Introduction

Population ageing has significant social and financial implications for families and governments, and this has led to considerable interest in the future support needs of the older population and the balance between support provided directly by families and that obtained from state or market sources. Recent commentators have distinguished ‘weak’ from ‘strong’ family societies, arguing that older people in less family-oriented societies receive less support from family members than those in countries with strong family ties (e.g. Southern Europe) (Reher 1998). However, as membership of the European Union (EU) may lead to greater convergence in social policies concerning support for older people, in addition to greater similarity in economic and labour market spheres, it is possible that this will prompt a trend towards convergence in family-related behaviours, including provision of support to older relatives. Some analysts have argued that convergence will never happen, and that intergenerational coresidence and family provision of care will always be much less prevalent in north-western Europe (and North America) than in southern Europe (Reher 1998). Others have pointed out that in many respects coresidence patterns in Scandinavian and northern European countries in the past were very similar to those currently prevalent in southern European countries today, suggesting a future trend towards convergence (Sundström 1994). In this study we explore the north-south divide in various dimensions associated with support for older people among the following participant countries in a European Scientific Foundation (ESF) network, ‘Family Support for Older People: Determinants and Consequences’ (FAMSUP): Austria, Belgium, Germany, Italy, The Netherlands, Portugal, Sweden, and the United Kingdom. Using principal components analysis (PCA) and cluster analysis we investigate patterns across countries in four dimensions designed to be indicative of the balance between family and formally provided resources for older people and the socio-economic, demographic and policy contexts in which these are provided. The aim of the study is to critically examine whether the traditional typologies that have been used to group countries according to family behaviour accurately capture differences between countries.

Background

Some researchers have based their categorisations of European countries on modified versions of Esping-Andersen’s (1990) immensely influential typology of welfare states (Daatland and Herlofson 2003; Iacovou 2002). Esping-Andersen distinguished groups of countries on the basis of broad conformity to the following welfare regime types: (a) social democratic countries, characterised by universal social insurance programmes (e.g. Sweden, Denmark, The Netherlands), (b) liberal societies largely characterised by private (rather than public) insurance programmes (e.g. the United States) and (c) conservative societies with a greater emphasis on occupation based social insurance schemes (e.g. Italy, France, Austria, Germany and Belgium). More recent studies have modified Esping-Andersen’s original typology to group the other southern European countries (Greece, Portugal, Spain), including Italy, into a separate ‘Mediterranean welfare’ category (Iacovou 2002).

Other researchers have derived typologies on historical studies of family systems in Europe rather than welfare regimes, but with the same result: delineation of southern European countries as more family oriented than northern ones (Reher 1998). In some senses these approaches—emphasis on family regime or emphasis on type of welfare state—may be regarded as two sides of the same coin as they clearly may have interactive effects, and it is difficult to determine whether families provide help because of a lack of other alternatives, or whether states provide services (including financial resources) because of a lack of family support or strong preferences for formal rather than family care. However, not all analysts concur with this demarcation, and while some studies suggest a clear north-south divide in extent of family care for older relatives, which supports the idea of a north-south divide in familistic orientation (Daatland and Herlofson 2003; Iacovou 2002; Reher 1998), others challenge this view (Alber and Köhler 2004). A major difficulty for all studies, including the present one, is the lack of cross-national comparative information on important indicators of family support for older people, such as actual hours of care provided to those in need. Analysts have therefore had to rely, as do we, on imperfect measures of family support such as intergenerational coresidence and extent of contact between older parents and their children.

Living arrangements

A very extensive literature has documented variations over time and space in the living arrangements of older Europeans (Grundy 1996; Palloni 2001; Pampel 1992; Sundström 1994; Tomassini et al. 2004; Wolf 1995). While coresidence does not in itself imply provision of support by adult children to elderly parents (and in some cases may in fact be indicative of support provided by elderly parents to adult children), coresidence may be regarded as a vehicle for the provision of social and financial support and thus retains some value as an indicator of family solidarity. Studies from the United States, for example, have shown that children who live at home provide greater assistance to their parents (both financially and with domestic tasks) than non-coresident children (Hoyert 1991; Ward et al. 1992).

