Confidant Network Types and Well-Being Among Older Europeans

Abstract

Purpose of the Study: To derive a typology of confidant networks among older adults in Europe and to examine them in relation to country differences and well-being (CASP-12). Design and Methods: The study population was composed of persons aged 65 and older in 16 countries from the 4th wave of the Survey of Health, Ageing and Retirement in Europe (N = 28,697). K-means cluster analysis was applied to data from a newly implemented name-generating network inventory. CASP-12 scores were regressed on network type controlling for country and potential sociodemographic and health confounders. Results: Six prototypical confidant network types were discerned, including proximal and distal family-based networks of varying configurations, as well as friend-based and other-based network types. Regional country differences in network type constellations were observed. Better well-being was found to be associated with network types with greater social capital. Respondents with no named confidants had the lowest CASP-12 scores, and those embedded in “other” network types also exhibited a negative association with well-being. Implications: The study demonstrates the utility of name-generating network inventories in understanding the social capital of older persons. It also shows that accessible family ties are strong correlates of well-being in this population. Finally, it documents the importance of improving the means to detect the small but significant subgroup of isolated older people—those who have no confidants on whom they may rely.

The social networks of older people are important factors in facilitating a “good old age” due to their association with better well-being (Pinquart & Sorensen, 2000), enhanced utilization of health and social services (Bowling, Farquhar, & Browne, 1991; Litwin, 2004), and greater longevity (Brown, Consedine, & Magai, 2005; Eriksson, Hessler, Sundh, & Steen, 1999). Social networks are a key measure of the extent of social capital that older adults may have, that is, the collection of social contacts that give them access to social, emotional, and practical support (Gray, 2009). On the basis of several studies, it can be said that older persons in networks endowed with greater interpersonal resources, or what is referred to as greater social capital, show greater well-being than those who are in lesser endowed networks.(Dominguez & Arford, 2010; Garcia, Banegas, Perez-Regadera, Cabrera, & Rodriguez-Artalejo, 2005; Guilley et al., 2005; McLaughlin, Vagenas, Pachana, Begum, & Dobson, 2010).

“Network type” is a relatively new construct in the realm of social network analysis. First coined in the gerontological context by Wenger (1991), the construct is a composite characterization of the nature and the extent of one’s social capital. Identification of network types permits analysis as to how relationships and emotional interconnectedness can interplay with health and emotional well-being in late life. Network types have been shown to predict such mental health outcomes among older people as depressive symptomatology (Fiori, Antonucci, & Cortina, 2006), morale (Litwin, 2001), anxiety, loneliness, and happiness (Litwin & Shiovitz-Ezra, 2011). They are also related to physical health, including vision, incontinence and self-rated health (Litwin, 1998), functional dependency (Doubova, Perez-Cuevas, Espinosa-Alarcon, & Flores-Hernandez, 2010), and mortality (Litwin & Shiovitz-Ezra, 2006).

There is some debate as to whether network types are culturally bound. Four network constellations, “diverse,” “family-focused,” “friend-focused,” and “restricted” network types, have been discussed in the literature based on research studies using various country data sources (Fiori, Antonucci, & Akiyama, 2008; Fiori et al., 2006; Fiori, Smith, & Antonucci, 2007; Litwin & Shiovitz-Ezra, 2006). The emergence of these same types across different settings might suggest that such groupings are not culturally dependent. However, other studies point to unique network types in different cultural settings, as for example, the widowed network type in Mexico (Doubova, Perez-Cuevas, Espinosa-Alarcon, & Flores-Hernandez, 2010), the isolated network type in Korea (Cheon, 2010), the distant family network in Hong Kong (Cheng, Lee, Chan, Leung, & Lee, 2009), and the congregant network type in the United States (Litwin & Shiovitz-Ezra, 2011). This diversity raises the possibility that network types may be conditioned by the culture in which one is embedded.

It should also be pointed out that most social network assessment inventories relate to the network in its broadest sense. This dominant approach encompasses the enumeration of all possible relationships regardless of the degree of affinity with them. The underlying assumption is that relationships have a meaningful role in persons’ lives simply because they exist (Litwin, 1996). However, without consideration of the true nature of the personal interactions, this indirect “role relational” basis for social network derivation may give an inaccurate portrayal of a person’s interpersonal environment.

