 (2.48) 0.29 (2.75) r= .203*** −0.43 (2.46) −0.85 (2.39) −0.18 (2.47) r= .132*** β=−0.31 M:β=−0.35 W: β=−0.31 Indicators Non-voting (%) 22.6 16.9 26.3 φ = .109* 24.8 16.7 30.3 φ = .156*** 16.1 13.1 17.9 φ = .064 OR = 0.65** M: OR = 0.71 W: OR = 0.61* No newspaper reading (%) 8.08 8.61 7.74 φ=−.016 15.4 10.7 18.6 φ = .108** 13.0 9.88 14.9 φ = .073* OR = 1.62* M: OR = 1.13 W: OR = 1.97** Low participation in organisations (0–5), M (SD) 3.91 (1.31) 3.94 (1.29) 3.89 (1.33) r=−.019 3.56 (1.52) 3.49 (1.46) 3.62 (1.56) r=.041 3.73 (1.48) 3.69 (1.46) 3.75 (1.49) r= .020 β=−0.23** M:β=−0.34* W: β= −0.17 Unable to deal with public authorities (%) 71.2 56.5 80.8 φ = .263*** 61.7 47.6 71.4 φ = .241*** 56.7 44.8 63.8 φ = .186*** OR = 0.52*** M: OR = 0.60* W: OR = 0.42*** Note 1: Due to missing data on individual variables, n will vary across analyses. Note 2: * p < .05; ** p < .01; *** p < .001. Note 3: p for trend is a test for significant difference between 1992 and 2011.

was primarily driven by a change in the habits of women. There was a significant association between gender and not reading newspapers in 2002 and 2011, with a higher proportion of women than men not reading newspapers in both waves.

There was a decrease in low participation in organisations between 1992 and 2011, with mean scores of 3.91 out of 5 in 1992 and 3.73 in 2011. There were no significant gender associations regarding partici-pation in organisations.

While a majority of participants lacked the ability to deal with public authorities, this proportion decreased over the study period from 71.2 percent in 1992 to 56.7 percent in 2011. Although this reduction in exclusion concerned both men and women, women remained at a substantially higher level of exclusion than men. By 2011, 63.8 percent of the women were unable to deal with public authorities compared to 44.8 percent of the men.

4.4. Exclusion from services

In this domain, data were only available for the latter part of the study period, that is, for 2002 and 2011. As presented inTable 5, the summative scale for exclusion from services indicated a reduction in exclusion between 2002 and 2011, a reduction that was significant for men only. However, there were no significant gender associations for 2002 or 2011.

When considering the individual indicators, at both data collection points approximately 11 percent of participants had not accessed health care despite need, with no significant gender associations.

The proportion of participants who had not accessed dental care despite need decreased from 8.44 percent in 2002 to 4.99 percent in 2011. There were no significant gender associations regarding access to dental care.

5. Discussion

Based on data from a national study covering the years of 1992, 2002 and 2011, our analyses examined trends and gender associations in social exclusion among adults aged 77 years or older living in Sweden. The analyses revealed general reductions in exclusion in the domains of material resources and services. In addition to these overall trends, there were changes in specific indicators of social exclusion between 1992 and 2011. Sometimes trends for indicators were in op-posing directions within the same domain, such as in the domains of material resources and civic participation.

The trend of reduced exclusion in the domain of material resources was mainly driven by the changing situations of women. There was an improvement in cash margin, that is, being able to withdraw SEK 14 000 from a bank account or similar within a week, among older women over the last two decades. At the same time, we found a marked in-crease in the proportion of women reporting financial difficulties, meaning that they had experienced difficulties in managing their cur-rent expenses for food, cur-rent, bills etc. over the last 12 months.

In Sweden, the level of absolute, fixed poverty has decreased over several decades, following a general increase in average income. However, there has been a change in the distribution of income within older age groups resulting in increasing inequalities/relative poverty in this group. In particular, one can note a relative decline for older women in comparison to the population at large (European Commission, 2018). Having cash margin may not necessarily influence people’s everyday consumption patterns and lifestyles, but provides a financial buffer to cope with unforeseeable expenses without depending on others (seeHeap, 2016). The increase in women reporting financial difficulties has been paralleled with an increase in older adults living in “severe and persistent poverty” observed in other countries (e.g. McKee, 2010, p. 20) and also with the increasing at risk of poverty rates among older adults in Sweden.