There has been a well-documented rise in the proportion of older people living alone throughout Europe, as well as in other industrialised countries, since the Second World War, and a correspondingly large decline in intergenerational coresidence (Palloni 2001; Pampel 1992; Sundström 1994; Tomassini et al. 2004). However, in the 1990s some countries showed a reversal of earlier trends toward solitary living among older people (Tomassini et al. 2004). Nevertheless, while trends in industrialised countries show similar substantial declines, intergenerational coresidence continues to vary and is higher in southern Europe than in other industrialised countries (Grundy 1996; Palloni 2001; Sundström 1994).

Contacts with family members

Coresidence is only one, and perhaps a decreasingly important, indicator of family resources. Data on contacts and support exchanges between non-coresidents is much sparser than data on household composition, but results from several studies suggest that intergenerational support and contact is high, although less is known about trends over time. Grundy and Shelton (2001) used data from three British surveys conducted in 1986, 1995 and 1999 to examine trends over time in contacts between adult children aged 22–54 years and their parents and, contrary to popular belief, found no indication of a trend towards reduced contact, but few comparable results from other countries are available (Grundy and Shelton 2001). With regard to differences between countries, international comparative data available, for example, through Eurobarometer surveys and relevant International Social Survey Programme modules (both with rather small sample sizes), suggest more frequent contact between family members in southern than in northern European countries (Höllinger and Haller 1990; Lowenstein et al. 2003; Sundström 1994).

Care from family members

‘Everyday’ help and emotional support is obviously very important, but from a policy perspective the availability of care for elderly people with high support needs is probably more so, if only because friends, neighbours and volunteers may be used to provide support and help with instrumental activities of daily living (IADL, such as shopping) but rarely provide much assistance with activities of daily living (ADLs, such as bathing). Deficits in family care in these areas may therefore have greater implications for the provision of formal services.

As with intergenerational contacts, there are few cross-national studies looking at variations in provision or receipt of care among older people (Alber and Köhler 2004; Sundström 1994; Walker and Maltby 1997). Results from the 1999 Eurobarometer survey show that provision of coresident care to older relatives is higher in southern than in north-western European countries (Alber and Köhler 2004). For example, 1% of Dutch respondents provide coresident care for an older relative in comparison with about 7% of Greeks (Alber and Köhler 2004). However, when total care provision is taken into account (i.e. caregiving both within and outside the household), Dutch adults show the highest prevalence of informal care activities among the EU member states and southern European countries the lowest (e.g. Italy, Portugal and Spain; Alber and Köhler 2004).

With regards to the recept of care, in most industrial societies only a small portion of care received by older people is provided by public services; the great majority of the care continues to be provided by family members including other older people, particularly spouses (Sundström 1994; Walker and Maltby 1997). Formal carers rarely constitute the majority of carers; only in Denmark (which has the most extensive home care services) do a large proportion of older individuals receive regular help or assistance from public social services (Walker and Maltby 1997).

Explanations for differences

Cultural values

The notion of the familistic culture has been used in the past to explain strong family ties in southern Europe (although, of course, this argument is to an extent tautological; Banfield 1958; Reher 1998). In a familistic society personal utility and family utility are seen as the same: the structure of the family and the relationships among family members are influenced by the strong ties that link them together. In northern Europe, it is argued, there is a more strongly individualistic culture, involving more emphasis on voluntaristic relationships and greater stress on personal autonomy (Phillipson et al. 2001). Evidence on attitudes lends some support to the idea of different cultural values in northern and southern Europe. Analyses of Eurobarometer data show little change in attitudes toward the role of the family in elder care between 1992 and 1999 and suggest that national differences in attitudes have remained (Tomassini et al. 2004). For example, Swedish and Austrian adults were more likely to consider the option of a residential or nursing home for frail older people (13% of Swedes compared with 7% of Italians). With regards to the potential needs of respondents’ own frail parents, 42% of Italians aged 40–64 years thought it would be best for an elderly father or mother who lived alone and could no longer manage on his or her own to move in with them or one of their siblings, compared with 30% of British and 9% of Swedish respondents (Tomassini et al. 2004).