An alternative but less common approach is to focus on named confidants. Confidant networks are composed solely of persons who are defined by focal respondents as especially meaningful to them (McPherson, Smith-Lovin, & Brashears, 2009). This direct approach to identifying social network uses a name generator that asks, for example, with whom one discussed important matters in the previous 12 months (McPherson, Smith-Lovin, & Brashears, 2006). Once confidants are nominated, subsequent additional information can be solicited on the cited persons, producing a more exact representation of the composition of and interactions with the identified personal social network.

The use of name generators for network identification is not widespread in large surveys. The approach has been applied to a general adult population in the American General Social Survey (Burt & Guilarte, 1986) and to older adult populations in the Longitudinal Aging Study Amsterdam (van Tilburg, 1994) and the National Social Life, Health and Ageing Project (Cornwell, Schumm, Laumann, & Graber, 2009). In 2010, the Survey of Health, Ageing and Retirement in Europe (SHARE) introduced a name generator in its fourth wave, making it the first harmonized cross-national study of older people to employ this approach (Litwin, Stoeckel, Roll, Shiovitz-Ezra, & Kotte, 2013). SHARE is a multinational panel survey that is gathered in 20 countries and currently includes more than 60,000 survey participants.

Little work has been reported to date on the derivation of network types using name-generated social network data. Work on this topic, thus far, stems mostly from role relationship–based inventories. The current study expands the field of inquiry in three directions. First, it considers whether confidant network types are similar to the network types previously derived through other methodological approaches. Second, through its use of a cross-national survey (SHARE), the study explores if confidant network types differ by country. Third, it examines whether network types maintain differing associations with well-being in late life and if older adults who lack a confidant network are at greater risk than those who have such a network.

The study examines three hypotheses:

1. Confidant network types differ by country.

2. Confidant network types with greater social capital are associated with better well-being.

3. Older adults without a confidant network have poorer well-being than those with such a network.

Design and Methods

This analysis utilizes data from the fourth wave of SHARE, executed in 2010 in 16 countries. The survey queried respondents aged 50 and older and their spouses of any age. The current analysis considered the subsample of respondents aged 65 and older in order to focus on the older population. The study sample size, by country, is presented in Table 1.

Table 1.

Study Sample by Country

Country	N	%
Northern
Sweden	1,342	4.5
Denmark	1,045	3.5
Western
Germany	962	3.2
Netherlands	1,389	4.7
Belgium	2,373	8.0
France	2,821	9.5
Switzerland	1,816	6.1
Austria	2,626	8.8
Southern
Portugal	993	3.1
Spain	2,057	6.9
Italy	1,981	6.7
Eastern
Estonia	3,675	12.4
Poland	935	3.1
Czech Republic	3,007	10.1
Hungary	1,380	4.6
Slovenia	1,306	4.4
Total	29,708	100.0

Table 2 presents the overall background characteristics. The sample had a majority of women (55.7%) and average age was 74 years. Approximately two thirds of the respondents were living with a spouse or partner. Based on the International Standard Classification of Education scale (ISCED-97), half the sample had only a primary level of education. Finally, respondents reported having, on average, between two and three difficulties among a listing of 10 basic mobility, arm, and fine motor functions.

Table 2.

Europeans Aged 65 and Older: Univariate Description of the Background Characteristics of the Sample

Characteristics	N	%	Mean	SD	Range
Age	29,708		74.3	6.9	65–104
Gender
Men	13,151	44.3
Women	16,557	55.7
Marital status
Live-in partner	18,928	63.7
No live-in partnera	10,340	34.8
Primary educationb	14,516	50.0	2.4	1.6	0–6
Secondary educationb	7,928	27.3
Postsecondary educationb	6,598	22.7
Mobility difficultiesc	29,527		2.3	2.6	0–10

Notes: ^aIncludes widowed, divorced, never married, and married living separately from spouse.

^bMeasured on the International Standard Classification of Education Scale (ISCED-97); primary (0–2), secondary (3), postsecondary (4–6).