Despite these trends, both lack of cash margin and financial

Table 5 Exclusion from services domain and indicators for women and men 1992, 2002 and 2011. Domain/Indicator 1992 (n = 537) 2002 (n = 621) 2011 (n = 904) Trend Total (n= 537) Men (n= 210) Women (n= 327) Gender association Total (n= 621) Men (n= 253) Women (n= 368) Gender association Total (n= 904) Men (n= 403) Women (n= 501) Gender association Total sample Men (M) Women (W) Domain (z-score), M (SD) — 0.10 (1.63) 0.12 (1.63) 0.09 (1.63) r=−.009 −0.04 (1.43) −.14 (1.29) .02 (1.50) r= .054 β=−0.16* M:β=−0.29* W: β=−0.09 Indicators Non-use of health care despite need (%) — 10.7 11.1 10.4 φ=−.011 10.7 8.38 12.2 φ = .059 OR = 0.98 M: OR = 0.69 W: OR = 1.18 Non-use of dental care despite need (%) — 8.44 8.43 8.45 φ=.000 4.99 4.18 5.48 φ = .029 OR = 0.54* M: OR = 0.46* W: OR = 0.60 Note 1: Due to missing data on individual variables, n will vary across analyses. Note 2: * p < .05; ** p < .01; *** p < .001. Note 3: p for trend is a test for significant difference between 2002 and 2011. Note 4: — = not measured that year.

difficulties were more common among women than men, which is in line with previous research (Becker & Boreham, 2009; Kneale, 2012) and current European data as described earlier (European Commission, 2018). There is thus ample evidence of a higher prevalence of poverty among women than men and research has also suggested that women are more likely to experience unfavourable outcomes of poverty, such as mental health problems (Kwan & Walsh, 2018). At a European level, it has been noted that while “social security is vital for those who can work […], social protection is inevitable for those who cannot work” and that reductions in social policy programmes place such groups at greater risk of poverty and social exclusion (ESPAS, 2019, p. 8).

There was an overall reduction also in the domain of exclusion from services, driven by a decrease in the proportion of older adults not accessing dental care despite need. This can be compared with a study from England that, on the contrary, found that difficulties in accessing dentists and general practitioners increased between 2002 and 2008 (Kneale, 2012). Whereas health care in Sweden is largely paid for via taxes, dental care is considerably more expensive for the individual patient. Issues on exclusion from health and social care services due to lack of availability and affordability has previously been reported in a study from the UK (Patsios, 2018).

Although there was no statistically significant change at domain level regarding civic participation, there were statistically significant reductions on most indicators in this domain, that is, there was an in-crease in the proportion of people who had voted in the previous election, participated in organisations, and had the ability to deal with public authorities. The exception to this pattern was the observed de-crease in the proportion of older adults who were reading newspapers. The trend in voting participation follows the general trend in Sweden, where there has been a decrease in voting up to 2002, after which there has been an increase (Statistics Sweden, 2019). The oldest age group included in our study had lower levels of voting than the general population in Sweden (for similar patterns in the UK, seeBailey, Fahmy, & Bradshaw, 2018), but this gap has reduced due to increasing levels of voting in the older age groups (Öhrvall, 2012; Statistics Sweden, 2019). Despite the increased voting participation, it should be noted that we found that non-voting was higher among older women than men, which is in line with data from Statistics Sweden and unique to the oldest age group, that is, people aged 75 years or older (Öhrvall, 2012; Statistics Sweden, 2015). While the voting gender gap in the general population diminished in many Western countries, including Sweden, in the 1980s and 1990s (Norris, 2002;Öhrvall, 2012), it still remains in some age groups, as found in our study and research cov-ering other Western countries (seeBhatti & Hansen, 2012). Tradition-ally, the gender gap has been explained by a higher interest in and knowledge about politics in men than women and by higher voting participation in more privileged groups (Kostelka, Blais, & Gidengil, 2019; Öhrvall, 2012). Age-specific explanations are that living with somebody and good health increase voting (Bhatti & Hansen, 2012).

Our finding of an increased participation in organisations can be contrasted with that of a Swedish population study, which found a stable pattern in participation in organisations over the period of 1992 and 2009 (Jegermalm & Grassman, 2012). The population study, however, included people aged 16–74 years and asked about vo-lunteering in a broader range of organisations than those covered in our study. It has previously been shown that participation in organisations is associated with education (Statistics Sweden, 2011). The increase in participation in organisations in our study may, thus, reflect higher levels of education in more recent birth cohorts, particularly among women (Statistics Sweden, 2018). Higher levels of education may also partly explain the increased ability to deal with public authorities in our study.