It has been suggested that religious affiliation and religiosity are related to cultural norms and values regarding family responsibility for elder care, with family support for frail elderly people being more prevalent in Catholic nations (Alber and Köhler 2004; Pampel 1992). Nevertheless, in a cross-national comparative study of solitary living among older people in Europe, national differences in living arrangements remained even when religious beliefs and attitudes, in addition to other individual-level explanatory factors, were controlled for (Pampel 1992).

Demographic trends

Demographic influences on the availability of family support are clearly of major potential importance. Trends in nuptiality, age differences between spouses, age at childbearing and the parity distribution of cohorts, as well as overall levels of fertility and mortality influence the composition of kin networks and the proportions with or without specified kin at different ages. As many of these parameters vary between developed countries, there are also quite large differences in the availability of particular kin. Differences between countries and time periods in the proportions of ever-married older people are also marked. For example, in Ireland in the 1980s nearly one-quarter of older people population had never been married, compared with only 5% in Bulgaria (Grundy 1996). In many European countries (and also North America) cohorts born in the inter-war and immediately post-World War II period had much lower rates of celibacy than either preceding or succeeding generations (Grundy 1996). This means that short-term prospects for the support of older people are in many instances favourable (as spouses are the main providers of support). However, longer term prospects are much less so, particularly as those born since the mid-1950s have experienced high rates of divorce, as well as including relatively high proportions of never-married persons.

Many studies have shown that older people with more children are less likely to live alone than those with fewer children (Wolf 1994). In addition, health also has an important impact on receipt of help among older people. Improvements in life expectancies have been well documented for many countries in Europe. Although life expectancy at age 65 years is a widely criticised measure of health, it nevertheless provides an easily available comparative indicator of the health status of older people.

Policy and socio-economic context

Both the wider socio-economic contexts and individual level socio-economic circumstances are likely to influence needs for, and reliance on, family support. For example, in some poorer countries with inadequate or low levels of pension provision, the proportion of older people who need material assistance, including coresidence, from family members is higher while in countries with generous pension provision and welfare services, needs for assistance may be met without extensive recourse to family members. At an individual level higher levels of education and higher incomes have often been associated with higher levels of residential independence, although there may be some country differences. In one study home ownership and high education (two proxies for higher social status) were found to be strongly negatively associated with coresidence in Britain but not in Italy (Glaser and Tomassini 2000). Variations in female labour-force participation may also be a relevant factor given arguments that increasing commitments outside the home may conflict with women’s ability and willingness to coreside with elderly relatives in need of care (Dooghe 1992). However, most studies have shown that individuals, especially women, appear to add elder care and other family responsibilities to their portfolio of daily tasks without cutting back substantially on paid work or other competing obligations (Evandrou and Glaser 2002; Spiess and Schneider 2003).

Policy environments vary greatly across Europe, reflecting differences in expenditure on pensions and health care, and in the provision of long-term care (here used to refer to both institutional and domiciliary care). For example, public expenditure on pensions varies from 4% of gross domestic product (GDP) in Ireland to 15% in Italy (Kubitza 2004). There is also considerable variation in the proportion of GDP devoted to public health care spending, which ranges from 5% of GDP in Greece to 8% in Germany (OECD health data 2002). The availability of long-term care provision varies greatly across Europe, with northern European countries (e.g. Denmark and Sweden) providing more extensive public services for older people requiring care, and having a higher proportion of older people resident in institutions of various kinds, than the countries of southern Europe (e.g. Portugal and Italy; Pacolet et al. 1999).

Given similarities in socio-demographic trends and differences in the balance between state- and family-provided resources for older people and in cultural traditions, comparative analyses between north-western and southern European countries have the potential for aiding our understanding of family support in later life.