^cCount of following: walking 100 m, sitting 2hr, rising from chair, climbing several flights of stairs, climbing one flight of stairs, stooping/kneeling/crouching, reaching above shoulders, pulling/pushing large objects, lifting/carrying weights > 10 lb/5kg, picking up small coin from table.

In SHARE Wave 4, name-generated social network confidants were identified in response to an open-ended enquiry, “Looking back over the last 12 months, who are the people with whom you most often discussed important things?” accompanied by the clarification that “These people may include your family members, friends, neighbors, or other acquaintances.” Survey participants were permitted to list up to seven names.

K-means cluster analysis was used to derive confidant network types among the sample that named one or more confidants (Milligan & Cooper, 1987; Rapkin & Luke, 1993). Eight criterion variables were used. The first five variables, which characterized the relationship compositional character of the confidant network grouping, reflected the proportion of the named network composed of the following relationship groupings: (a) spouse or partner, (b) children, (c) other family, (d) friends, and (e) others. Other family members included parents, parents-in-law, siblings, grandchildren, or extended family members. The other category was composed of neighbors, (ex-)colleagues, or formal helpers. The remaining three criterion variables took into account the relational dynamics. Proximity was measured as the proportion of cited confidants who resided within 5 km of the respondent’s residence (the survey response categories were <1, 1–5, 5–25, 25–100, 100–500, and >500 km). Contact was calculated as the proportion of named confidants with whom the respondent maintained daily contact (the survey response categories were daily, several times a week, about once a week, about every 2 weeks, about once a month, and less than once a month). Emotional closeness was indicated as the proportion of cited persons with whom the respondent felt very or extremely close (the survey response categories were not very close, somewhat close, very close, and extremely close). The results of the clustering procedure are displayed in Table 3. Overall mean values for each of the criterion variables are presented in the last column.

Table 3.

Confidant Network Types Among Europeans Aged 65 and Older by Criterion Variables: K-Means Cluster Analysis

Criterion (%)	Network type							Total/meana
	Spouse and children	Children	Spouse	Other family	Friend	Other	No network
Spouse	27.4	9.3	97.2	12.3	10.9	10.4	—	30.4
Children	61.6	73.5	0.0	13.0	9.7	12.7	—	34.3
Other family	6.6	8.1	1.3	63.8	7.0	9.1	—	15.5
Friends	2.7	7.0	0.9	7.7	69.7	6.8	—	13.8
Others	1.7	2.1	0.7	3.2	2.6	61.1	—	6.0
5-km proximity	85.0	35.0	99.9	56.1	62.5	74.7	—	69.3
Daily contact	77.6	23.9	100.0	36.8	26.6	35.1	—	54.5
Very close	93.7	89.2	94.5	83.6	66.2	44.9	—	84.0
% of sample	23.4	18.4	17.4	15.3	13.5	6.1	5.9	100.0
Network sizeB	2.5	3.1	1.1	3.0	3.0	2.7	0.0	2.6

Notes: N = 28,697.

^aMean is computed for all respondents with at least one confidant.

^bConfidant network size range: 0–7.

Respondents without a confidant network were defined as those who did not name anyone. Six percent of the sample had no confidants. For purposes of the current analysis, the “no network” category was added to the collection of derived network clusters. Thus, the “confidant network type” variable encompassed the entire 65+ sample, including those having no confidants at all, in the relevant analyses.

Well-being was measured on the validated 12-item version of the CASP scale that is employed in SHARE (Wiggins, Netuveli, Hyde, Higgs, & Blane, 2008). The scale items encompass four quality of life domains: control, autonomy, self-realization, and pleasure, which combined provide a comprehensive summary of the state of well-being. Information for each of the 12 statements is collected on 4-point Likert scales identifying a person’s feelings about how frequently the statement applies to his or her life: “often,” “sometimes,” “not often,” or “never” (total score range = 12–48; higher scores indicate better quality of life). The CASP-12 score had a good level of internal consistency in this analysis (α = .82).