Men were generally in a better position than women regarding ex-clusion from civic participation, which echoes findings in previous re-search (Del Bono et al., 2007;Kneale, 2012). A higher proportion of men than women had the ability to deal with public authorities, voted

in the previous election (significant in 1992 and 2002), and read newspapers (significant in 2002 and 2011). Civic inclusion, not the least the ability to deal with public authorities, is crucial when navi-gating a care system in which the user is expected to make informed choices about care providers (Meinow, Parker, & Thorslund, 2011; Szebehely & Trydegård, 2012). This is particularly problematic for women, as they rely on formal care in old age to a greater degree than men (Dahlberg, Berndt, Lennartsson, & Schön, 2018).

Regarding exclusion from social relations and leisure activity, the most evident finding was an increase in participation in leisure activ-ities over the study period. It has previously been reported that women have a lower risk of exclusion from social relations (Del Bono et al., 2007;Kneale, 2012). This was confirmed in our study with regard to contact with children in 2011. While we did not identify any gender associations regarding contact with friends, our study shows a higher level of engagement in leisure activities among men.

In summary, this study showed statistically significant reductions in exclusion in two out of four domains, as well as reductions in exclusion indicators across all domains. One factor that may influence the re-duction of exclusion across several domains is the increase in functional ability among the oldest old (Fors, Lennartsson, Agahi, Parker, & Thorslund, 2013), as this enables individuals to, e.g., be active in civic and leisure participation. The higher levels of exclusion among older women may be a reflection of the poorer health of women than men (Schön & Parker, 2009) and a higher proportion of women than men living in institutions (NBHW, 2019).

5.1. Strengths and limitations

Despite being identified as a risk group, older adults have received relatively little attention in social exclusion research and there is a particular dearth of studies on trends in social exclusion, at least outside the UK. Instead, research has mainly focused on people of working age and on topics such as unemployment, immigration and mental health. Our study therefore contributes substantially to knowledge on trends in exclusion of older adults. In addition, this study contributes with its focus on gender, as separate analyses of older women and older men are rare in research on social exclusion in older adults. Given that some trends in exclusion at domain level concealed opposing trends at in-dicator level, a further contribution of our study was its detailed ap-proach demonstrating that analyses of exclusion purely at domain level may lead to erroneous conclusions.

One strength of this study is its high response rates, the inclusion of institutionalized persons and the use of proxy informants for people unable to be interviewed directly, thus ensuring that the SWEOLD sample is representative of older adults in Sweden in each interview wave (Lennartsson et al., 2014). A high response rate is a factor of particular importance if one wishes to validly analyse social exclusion processes: those who are experiencing exclusion of one form or another are most likely to be poorly represented in studies with low response rates. Another strength of this study is that, in contrast to community studies of social exclusion, it is based on a national sample. In addition, this study had several waves of data collection and included a wide range of measures of exclusion, enabling the analysis of trends in social exclusion over two decades. One weakness of this study is that the analyses are based on a study not originally designed to examine social exclusion. This limited the range and quality of indicators of social exclusion that could be used. For example, the domain of exclusion from services only contains indicators of access to health and dental care, and only for the latter part of the study period. Although access to health and dental care is important to older adults, there were no in-dicators on access to services such as shops, banks or transport. 5.2. Conclusions and policy implications

novel analyses on trends and its focus on gender. Although our study found a reduction in exclusion in most domains, there are still older adults who are at risk of social exclusion. There are individuals who do not have any financial buffer, a small but increasing group who have difficulties managing their current expenses for food, rent, bills etc. There are also individuals with low levels of social relations and social activities, and a significant group that does not access health and dental care despite need. In addition, individuals are excluded from the do-main of civic participation, where the majority of the participants do not have the ability to deal with public authorities. With regard to gender, women were found to have higher levels of exclusion from material resources and civic participation.

The findings that there are still people at risk of exclusion challenge the notion of a universal welfare state where care and services are provided on equal terms to the whole population and where equal living conditions are promoted (Hälso-och sjukvårdslagen, 1982; Socialtjänstlag, 2001). Thus, this study underlines the continuing im-portance of a well-developed welfare and social security system, where people are guaranteed a certain material standard, where not only medical but also social needs are met, and where care is affordable and easily accessible for those in need.

Author contributions

Development, reviewing, revision and final approval of manuscript: all authors. Data analyses: LD and CL. Original draft: LD. Funding ac-quisition: LD, JF, CL, KM. Project management: LD.

