Bonding personal social capital as an ingredient for positive aging and mental well-being. A study among a sample of Dutch elderly

Abstract

Objectives

The current study aims to add to the limited empirical research of possible benefits of personal social capital for the well-being of elderly. A validated personal social capital scale, measuring both bonding and bridging social capital in a general population, was adjusted to fit the characteristics of the social environment of elderly, to explore the association between social capital and well-being of elderly, as well as the mediating role of loneliness.

Method

A sample of 328 Dutch adults, varying in age from 65 to 90 years (Mean = 72.07; SD = 4.90) filled out an online questionnaire including the adapted personal social capital scale for elderly (PSCSE), as well as validated scales measuring social, emotional, and psychological well-being and loneliness. Relevant other (demographic) variables were included for testing construct and criterion validity.

Results

CFA analysis revealed the subdimensions bonding and bridging social capital with reliability scores of respectively α = .88 and α = .87, and α = .89 for the total scale. Regression analyses confirmed construct and criterion validity. Subsequently, significant positive associations between bonding social capital and respectively social, emotional and psychological well-being were found. These associations were mediated by loneliness. Bridging social capital was only found to be significantly associated with social well-being, not mediated by loneliness.

Conclusion

Our findings have enhanced our understanding of the association between social capital and mental well-being of elderly and indicate that bonding personal social capital in particular may be considered an ingredient for positive aging.

Keywords:

• personal social capital
• bonding social capital
• mental well-being
• positive aging
• elderly

Introduction

The well-being of elderly has become an important topic of research, with a focus on factors that may add to the process of positive aging (Hill, 2011). Personal social capital, described by Putnam (1995) as 'features of social organization such as networks, norms, and social trust that facilitate coordination and cooperation for mutual benefit’ (p. 67), may well be one of these factors. In older populations (60 and older) a relation was found between decreased social capital and a diminished sense of well-being (Biddle, 2012, Keating, Swindle, & Foster, 2005) and experiences of loneliness (Nyqvist, Viktor, Forsman & Cattan, 2016; Santini et al., 2016). Litwin (2001) showed positive correlations between different aspects of social capital and positive mental health in older adults. Greaves and Farbus (2006) observed an increase in quality of life and a decrease in depressive symptoms after providing various social activities and stimulating social networking among older people. Although the potential benefits of social capital for the well-being of elderly are increasingly recognised (Chipps & Jarvis, 2016; Forsman, Herberts, Nyqvist, Wahlbeck, & Schierenbeck, 2013), empirical results are still limited (Nyqvist, Forsman, Giuntoli, & Cattan, 2013). As this might be explained by the lack of a valid quantitative measure of personal social capital for this particular group (Chipps & Jarvis, 2016), we aim with the present study to add to the empirical research on this topic by adjusting a validated personal social capital scale (Chen, Stanton, Gong, Fang, & Li, 2009) for use in an elderly population and subsequently examine associations between personal social capital and mental well-being in a Dutch sample of older adults.

The aforementioned definition of Putnam (1995) is considered the most used approach of social capital in health related and behavioural research literature (Forsman et al., 2013; Nyqvist et al., 2013) and allows for a distinction between a focus on external relationships with a bridging function and a focus on internal relationships involving emotional bonding (Putnam, 2000). Bridging social capital refers to the creation and maintenance of relatively weak and business-like connections in mostly larger networks (Bourdieu & Wacquant, 1992), assumed to be valuable for economic advancement (Zhang, Anderson, & Zhan, 2011), and found positively associated with education (Coleman, 1988; Chen et al., 2009) and financial resources or earnings (Beugelsdijk & Smulders, 2003; Zhang et al., 2011). Bonding social capital concerns close and intimate relationships in a smaller social circle and is argued to be effective as a social and emotional resource (Liu, Ainsworth, & Baumeister, 2016). Furthermore, social capital can either be studied as a collective property or an individual asset (Chen et al., 2009; Nyqvist et al., 2013). In the current study we focus on the latter, also referred to as personal social capital (Bowling, Banister, Sutton, Evans, & Windsor, 2002; Chen et al., 2009). This concept has been operationalised by Chen et al. (2009) as ‘…his or her accumulated network connections that are durable, trustworthy, reciprocal and full of socioeconomic resources….’ (Chen et al., 2009, p. 307), which is in line with Putnam's approach and also distinguishes between bonding and bridging social capital.