The analysis proceeded in several stages. First, confidant network types were derived by means of K-means cluster analysis applied to the eight criterion variables explained earlier (Table 3). All of the criterion variables were measured on identical scales ranging from 0 to 100, enabling appropriate utilization of the clustering procedure. Next, the confidant network types and the well-being outcome (CASP-12) were examined in relation to the respondents’ background characteristics and functional health, specifically age, gender, marital status (coresiding with partner), education, and mobility difficulties. Subsequently, network type frequency was examined by country. A chi-square test was performed to test for significance of country differences. Analysis of adjusted standardized residuals (ASRs) was then conducted to identify which cluster distributions by country differed significantly from expected values.

In the last stage of the analysis, the well-being outcome was regressed on the confidant network types by means of hierarchical regressions. In the first model, the CASP-12 scores were regressed on confidant network type. The second model did the same, controlling for country. The third model added the background and functional health variables as additional controls. Effect coding was employed for both network type and country, such that the scores for each of the values in these two variables reflect the deviation of the CASP-12 score for each value from the CASP-12 mean for the variable as a whole.

Results

The average number of named confidants was 2.6, among the subsample of respondents who reported one or more confidant network members (Table 3). The optimal cluster solution identified six distinct confidant network types. The first four types were mainly family based. The two most predominant clusters that emerged were composed of close family members. The “spouse and children” network type was composed almost entirely of spouse and children confidants. Averaging between two and three confidants, this grouping also reflected high proximity, contact, and emotional closeness. Children comprised almost three quarters of the “children” network type, which numbered, on average, three confidants. Although similar in emotional closeness with their confidants as respondents in the “spouse and child” network type, respondents in the “children” network cluster reported lower proximity and frequency of daily contact.

The third family-based network was termed the “spouse” network, in that it numbered about one confidant and, in almost all cases, the named confidant was a spouse or partner. This grouping exhibited the highest proximity, contact, and emotional closeness. Other family members (such as parents and siblings) comprised the majority category in the last of the family-based confidant networks, termed the “other family” type. The average size was three members. Respondents in this cluster reported somewhat high emotional closeness with their confidants, but only moderate proximity and lower frequency of contact.

The two primarily nonfamily confidant network types were named “friend” and “other” networks, respectively. The friend grouping had three confidants, of which more than two thirds were friends. Respondents in the “friend” cluster reported moderate proximity and moderate emotional closeness to their confidants, but low frequency of contact. The dominant categories in the “other” network type, in contrast, were neighbors, colleagues, or formal helpers. Although low in contact and in emotional closeness, these confidants (almost three on average) were in moderate proximity to the focal respondent.

Table 4 presents the background and functional health characteristics of respondents by confidant network type. Those in the “spouse” and “friend” networks were younger than the respondents in the other confidant network types. Women were much less prevalent in “spouse” networks. In comparison, they accounted for approximately two thirds of all respondents in the “children,” “other family,” and “friend” networks. Respondents who lived with a spouse or partner were relatively less represented in the “children,” “other family,” “friend,” and “other” network types. As for schooling, those in “friend” networks had the highest level of education. Finally, mobility difficulties were higher in the “spouse and children” and the “other” networks. Respondents with no reported confidants were older, on average, and fewer of them lived with a spouse or partner. They also reported the least schooling and the greatest number of mobility difficulties among the respondents.

Table 4.

Background Characteristics, Functional Health, and Well-Being of Europeans Aged 65 and Older by Confidant Network Type: Analysis of Variance

Characteristics	Network type							F
	Spouse and children	Children	Spouse	Other family	Friend	Other	No network
Average age	74.3	75.1	73.1	74.3	73.0	74.9	76.9	107.0***
% Women	56.1	65.7	33.2	64.4	60.8	58.4	51.9	249.8***
% Living with partner	73.7	51.4	96.3	54.0	54.4	47.6	50.7	674.3***
% Primary educationa	57.7	48.3	47.6	48.1	37.6	49.7	62.8	88.7***
% Secondary educationa	24.9	26.6	30.1	27.7	31.6	25.9	22.9	15.5***
% Postsecondary educationa	17.4	25.1	22.3	24.2	30.8	24.4	14.3	57.4***
Mobility difficultiesb	2.6	2.2	1.9	2.2	1.8	2.6	3.3	107.1***
CASP-12	35.4	36.9	36.5	36.5	37.6	35.4	34.1	77.8***

^aMeasured on the International Standard Classification of Education Scale (ISCED-97); primary (0–2), secondary (3), postsecondary (4–6).