Funding

This research was supported by the Swedish Research Council for Health, Working Life and Welfare (Forte), reg. no. 2017-00668 and the Swedish Research Council (VR), reg. no. 2018-01922.

Declaration of Competing Interests

The authors declare that they have no competing interests. Appendix A

In indirect interviews in 1992, questions were not asked on voting in previous election, participation in organisations, and ability to write a letter of appeal. For people living in an institution in 1992, questions were not asked on social contacts with friends, leisure activities, and reading newspapers. The question on financial difficulties was not asked of people living in an institution or indirectly interviewed in 1992. For these individuals, values were imputed for these questions as follows. For each wave, participants were grouped into four categories, based on combining their level of activities of daily living (no limita-tions vs. one or more limitalimita-tions in eating, toilet visits, hair washing, un/dressing and getting in/out of bed) with their gender (male vs. fe-male). Within these groups the samples of 2002 and 2011 were com-bined and ratios computed for those who received direct vs. indirect interviews and those living in ordinary housing vs. in institutions for the values on the questions for which data were missing in 1992. Those ratios were then applied to the 1992 participants who received direct interviews or lived in ordinary housing to compute the values on the missing questions for participants receiving indirect interviews or living in institutions, respectively. For those both receiving indirect interviews and living in institutions an average of the two ratios was applied. For dichotomous variables, the imputed values were rounded off so that proportions could be presented.

References

Bailey, N., Fahmy, E., & Bradshaw, J. (2018). The multidimensional analysis of social exclusion. In G. Bramley, & N. Bailey (Eds.). Poverty and social exclusion in the UK.

Volume 2 - The dimensions of disadvantage (pp. 311–342). Bristol: Policy Press.

Becker, E., & Boreham, R. (2009). Understanding the risks of social exclusion across the life

course: Older age. London: Cabinet Office and National Centre for Social Research. Bhatti, Y., & Hansen, K. M. (2012). Retiring from voting: Turnout among senior voters.

Journal of Elections, Public Opinion and Parties, 22(4), 479–500.https://doi.org/10. 1080/17457289.2012.720577.

Burchardt, T., Le Grand, J., & Piachaud, D. (2002). Degrees of exclusion: Developing a dynamic, multidimensional measure. In J. Hills, J. Le Grand, & D. Piachaud (Eds.).

Understanding social exclusion (pp. 30–43). Oxford: Oxford University Press. Council of European Union (2004). Joint report by the Commission and the Council on social

inclusion. Retrieved from.http://europa.eu.int/comm/employment_social/soc-prot/ soc-incl/final_joint_inclusion_report_2003_en.pdf.

Dahlberg, L., & McKee, K. J. (2018). Social exclusion and well-being among older adults in rural and urban areas. Archives of Gerontology and Geriatrics, 79, 176–184.https:// doi.org/10.1016/j.archger.2018.08.007.

Dahlberg, L., Agahi, N., & Lennartsson, C. (2018). Lonelier than ever? Loneliness of older people over two decades. Archives of Gerontology and Geriatrics, 75, 96–103.https:// doi.org/10.1016/j.archger.2017.11.004.

Dahlberg, L., Berndt, H., Lennartsson, C., & Schön, P. (2018). Receipt of formal and in-formal help with specific care tasks among older people living in their own home. National trends over two decades. Social Policy & Administration, 52(1), 91–110.

https://doi.org/10.1111/spol.12295.

Del Bono, E., Sala, E., Hancock, R., Gunell, C., & Parisi, L. (2007). Gender, older people and

social exclusion. A gendered review and seconday analusis of the data. Retrieved from. ESPAS (2019). Global trends to 2030: New ways out of poverty and social exclusion.

Retrieved 27 June 2019 from European Strategy and Policy Analysis System website:.

https://espas.secure.europarl.europa.eu/orbis/sites/default/files/generated/ document/en/Ideas%20paper_poverty%20exclusion.pdf.

Esping-Andersen, G. (2016). Families in the 21st century. Stockholm: SNS Förlag. European Commission (2017). European pillar of social rights. Retrieved from.https://ec.

europa.eu/commission/sites/beta-political/files/social-summit-european-pillar-social-rights-booklet_en.pdf.

European Commission (2018). Pension adequacy report 2018.\ Brussels: European Commission.