We examined associations between personal social capital and three dimensions of mental well-being, that have been elaborately described in previous literature: emotional, psychological and social well-being (Keyes et al., 2008; Lamers, Westerhof, Bohlmeijer, ten Klooster, & Keyes, 2011; Robitschek & Keyes, 2009). Emotional well-being refers to one's own subjective experience of well-being, which manifests itself in the presence of positive affect, absence of negative affect and someone’s cognitive appraisal of satisfaction with life (Diener, 1984; Diener, Suh, Lucas, & Smith, 1999; Keyes, 2009). Psychological well-being refers to effective functioning and self-realisation (Ryff, 1989; Ryff & Singer, 1998) and manifests itself on dimensions such as self-acceptance, personal growth, purpose in life, positive relations with others, autonomy, and environmental mastery, the latter referring to a sense of control, competence and effective use of possibilities (Ryff, 1989; Ryff & Singer, 1998). Social well-being can be understood as someone's functioning in their community and society in general (Keyes, 1998) and concerns aspects as social integration, social contribution, social coherence, social actualization, and social acceptance (Keyes, 1998).

Social capital might help facilitate these three subdimensions of well-being by providing for positive relations with others and offering elderly resources for autonomy and environmental mastery; facilitating feelings of belonging and a sense of coherence through social norms and trust of reciprocal relationships (Putnam, 2000); and providing for emotional resources and experiences of positive affect.

According to the socioemotional selectivity theory (Carstensen, Fung, & Charles, 2003) people tend to prefer social relationships with trusted others, that are emotionally important to them and may add to positive experiences (Carstensen, Mikels, & Mather, 2006; Carstensen, Pasupathi, Mayr, & Nesselroade, 2000). This may imply that at an older age, bonding social capital becomes more important for mental well-being than bridging social capital. Consequently, loneliness, which has been described as a perceived lack of certain quantity and quality of personal relationships (De Jong Gierveld & Van Tilburg, 2006) or a discrepancy between the desired emotional support or companionship and its actual presence in someone's social environment (Blazer, 2002) may well (partly) explain the assumed relationship between personal social capital and mental well-being. Loneliness has been found to be negatively related to aspects of well-being in older age (Golden et al., 2009; Dong, Chang, Wong, & Simon, 2012), and has shown to mediate associations between concepts related to social capital such as social support from friends and children and social network integration on the one hand and depressive symptoms on the other, in samples of older adults (Chen & Feeley, 2014; Santini et al., 2016). In our study we expected bonding social capital to be more strongly associated with mental well-being than bridging social capital is, and that this association would be (partly) mediated by the experience of loneliness.

Method

Procedure and sample

After the study was approved by the local research ethics committee, several Dutch senior associations – linked to the Dutch Federation of General Senior Associations (www.FASv.nl) – were sent a request to invite their members to take part in this research. No inclusion criteria other than age (65 and above) and sufficient command of the Dutch language to understand instructions and give informed consent were applied. The questionnaire was filled out and submitted by a sample of 328 respondents, varying in age from 65 to 90 years (Mean = 72.07 and SD = 4.90).

Data collection

Respondents were asked to fill out an online questionnaire (Limesurvey), from which the data was anonymously collected after they submitted their answers. The online questionnaire included a consent page that, in accordance to the American Psychological Association Ethical principles of psychologists and code of conduct (American Psychological Association, 2010), explained that participation was on a voluntary basis and that the participant could stop any time without giving any reason and without (adverse) effects. It also explained that by submitting their answers, participants consented to the careful and secure use of the data for this study, in compliance with their privacy rights. Only fully completed questionnaires could be submitted.