^bCount of mobility difficulties; range 0–10.

***p < .001.

The last line in Table 4 shows the CASP-12 well-being scores for each network type. Tukey tests of the unadjusted group mean differences (not shown) revealed four distinct levels of well-being. The “friend” network type had the highest scores. The lowest well-being scores among those with a confidant network were respondents in the “spouse and children” network and “other” network. The “spouse,” “children,” and “other family” network types fell in between. The lowest of the well-being scores among all respondents was found among those with no confidants at all.

The frequency distribution of confidant network type by country is displayed in Table 5 and reveals that all network types were present in all countries represented in the SHARE survey. Pearson chi-square results (χ²(28,697, 90) = 3229.63; p < .001) indicate that relative distribution of network types differed across countries. Adjusted standardized residuals (ASRs) were examined to determine which specific distributions differed significantly from the expected value. The overall desired significance (.01) was adjusted for the multiple post hoc tests using the Sidak correction formula and then used to determine the critical ASR z value (3.92). Differences between cells were assumed to be significantly different when standardized adjusted residuals were less than −3.92 or greater than 3.92. Findings indicate that the Northern and Western countries had more “children” networks and fewer “spouse and children” networks in comparison. The opposite was found in the Southern countries where “spouse and child” network types were more predominant. Exclusive “spouse” networks stood out in the Eastern countries as did the lesser frequency of having “friend” networks (the latter of which were more prevalent in the Northern and Western countries).

Table 5.

The Confidant Networks of Europeans Aged 65 and Older by Country: Mean Percentages

Country	Network type
	Spouse and children	Children	Spouse	Other family	Friend	Other	No network
Sweden	14.9*	25.7*	17.2	15.8	17.8*	5.3	3.4*
Denmark	13.3*	22.9	16.5	15.5	19.4*	7.3	5.2
Germany	24.2	20.2	13.7	16.6	13.9	6.6	4.9
Netherlands	15.4*	24.4*	17.5	18.6	14.6	6.9	2.7*
Belgium	14.4*	21.8*	10.2*	19.0*	19.5*	8.8*	6.4
France	12.9*	22.8*	11.9*	17.2	18.1*	7.6*	9.4*
Switzerland	11.5*	21.1*	14.2*	16.6	24.0*	7.8	4.7
Austria	22.7	22.5*	17.7	15.7	15.4	3.0*	2.9*
Portugal	36.5*	12.7*	18.6	16.1	6.4*	5.2	4.6
Spain	42.8*	8.6*	18.1	12.1*	9.2*	3.2*	6.1
Italy	35.8*	8.5*	16.4	13.5	12.1	3.2*	10.4*
Estonia	20.8	19.1	18.9	17.3*	11.0*	8.3*	4.6*
Poland	34.7*	13.1*	26.7*	10.9	4.0*	3.7	6.9
Czech Republic	24.9	19.4	21.0*	11.2*	10.1*	6.3	7.1
Hungary	41.8*	16.2	13.5*	14.8	5.3*	4.7	3.8
Slovenia	22.0	8.2*	34.3*	11.4*	8.1*	6.8	9.3*

Notes: χ²(28,697, 90) = 3229.63; p < .001.

*Adjusted standardized residuals (ASRs). <−3.92 or > 3.92.

Table 6 presents the results of the multivariate regressions. In the first model, all the network types were entered using effect coding. The confidant network type variables accounted for 1.6% of the variance in the well-being outcome. Cohen’s f ² was 0.02, indicating that, although significant, the effect size of confidant network type was small.

Table 6.