Fors, S., Lennartsson, C., Agahi, N., Parker, M. G., & Thorslund, M. (2013). Äldre har fått fler hälsoproblem, men klarar vardagen bättre [More older people have health pro-blems, but manage their daily life better]. Läkartidningen, 110, CA33.

Hälso-och sjukvårdslagen 1982:763 [Swedish Health Care Act] (1982).

Heap, J. (2016). Living conditions in old age: Coexisting disadvantages across life domains. Stockholm: Stockholm University.

Heap, J., Fors, S., & Lennartsson, C. (2017). Coexisting disadvantages in later life: Demographic and socio-economic inequalities. Journal of Population Ageing, 10(3), 247–267.https://doi.org/10.1007/s12062-016-9158-y.

Jegermalm, M., & Grassman, E. J. (2012). Helpful citizens and caring families: Patterns of informal help and caregiving in Sweden in a 17-year perspective. International Journal of Social Welfare, 21(4), 422–432.https://doi.org/10.1111/j.1468-2397.2011. 00839.x.

Kelfve, S., Thorslund, M., & Lennartsson, C. (2013). Sampling and non-response bias on health-outcomes in surveys of the oldest old. European Journal of Ageing, 10(3), 237–245.https://doi.org/10.1007/s10433-013-0275-7.

Kneale, D. (2012). Is social exclusion still important for older people? London: International Longevity Centre - UK.

Kostelka, F., Blais, A., & Gidengil, E. (2019). Has the gender gap in voter turnout really disappeared? West European Politics, 42(3), 437–463.https://doi.org/10.1080/ 01402382.2018.1504486.

Kwan, C., & Walsh, C. A. (2018). Old age poverty: A scoping review of the literature. Cogent Social Sciences, 4(1),https://doi.org/10.1080/23311886.2018.1478479. Lee, J., & Cagle, J. G. (2018). Social exclusion factors influencing life satisfaction among

older adults. Journal of Poverty and Social Justice, 26(1), 35–50.https://doi.org/10. 1332/175982717x15127351091521.

Lennartsson, C., & Lundberg, O. (2007). ’What’s marital status got to do with it?’: Gender inequalities in economic resources, health and functional abilities among older adults. In J. Fritzell, & O. Lundberg (Eds.). Health inequalities and welfare resources.

Continuity and change in Sweden (pp. 179–198). Bristol: The Policy Press. Lennartsson, C., Agahi, N., Hols-Salén, L., Kelfve, S., Kåreholt, I., Lundberg, O., Parker, M.

G., & Thorslund, M. (2014). Data resource profile: The swedish panel study of living conditions of the oldest old (SWEOLD). International Journal of Epidemiology, 43(3), 731–738.https://doi.org/10.1093/ije/dyu057.

Levitas, R., Pantazis, C., Fahmy, E., Gordon, D., Lloyd, E., & Patsios, D. (2007). The

multi-dimensional analysis of social exclusion. London: Social Exclusion Task Force.

McKee, S. (2010). The forgotten age: Understanding poverty and sociale xclusion in later life. London: Center for Social Justice.

Meinow, B., Parker, M. G., & Thorslund, M. (2011). Consumers of eldercare in Sweden: The semblance of choice. Social Science & Medicine, 73(9), 1285–1289.https://doi. org/10.1016/j.socscimed.2011.08.015.

Miranti, R., & Yu, P. (2015). Why social exclusion persists among older people in Australia. Social Inclusion, 3(4), 112–126.https://doi.org/10.17645/si.v3i4.214.

NBHW (2019). Vård och omsorg om äldre. Lägesrapport 2019 [Health and social care for

older people. Progress report 2019].\ Stockholm: National Board for Health and

Welfare.

(Eds.). Voter turnout from 1945 to 2000: A global report on political participation (pp. 95–104). Stockholm: IDEA.

Ogg, J. (2005). Social exclusion and insecurity among older Europeans: The influence of welfare regimes. Ageing and Society, 25, 69–90.https://doi.org/10.1017/ Soi44686X04002788.

Öhrvall, R. (2012). Svenskt valdeltagande under hundra år [One hundred years of Swedish

voter turnout]. Stockholm: Statistics Sweden.

Patsios, D. (2018). Improvement for some: Poverty and social exclusion among older people and pensioners. In E. Dermott, & G. Main (Eds.). Poverty and social exclusion in

the UK. Volume 1 - The nature and extent of the problem (pp. 61–93). Bristol: Policy

Press.

Peeters, H., Debels, A., & Verpoorten, R. (2013). Excluding institutionalized elderly from surveys: Consequences for income and poverty statistics. Social Indicators Research, 110(2), 751–769.https://doi.org/10.1007/s11205-011-9957-8.