Measurement

Personal social capital

Personal social capital was measured with the Personal Social Capital Scale for the Elderly (PSCSE). The subitems of the ten items of the validated Personal Social Capital Scale of Chen et al. (2009), measuring overall personal social capital (α = .87) as well as bonding (α = .85) and bridging personal social capital (α = .84) were not only translated but also adjusted and reformulated into statements to optimize fit with characteristics of the social environment of elderly. These adjustments, that are further illustrated below, made the initially intended translation – back translation procedure no longer useful and was therefore omitted.

The categories ‘family members’ and ‘relatives’ were merged, as older people usually live with their partner or alone, which makes it more plausible to use one category of family and relatives, including children, grandchildren, siblings, cousins etc. Categories ‘friends’ and ‘people in the neighbourhood’ were renamed as respectively ‘close friends’ and ‘acquaintances’, to better represent the possible social groups as older adults may live in a retirement or care home, giving the concept of ‘neighbourhood’ a different meaning. We removed the categories ‘coworkers/fellows’ and ‘country fellows/old classmates’ as making these distinctions in social groups are less relevant for elderly. To make the scale more comprehensible and straightforward, the original items were reformulated as statements, clarifying the social group as well as the question per statement, and preventing the respondent from having to look back each time to the original question. In addition, the items measuring bridging social capital were grouped per organization type, assuming that not having to repeatedly switch between different types of organizations while completing the items, makes an easier task. Finally, the scale was presented to a small panel of Dutch elderly (n = 31, age > 65 years) and a few more adjustments were made to the list of assets/resources that members of someone’s social network may possess. The categories ‘certain political power’ and ‘high reputational/influential’, were merged to prevent for the potential problem that to an older population this distinction is less/not relevant.

Items were scored on a 5-point Likert scale (5 = many; 4 = reasonably many; 3 = some; 2 = few; 1 = none). Bonding social capital was calculated by adding the mean scores on 4 statements with regard to family members/relatives, close friends, acquaintances, and others that were formulated for four of the five items, that is bond1 (i.e. I have ……. close friends), bond2 (i.e. I keep a routine contact with ……. family members/relatives), bond3 (i.e. I have ……. acquaintances that I can trust), and bond4 (i.e. I can ask ……. others for help) and the mean score on 5 statements for the fifth item bond5 (i.e. ‘I know ……. people with certain political or other influential power’ or ‘I know ……. people with broad connections with others’.)

Bridging social capital was calculated by adding the mean scores of two statements regarding government/corporate/social organizations and cultural/recreational/leisure organisations in someone’s environment, that were formulated for respectively item bridg1 (i.e. As far as I know, ……. government/corporate/social organizations can be found in my area); bridg2 (i.e. ……. of these organisations represent my rights and interests; bridg3 (i.e. I participate in or am a member of ……. of these organisations); and bridg4 (If I need help, I can call upon ……. of these organizations), as well as the mean score on 8 statements with regard to connections and influence these organisations may have for item bridg5 (i.e. ……. of these organisations have (political) power for or influence on (local) decision making; ……. of these organisations have great social influence).

Adding the scores of bonding and bridging social capital resulted in a score for someone’s total personal social capital.

Mental well-being

The Dutch Continuum Mental Health Short Form (MHC-SF, Lamers et al., 2011) was used to measure social well-being (5 items; i.e. ‘During the past month how often did you feel that you belonged to a community, like a social group, or your neighbourhood’), psychological well-being (6 items; i.e. ‘During the past month, how often did you feel good at managing the responsibilities of your daily life’) and emotional well-being (3 items; i.e. ‘During the past month how often did you feel happy?’). Participants were asked to respond to these items on a 6-point Likert scale, (1 = never; 2 = once or twice; 3 = about once a week; 4 = 2 or 3 times a week; 5 = almost every day; and 6 = every day), based on their experiences of the last month. Mean scores for each subscale (by dividing the total score on each subscale by the number of items of the respective subscale) were computed.

Loneliness

We used the (shortened) Loneliness scale (De Jong Gierveld & Van Tilburg, 2006) to measure someone’s perspective on his or her degree of loneliness. This scale consists of 6 items (such as ‘I miss having people around me’ and ‘There are many people I can trust completely’), that can be responded to by either ‘yes’, ‘more or less’, or ‘no’. Neutral and positive answers on negatively stated items are scored as ‘1’and negative answers as ‘0’. Consequently, neutral and negative answers on positively stated items are scored as ‘1’ and positive answers as ‘0’. This gives a range of scores from 0 (least lonely) to 6 (most lonely).