Confidant Network Type Correlates of Well-Being Among Europeans Aged 65 and Older, Controlling for Country and Background Characteristics: Weighted OLS Hierarchical Regressions

Variables	Model 1	Model 2	Model 3
	Beta	Beta	Beta
Network type
Spouse and children	−.062***	.019*	.032***
Children	.075***	.038***	.053***
Spouse	.039***	.055***	−.008
Other family	.037***	.027**	.028***
Friend	.129***	.073***	.034***
Other	−.047***	−.058***	−.024**
No network	−.127***	−.106***	−.082***
Sweden		.121***	.069***
Denmark		.236***	.167***
Germany		.123***	.085***
Netherlands		.247***	.187***
Belgium		.015	.016*
France		.029**	.025**
Switzerland		.265***	.172***
Austria		.159***	.139***
Portugal		−.294***	−.258***
Spain		−.135***	−.060***
Italy		−.251***	−.209***
Estonia		−.130***	−.113***
Poland		−.140***	−.082***
Czech Republic		−.169***	−.161***
Hungary		−.166***	−.112***
Slovenia		.095***	.138***
Age			−.039***
Women			.015**
Married			.039***
Secondary educationa			.054***
Postsecondary educationa			.073***
Mobility difficultiesb			−.422***
R 2	.016	.169	.368
ΔR 2	—	.153	.198

Notes: OLS = ordinary least squares.

^aReference category: Primary education.

^bRange: 0–10.

*p < .05. **p < .01. ***p < .001.

Model 2 controlled for country differences, which explained an additional 15% of the variance in the outcome measure. The effect size of country on CASP-12 scores was at a medium level (Cohen’s f ² = 0.18). After taking country into account, the associations between network type and the well-being outcome changed somewhat. The findings also revealed that, in general, Northern and Western European countries had positive associations with CASP-12 well-being. Negative associations were found for Southern and Eastern European nations, with the exception of Slovenia.

Model 3 of the regression entered the background and functional health variables into the analysis, explaining an additional 20% of the variance. Cohen’s f ² was 0.31, indicating that the effect size of background and functional health on well-being, as expected, was considerable. After taking these variables into account, the confidant network types revealed their unique net associations with the outcome measure. Respondents in the “children” network emerged as the network type most positively related with well-being, all things considered. The “friend” and “spouse and children” networks were the second and third most positively related confidant networks. “Other family” network types had a slightly lesser but still positive association with well-being. The “spouse” network type was statistically insignificant when controlling for country, background, and functional health. Respondents in the “other” network showed a modest negative association with the well-being outcome, all else considered, and those having no confidants had the most negative such association.

As for the control variables, the analysis showed that functional mobility difficulties were strongly and negatively associated with well-being. Age was also negatively associated, but to a much lesser degree. Gender (female), higher education, and living with a partner were positively related to the well-being outcome. The regional country differences found in the second model remained the same after controlling for background and health.

Discussion

We examined the confidant networks of older Europeans using a name-generating network inventory that was newly implemented in the fourth wave of SHARE. Our aim was to derive a series of prototypical network types among the collection of named confidants. The inquiry sought to clarify whether confidant network types are similar to the social network types that appear in the literature and whether they are similarly associated with the well-being of older adults. In addition, the study explored whether such name-generated network typing can identify older adults who lack meaningful social ties, and if persons without confidants are at greater risk of poor well-being than those who have a confidant network of any kind.

The inquiry revealed that the confidant networks numbered two and a half members, on average, and were hence smaller than the social networks that are derived from more general inventories of social relationships (Fiori, Consedine, & Merz, 2011; Hawkley et al., 2008; Mendes de Leon, Gold, Glass, Kaplan, & George, 2001). Six prototypical confidant network types were discerned. Moreover, almost three quarters of the respondents were embedded in family-based networks, labeled “spouse and children,” “children,” “spouse,” and “other family,” respectively. Two additional nonfamily-oriented network types were labeled the “friend” network and the “other” network. The latter was composed largely of neighbors, current or former colleagues, and formal helpers. Together, these two types accounted for a fifth of the respondents. Another 6% of the respondent sample reported having no confidant at all.

The four family-based confidant network types differed not only by their composition but by their modes of interaction. All were very close emotionally, but only the “spouse and children” and the “spouse” networks maintained high proximity and frequent contact. The typology thus distinguishes between proximal family-based confidant networks and distal family-based confidant networks, those with whom one might practice “intimacy at a distance.” In comparison, the “friend” network was characterized by relatively less emotional closeness and contact but also by a moderate degree of proximity. It seems, therefore, that friend-based confidant networks do not share the degree of daily contact in the lives of their focal members that most of the family-based networks do. However, older adults with this type of confidant network live in closer proximity to their confidants than those with a children confidant network, which makes the friend networks more accessible if immediate need for support should arise. In terms of proximity and contact, the “other” network type was more available than its friend-based counterpart. This latter network type, of which the majority of members were neither family nor friends, was less emotionally accessible to the older adults. This suggests that even when identified as confidants, persons who are formal helpers, current or former colleagues, and neighbors may play a less emotionally supportive role in the lives of older people despite nearer proximity and more regular contact.