Data and methods

Data

A FAMSUP data file was created by collating information drawn from nationally representative surveys, censuses, and official publications provided by FAMSUP members; Table 1 lists the variables in the data file. Additionally, personal contacts with national offices of statistics were made in order to gather data from national surveys or censuses not electronically available. Sources for data used in this study were the following:

Table 1 Principal components analysis (PCA) of selected dimensions associated with support for older people: factor loadings
Full size table

Plan of analysis

To better understand the underlying structure of European countries when several indicators related to the support of older people are to be analysed we used PCA and cluster analysis (Bouroche and Saporta 1980). With PCA we were able to identify underlying ‘factors’ (reflecting correlations among our set of indicators) capturing various dimensions associated with support for older people. With cluster analysis we were further able to analyse similarities and differences between the European countries in terms of dimensions related to support for older individuals by measuring the distance between indicators. The indicators used in both analyses reflect important elements associated with family support: attitudinal, demographic and policy and socio-economic factors. Using these two different approaches we investigated differences and similarities between groups of countries with respect to these indicators.

Measures

Family contact and care variables

The family contact and care variables used in this analysis represent elements of family support, although it is acknowledged that they are only partial indicators. We considered the following, the percentage of: (a) women aged 65 years or over living alone in 2000/2001 (FAMSUP database, derived from national censuses or survey data); (b) individuals aged 60 years or over with daily family contacts in 1992 (‘special’ Eurobarometer Survey, authors’ calculation, from the question ‘How often do you see your family these days?’ with response categories every day, two or more times a week, once a fortnight, once a month, less often, never, don’t know); and (c) persons aged 60 years or over receiving family care only in 1992 (‘special’ Eurobarometer Survey, authors’ calculation, from the question, ‘Does anyone give you regular help or assistance, with personal care or household tasks, because you find it difficult to do these by yourself?’ If the answer was yes, respondents were asked ‘Who gives you regular help or assistance?’ and response categories included spouse, children, other relatives, friends, private paid help, neighbour, public social service, someone from voluntary organisation or charity group, don’t know; assistance was considered both from within and outside the household).

Information on solitary living for most European countries is readily available from published statistical sources based on household surveys and/or census data. Obtaining information on coresidence with children, however, is more complicated given considerable cross-national differences in how a ‘child’ is defined in published statistical material (e.g. in the United Kingdom a child is considered to be coresident only if the child has never married regardless of age, whereas in Italy the definition encompasses all children regardless of age or marital status).

Attitudinal and cultural variables

We used here attitudinal data, as specified below, and also data on religiosity as it has been argued that religion and religiosity may be associated with perceptions of family responsibilities. We considered the following attitudinal and cultural variables: (a) the percentage of persons aged 45–64 years that think that parents should move to a nursing home if they become frail in 1998 (Eurobarometer Survey, authors’ calculation, from the question ‘Let’s suppose you had an elderly father or mother who lived alone. What do you think would be best if this parent could no longer manage to live on his/her own?’ with responses being myself or one of my brothers or sisters should invite my father or mother to live with one of us, I or one of my brothers or sisters should move in with my father or mother, one should move closer to the other, my father or mother should move into an old people’s home or a nursing home, my father or mother should stay at home, and receive visits there, as well as appropriate health care and services; and (b) religiosity median score from a scale of 0–10 in 2002 among persons aged 65 years or over (European Social Survey; authors’ calculation from the question ‘Regardless of whether you belong to a particular religion, how religious would you say you are?’ response categories ranged from 0 not at all religious to 10 very religious).

Demographic variables

As noted above, the demographic context of a country has an important impact on family support. First, there are known to be variations between European countries in the proportion of older people lacking close relatives (e.g. living children). Second, factors such as age composition may influence state provision of resources for older people. We used the following demographic indicators: (a) percentage of persons aged 65 years or over in 1970 and 2001 (Eurostat New Cronos database); (b) female life expectancy at age 65 years in 1970 and 2000 (Eurostat New Cronos database); (c) percentage childless among cohorts born around 1930 (Eurostat New Cronos database; Iacovou 2000; Rowland 1998); (d) total fertility rate for cohorts born around 1930 (Eurostat New Cronos database and Council of Europe); (e) percentage of married women aged 65 years or over in 1970 and 2000/2001 (FAMSUP database, Census data, United States International Database); and (f) percentage of divorced women aged 65 years or over in 2000/2001 (FAMSUP database, Census data).