Demographic variables and confounders

In addition to gender and age, having a partner (0 = no partner; 1 = with partner), and perception of physical health on a 5-point scale from 1 (very poor) to 5 (excellent) were measured. Both variables have been found to be associated with mental well-being (resp. Hooghe & Vanhoutte, 2011; Cho, Martin, Margrett, MacDonald & Poon, 2011) and are likely to undergo changes in older age.

Three additional (demographic) variables were added to examine the construct validity (Thorndike, 2004) of the PSCSE, of which two were already discussed as being positively associated with social capital in the Introduction: level of education (0 = low: up till high school and/or vocational education, 1 = high: at least an undergraduate degree) and financial resources to provide for oneself and (future) needed care (1 = definitely enough, 2 = probably enough, 3 = probably not enough, 4 = definitely not enough). This operationalisation of financial resources was preferred over the assessment of income, as it was assumed that the majority of our elderly sample was no longer employed. Finally, respondents were asked to indicate whether they participated in volunteer work. This variable is not only a useful operationalisation of altruism, which has been found to be positively related to social capital (Theurer & Wister, 2010; Zhang, Zheng & Wei, 2009), it also can be assumed to be valuable for the social network of older adults in particular (Li & Ferraro, 2006), since they usually do not participate anymore in a regular working environment.

Social participation

With the Maastricht Social Participation Profile (MSPP; Mars et al., 2009) social participation was measured, which can serve as a theoretically concurrent relevant measure to test criterion validity of the PSCSE (Thorndike, 2004). Social participation has been found closely related to and sometimes interchangeable with social capital (Engbers, Thompson & Slaper, 2017; Hooghe & Vanhoutte, 2011). The MSSP measures three dimensions, that is ‘doing things’ (10 items), ‘contact with acquaintances’ (8 items) and ‘contact with family’ (8 items), with items such as respectively “How often did you go out for dinner during the last four weeks?’; How often have you had acquaintances coming over the last four weeks?’; and ‘How often did you contact your family by either phone, letter, email or chat?’. All items are scored on a 4-point scale (0 = not at all; 1 = less than once a week; 2 = 1 or 2 times a week; and 3 = more than 2 times a week) and can be used to compute both the frequency (presented by the mean score, dividing the total score by the number of items) and the diversity (addition of items scored >'0') of social participation (Mars et al., 2009). Social capital can be assumed to be associated with both frequency and diversity of participation, since the PSCSE measures both the (subjective) quantity of relations and frequency of contact.

Analysis

First internal consistency and validity of the PSCSE were examined. Crohnbach’s alpha and item-total correlations were computed and the assumed two-factor structure of the scale (bonding and bridging social capital) was explored by exploratory factor analysis (EFA) and further assessed by confirmatory factor analysis (CFA). Sampling adequacy was tested with Kaiser–Meyer–Olkin (KMO) measure, for which scores above .7 were considered good (Hutcheson & Sofroniou, 1999). Additionally, anti-image correlations above .05 were considered acceptable (Field, 2013). For the CFA the goodness of fit (GFI) > .9, the comparative fit index (CFI) ≥ .9 and the standardized root-mean-square residual (SRMR) ≤ .10 (Hu & Bentler, 1999) were used as evaluation criteria. Construct validity was tested with the Student’s t-test using the theory-based group difference approach. Criterion validity was tested with multiple linear regression analyses, including the demographic variables level of education, financial resources and participation in volunteer work as covariates.

Associations between personal social capital, on the one hand, and emotional, psychological and social well-being on the other, were examined with separate regression analyses for each well-being subscale (dependent variable), including both bonding and bridging social capital as predictor variables, to examine differences in their individual contribution. The role of loneliness in the association between bonding social capital and the subscales of well-being, was examined by mediation analysis (PROCESS, Preacher & Hayes, 2004; Hayes, 2013), testing the direct (between social capital and well-being) and indirect associations (via loneliness), using the bootstrapping method as well as the Sobel-test. A priori confounders gender, age, having a life partner, and perception of own physical health were included in all analyses as covariates.