Comparing this typology with the ones reported in the literature reveals that confidant networks seem to be more heavily family based than are the more general social network types (Dykstra, 1993; Gray, 2009). Moreover, they point to a more specific differentiation among the respective family ties themselves. This result may be related to the measurement method, insofar as general social network inventories ask broader questions about contact with people defined as relatives, friends, or acquaintances, with little distinction about the nature of the relationships. In comparison, name-generated confidant lists are identified by the focal respondent as the most central persons in their lives. For this reason, we consider the confidant network listing to more accurately reflect the supportive ties upon which older people may call in time of need.

The confidant network typology also identifies friend-based and other-based groupings alongside the more prevalent family-based groupings. The identification of nonfamily members as confidants and the inclusion of these nonfamily network types in the cluster analysis highlight the integral role of nonfamily relationships in the lives of some older adults. Network types of which nonfamily members comprise the majority of the confidant network also appear, to varying degrees, in the general social network type literature. However, the distinction offered by the current inquiry is that named confidant network clustering reveals that nonfamily networks ties are less prevalent than was previously found in more general social network typologies.

Although all of the network types were found in all of the countries, their relative distribution differed. This finding supports our first hypothesis that confidant network types differ by country. In general, proximate familial network types were more prevalent in Southern countries, whereas distal family-based and nonfamilial network types were more prevalent in Northern and Western countries. These findings align with previous research, which highlighted Mediterranean and non-Mediterranean regional distinctions of indirect role-relationship social network characteristics (Litwin, 2010). This suggests that cultural context has an influential role on the types of social relationships maintained by older Europeans and needs to be taken into account when considering the interpersonal milieus of older people.

The second research hypothesis considered the association between confidant network types and well-being. Comparing across different studies of social network types is a challenge, in this regard, due to the wide range of well-being measures that are employed in different studies and the presence of cultural differences as to what constitutes quality of life. Results may also vary according to the way social capital is measured. In the current inquiry, we focused on confidant network types as the social capital variable of interest, and we controlled for country to take cultural differences into account.

The findings from the present study support our second hypothesis that network types with greater social capital are associated with better well-being. The final regression model revealed that the respondents with the least “network capital,” those in the “other” network type, had significantly lower well-being scores, all else considered. The number of confidants in the “other” network grouping was slightly above the average (among those with a confidant network) and their geographic proximity to the respondent was quite close—three quarters of them resided within 5 km. However, a relatively small proportion of these confidants were in daily contact with the respondent. More importantly, they were ranked as having the lowest degree of emotional closeness to the respondent among the confidants in all the network types. As such, the respondents in the “other” network type did, indeed, have less social capital in terms of their confidant networks.

Another finding that supports the second hypothesis, to some degree, emerged in relation to the spouse network. By definition, confidant networks, which include mainly or only a spouse, have high proximity and daily contact. Our findings also showed that they were ranked as having a high degree of emotional closeness. Nonetheless, after taking the other study variables into account, this network type was found to be unrelated to well-being. The reason for this outcome may well be the limited size of this network type—only one confidant. Stated differently, irrespective of their quality, networks that are based upon a single tie do not have as much social capital as network types having a greater number of confidants.

The final point of inquiry examined the distinguishing characteristics of those who named no confidant at all and investigated the third research hypothesis of the study, which conjectured that this socially isolated subgroup of older Europeans differed significantly in terms of well-being. Identification of such persons is a challenge in more general social network inventories because most respondents can report having at least some relatives or neighbors. As such, the truly isolated may be overlooked. By employing a name-generating mechanism, it is possible to more clearly identify older people who do not consider themselves as having any meaningful relationships and are therefore at-risk.