Socio-economic and policy-related contextual variables

These include indicators both of the socio-economic status of older people and indicators relating to state provision of services and other supports for older individuals. We included the following variables: (a) subjective income measure, percentage of individuals aged 65 years or over feeling comfortable or coping on present income in 2002 (European Social Survey; authors’ calculation from the question ‘Which of the descriptions on this card comes closest to how you feel about your household’s income nowadays?’ with response categories living comfortably on present income, coping on present income, finding it difficult on present income, finding it very difficult on present income); (b) percentage of women aged 45–60 years in the labour force in 2000 (International Labour Organisation); (c) country level expenditure on pensions as a percentage of GDP in 2000 (Eurostat 2004); (d) country level public expenditure on health care as a percentage of GDP in 2000 (European Commission 2003); (e) percentage of persons aged 65 years in institution in 2000/2001 (FAMSUP database, derived from census or survey data); and (f) percentage of older people receiving public care only in 1992 (Eurobarometer Survey; authors’ calculation, from question on regular help and assistance discussed above with regards to family care only).

Results

We used PCA to map these dimensions in FAMSUP countries and also included Denmark, France, Greece, Ireland and Spain (results are presented in Table 1 and Figs. 1, 2). PCA yielded three factors which together explained 75% of the total variance: a good synthesis of the matrix containing 13 countries and 20 variables. We used varimax rotation to better plot the results. Details of the correlations of the varimax rotated factors and the original variables are in Table 1.

The first factor explains 45% of the total item variance. We discuss the five most prominent items, in descending order based on the absolute magnitudes of factor loadings. As can be seen in Table 1, the first factor is negatively correlated with (a) family contacts and (b) family care and positively correlated with (c) adults who think that parents should go into institutions if they become frail, (d) older people receiving formal care only and (e) percentage of divorced older women. The second factor is largely defined by socio-economic measures: high fertility and levels of childlessness among older cohorts load negatively on the second factor, and proportions of persons aged 65 years or over in 2001, proportion of GDP spent on pensions, and life expectancy at age 65 load positively on it. The third factor is largely characterised by demographic measures: negatively correlated with high proportions of older married women in 1970 and in 2000/2001 (due to past and present low divorce rates); and positively correlated with older ageing processes (e.g. high proportions of older people in 1970), the subjective income indicator, and proportion of older women living alone in 2000/2001.

Figure 1 shows the position of each country on the factorial plane that has the first factor on the x-axis and the second on the y-axis. The southern European countries (in addition to Germany, Austria and Ireland) are grouped together on the left side of the factorial plane (reflecting high levels of family contact and care, and religiosity) in contrast to Sweden, Denmark and The Netherlands, which are close together on the right side of the factorial plane: defined by a high proportion of (a) older people in institutions and receiving formal care, (b) divorced older people, (c) adults who think parents should go to institutions if they become frail and (d) older people living alone. The positions of Belgium, France and the United Kingdom in this plane are not as strongly defined, given that these countries demonstrate an ‘average’ level for the variables used in this analysis. The second factor represented by the y-axis essentially contrasts Ireland with the other countries, reflecting Ireland’s greater religiosity, higher fertility among older cohorts, higher levels of family contact and care, younger population and lower life expectancy.

Fig. 1
figure 1

Results of principal components analysis: x-axis first factor; y-axis second factor

Full size image

Figure 2 shows the countries’ position on the factorial plane that has the first factor on the x-axis and the third on the y-axis. Scandinavian countries are again in the first quadrant. Germany and Austria are in the second quadrant, together with Ireland and Italy, which is characterised by countries with relatively high (a) proportions receiving family care, (b) health expenditure and (c) proportions reporting living comfortably or coping on their present income. The other southern European countries are grouped together in the third quadrant reflecting lower proportions of older people in 1970 and high proportions of married older women in 2000/2001. In the fourth quadrant we find The Netherlands (due mainly to the high proportion of older women in 1970). The positions of Belgium, France and the United Kingdom in this plane are, again, close to the origin.