Percy-Smith, J. (2000). Introduction: The contours of social exclusion. In J. Percy-Smith (Ed.). Policy responses to social exclusion: Towards inclusion? (pp. 1–21). Buckingham: Open University Press.

Popay, J., Escorel, S., Hernández, M., Johnston, H., Mathieson, J., & Rispel, L. (2008). Social exclusion. Final report to the WHO commission on social determinants of health from the social exclusion knowledge network. Retrieved from.http://www.who.int/ social_determinants/knowledge_networks/final_reports/sekn_final%20report_ 042008.pdf.

Sacker, A., Ross, A., MacLeod, C. A., Netuveli, G., & Windle, G. (2017). Health and social exclusion in older age: Evidence from understanding Society, the UK household longitudinal study. Journal of Epidemiology and Community Health, 71(7), 681–690.

https://doi.org/10.1136/jech-2016-208037.

Saito, M., Kondo, N., Kondo, K., Ojima, T., & Hirai, H. (2012). Gender differences on the impacts of social exclusion on mortality among older Japanese: AGES cohort study. Social Science & Medicine, 75(5), 940–945.https://doi.org/10.1016/j.socscimed. 2012.04.006.

Scharf, T., Phillipson, C., & Smith, A. E. (2005). Social exclusion of older people in de-prived urban communities of England. European Journal of Ageing, 2(2), 76–87.

https://doi.org/10.1007/s10433-005-0025-6.

Schön, P., & Parker, M. G. (2009). Sex differences in health in 1992 and 2002 among very old Swedes. Journal of Population Ageing, 1(2), 107–123.https://doi.org/10.1007/

s12062-009-9009-1.

Silver, H., & Miller, S. M. (2003). Social exclusion. The European approach to social disadvantage. Indicators, 2(2), 5–21.https://doi.org/10.1080/15357449.2003. 11069166.

Socialtjänstlag (2001:453) (2001).

Statistics Sweden (2011). Medborgerliga aktiviteter 2008–2009 [Civic activities 2008-2009]. Stockholm: National Statistics.

Statistics Sweden (2015). Allmänna val, valdeltagandeundersökningen [General elections, participation survey]. Retrieved 22.08.2019 from.https://www.scb.se/hitta-statistik/

statistik-efter-amne/demokrati/allmanna-val/allmanna-val-valdeltagandeundersokningen/.

Statistics Sweden (2018). Befolkningens utbildning 2018 [Educational attainment of the po-pulation in 2018]. Retrieved 28 June 2019 from.https://www.scb.se/contentassets/ c4f928a71e934437b1320762371c35a1/uf0506_2018a01_sm_uf37sm1902.pdf. Statistics Sweden (2019). Valdeltagande i Sverige [Voting participation in Sweden]. Retrieved

28 June 2019 from.https://www.scb.se/hitta-statistik/sverige-i-siffror/ manniskorna-i-sverige/valdeltagande-i-sverige/.

Szebehely, M., & Trydegård, G. B. (2012). Home care for older people in Sweden: A universal model in transition. Health & Social Care in the Community, 20(3), 300–309.

https://doi.org/10.1111/j.1365-2524.2011.01046.x.

Tinson, A., Aurton, C., Baker, K., Born, T. B., Aldridge, H., & Kenway, P. (2016).

Monitoring poverty and social exclusion 2016. York: Joseph Rowntree Foundation.

United Nations (2010). Analysing and measuring social inclusion in a global context. New York: United Nations.

van der Heijden, G., Donders, A. R. T., Stijnen, T., & Moons, K. G. M. (2006). Imputation of missing values is superior to complete case analysis and the missing-indicator method in multivariable diagnostic research: A clinical example. Journal of Clinical Epidemiology, 59(10), 1102–1109.https://doi.org/10.1016/j.jclinepi.2006.01.015. Van Regenmortel, S., De Donder, L., Dury, S., Smetcoren, A.-S., De Witte, N., & Verté, D.

(2016). Social exclusion in later life: A systematic review of the literature. Journal of Population Ageing, 9(4), 315–344.https://doi.org/10.1007/s12062-016-9145-3. Walsh, K., Scharf, T., & Keating, N. (2017). Social exclusion of older persons: A scoping

review and conceptual framework. European Journal of Ageing, 14(1), 81–98.https:// doi.org/10.1007/s10433-016-0398-8.