The current inquiry found 6% of older Europeans having no confidant at all. Identification of this group is important because, as the present study showed, respondents with no named confidants were also at risk according to other parameters. They were the oldest, on average, few lived with a spouse or partner and they had the lowest education, as well as the greatest number of mobility difficulties. Moreover, those with no confidants had the lowest reported well-being scores in comparison with the other network groupings, confirming out third study hypothesis.

Although the inquiry reported here provides important innovations for gerontological social network assessment, a few limitations of the study should be acknowledged. Firstly, this was an initial effort to delineate confidant network types using name-generated network data. We chose what we believed were the most important indicators as criterion variables for the clustering procedure and we identified what we understood to be the optimal solution—the six-cluster set. Nevertheless, additional clustering efforts might produce different results. We encourage future research using the same database, and similar databases among other populations, in order to further differentiate unique named confidant network types among the older population.

A second limitation is the cross-sectional nature of the inquiry in which the relationship data and the well-being outcome measure were collected at the same point in time. It is not implausible that reduced well-being might have delimited respondents’ exchange with significant others and, consequently, might have influenced the nomination of confidants. Longitudinal follow-up is required, therefore, in order to more fully ascertain the nature and the direction of the association between confidant network type and well-being among older adults.

Yet another challenge to the current inquiry is a possible variation in the understanding of the name-generating probe that was administered in several different languages. The data in the current analysis were gathered in 16 countries in almost as many languages. Although there were careful checks carried out during the online translation of the instrument and in the pretest, it might still be the case that linguistic nuances influenced some of the respondents. For this reason, we controlled for country differences in the multivariate phase of the analysis.

A final limitation stems from the limited variance accounted for by the network type construct. Moreover, its statistical significance may have been, at least partly, a concomitant of the large sample size. Future analysis will hopefully further clarify the contribution of the confidant network construct in relation to the well-being of older people.

The limitations notwithstanding, the current inquiry has several strengths. The extent and breadth of the SHARE survey allows for analysis to be performed on a very large international respondent sample. In addition, the introduction of the social network name-generating instrument in Wave 4 was accompanied by extensive training of interviewers and extensive treatment of the data in order to guarantee a reliable database. The unique analytical opportunity provided by these new SHARE data offsets the potential limitations of the data set.

The current inquiry is important not only because it sheds new light on the possible role of social relationships in furthering well-being in late life but because it raises new directions for gerontological practice. The significance of social networks is already evident in the adoption of network inventories for assessment and for care planning (Lubben et al., 2006; Wenger, 1997; Wenger & Tucker, 2002). However, our findings suggest that it may not be enough to ask the usual role-relational network questions that are typical of surveys and assessment forms. Rather, efforts should be made to introduce name-generating relationship inventories into day-to-day clinical practices with older adults. Not only is this approach helpful in better assessing the availability of social capital, it is essential in order to successfully identify socially isolated older adults. The current research empirically highlights the vulnerability of this subgroup of older persons. Detection of this subgroup is critical in order for practitioners to identify and best meet the unique needs of socially isolated older adults among their clientele.

In conclusion, this study of named social relationships underscores the presence of different types of confidant networks among older adults in Europe and clarifies what personifies social capital in later life. It also demonstrates that accessible family ties are strong correlates of well-being in this population. Finally, it documents the importance of improving the means to detect the small but significant subgroup of isolated older people—those who have no confidants on whom they may rely.

Funding

The SHARE data collection has been primarily funded by the European Commission through the 5th Framework Programme (QLK6-CT-2001- 00360 in the thematic programme Quality of Life), through the 6th Framework Programme (SHARE-I3, RII-CT-2006-062193, COMPARE, CIT5-CT-2005-028857, SHARELIFE, CIT4-CT-2006-028812), and through the 7th Framework Programme (SHARE-PREP, 211909, SHARE-LEAP, 227822, SHARE M4, 261982). Additional funding from the U.S. National Institute on Aging (U01 AG09740-13S2, P01 AG005842, P01 AG08291, P30 AG12815, Y1-AG-4553-01, IAG BSR06-11, OGHA 04-064) and the German Ministry of Education and Research as well as from various national sources is gratefully acknowledged (see www.share-project.org for a full list of funding institutions).