Fig. 2
figure 2

Results of principal components analysis: x-axis first factor; y-axis third factor

Full size image

We performed a hierarchical agglomerative cluster analysis on the same data as were used for the PCA, using between group linkage as the cluster method and squared Euclidean distance as the interval measure. We considered a four-cluster solution given that this is where the largest gap in the distance coefficients occurred. In the four-cluster solution the first cluster was formed by Austria, Belgium, France and the United Kingdom; the second by Denmark, The Netherlands and Sweden; the third by Germany, Ireland, Italy, and Spain; and the last cluster by Greece and Portugal.

Using the four-cluster solution we compared the average of each variable within the cluster with the overall average to identify key variables for each cluster (results not shown). For example, in the four-cluster solution in the first cluster (Austria, Belgium, France, United Kingdom), the average values of the variables were very close to the overall average values. The second cluster (Denmark, The Netherlands, and Sweden) is characterised by a higher level of provision of formal care, a lower level of family care and daily contacts with family, lower cohort fertility, higher proportion of divorced women and of women living alone, higher prevalence of older people feeling comfortable with their income, greater proportion of women in the labour force, and higher levels resident in institutions and receiving public care. The third cluster includes Germany, Ireland, Italy, and Spain and is characterised by a lower proportion of older people receiving formal care, a higher prevalence of daily contacts with family, a lower prevalence of divorced women and a lower proportion of adults who believe their parents should go into nursing home care if needed. The fourth cluster includes Greece and Portugal and is very similar to the third cluster except for the high level of religiosity and lower proportion of older people that feel comfortable with their income.

Summary and discussion

Our analyses show that attitudinal and cultural indicators are most important for explaining variation across the countries studied. Variations in attitudes may reflect cultural norms and values which emphasise obligations of mutual aid between parents and children throughout life, resulting not only in higher levels of family support to elderly relatives in need but continued levels of assistance from elderly parents to their adult children. However, national differences in family support are not wholly explained by cultural differences. Furthermore, it is important to remember that attitudes toward family responsibility for the care of frail older people are also likely to reflect differences in the policy environments across the selected countries, i.e. in the availability, cost and quality of public service provision.

Demographic constraints also play an important role: proportions married and childless are key indicators of the availability of potential carers. In contrast to earlier cohorts, today’s older people are more likely to have married and survived into later life with a spouse. In addition, current cohorts of older people have more children than with those in the past had and those in the future will have.

Rather than a clear-cut north-south division European countries reflect a more complex classification in terms of support for older individuals. While it is true that there have been similar demographic and socio-economic trends across Europe (e.g. increases in the proportion of older people and increasing female labour force participation) differences between countries in indicators of support for older people remain. When a wide range of measures are used which are associated with different dimensions of support for older people it is difficult to categorise European countries according to a clear north-south divide. For example, Ireland seems to have a unique position, due mainly to its younger age structure but also high religiosity (in common with southern European countries). Italy is not always close to the other southern European countries, but shows similar characteristics to Germany and Austria (i.e. countries with high levels of family support and high expenditure on social provision for older people). France, Belgium and the United Kingdom were not markedly defined by the measures used. Based on the variables considered The Netherlands appears to be more similar to the Nordic countries (e.g. Denmark and Sweden) than to the other western European countries (e.g. Germany) with whom it is very often associated. These results were supported by the cluster analysis which showed similar groupings to the PCA analysis. In both analyses Portugal and Greece behaved differently than Italy and Spain in terms of support for older people.

Our analysis is limited by available comparable measures of family support. Better cross-national indicators of family support would have significantly improved the analysis. For example, projects such as the Survey of Health, Ageing and Retirement in Europe (SHARE), which involves the collection of household surveys based on the United States Health and Retirement Survey (HRS) and the English Longitudinal Study of Ageing (ELSA) will provide future comparable data on the availability of kin and social support (e.g. living arrangements, family contact, and type, frequency and number of hours of household assistance including care; http://www.share-project.org).