Analyses were conducted using SPSS version 24 (IBM Corp. in Armonk, NY), except for the CFA for which Lavaan (Rosseel, 2012) was used in R/Rstudio. As only fully completed questionnaires could be submitted, no procedure for missing values was needed.

Results

Demographic information is presented in Table 1. Descriptive statistics and reliability of all scales are presented in Table 2. The homogeneity of both the social capital scales and the other included scales computed by Crohnbach's alpha, measured well above .65, indicating sufficient reliability for the purpose of this study (Nunnally & Bernstein, 1994).

Table 1. Demographic variables (n = 328).

Gender	Male 68.6% (n = 225)	Female 31.4% (n = 103)
Partner	yes 78% (n = 256)	no 22%(n = 72)
Residential situation	independent 97% (n = 317)	independent with care at home .01% (n = 4)	living in with son/daughter .003% (n = 1)
Work	yes .03% (n = 10)	retired 93% (n = 305)	never worked .04% (n = 13)
Health (self-report)	good to excellent 91% (n = 300)	moderate .08% (n = 27)	poor .003% (n = 1)

Table 2. Descriptives and reliability of scales (n = 328).

	α	N (items)	Mean	SD	Min	Max
PSCSE	.89	10	27.36	5.22	13.25	44.13
PSCSE-bonding	.88	5	15.17	2.81	7.75	22.75
PSCSE-bridging	.87	5	12.19	3.17	5.50	22.38
Social participation (MSSP)	.84	25	1.03	.39	.16	2.52
Loneliness	.76	6	1.33	.38	1.00	3.00
Social well-being	.76	5	3.59	1.08	1.00	6.00
Psychological well-being	.85	6	4.34	1.06	1.00	6.00
Emotional well-being	.83	3	4.93	.90	1.00	6.00

Internal consistency and validity of the PSCSE

Correlation analysis showed that all social capital item scores positively correlated with the total score. Correlation coefficients varied for PSCSE, PSCSE-bonding and PSCSE-bridging respectively from .55 to .74, .60 to .76, and .56 to .81.

Sampling adequacy was confirmed by a KMO = .87 and anti-image correlations > .82. The rotated matrix showed that the bonding and bridging items indeed loaded on two different factors, with the exception of bonding item bond5, which loaded on both factors (Table 3). CFA of these two factors, resulted in a goodness of fit of the model with scores of χ² (34, n = 326) = 224.482, p < .001, GFI = .88, CFI = .90, and SRMR = 0.08. These results mostly supported the two-factor structure of the PSCSE, although caution must be exercised regarding item bond5.

Table 3. Pattern matrix with loadings of PSCSE items (n = 328).

	Component
	1	2
bridg5	.872
bridg2	.842
bridg3	.757
bridg4	.754
bridg1	.579
bond5	.440	–.409
bond4		–.858
bond2		–.847
bond3		–.812
bond1		–.726

Note: Extraction: Maximum Likelihood. Rotation: Oblimin with Kaiser Normalization.

Results of Student’s t-tests to examine construct validity of the PSCSE are presented in Table 4. Respondents with a higher level of education possessed more overall social capital and bridging social capital than those with a lower level of education. Respondents who were confident that their financial resources will be sufficient to provide for themselves and (future) needed care possessed significantly both more bridging and bonding social capital than respondents who did not. Finally, differences in social capital were observed between respondents who participated in volunteer work and those who did not. The latter group possessed less social capital and this difference was significant for both bridging and bonding social capital. These findings supported the construct validity of the PSCSE.

Table 4. Results of testing construct validity of the PSCSE.

	Total Social Capital			Bridging Social Capital			Bonding Social Capital
	Mean	SD		Mean	SD		Mean	SD
Education
Low (high school or lower) (N = 193)	26.92*	5.05		11.87*	3.08		15.04	2.73
High (undergraduate degree or higher) (N = 135)	27.99	5.40		12.63	3.25		15.36	2.92
	t(326) = –1.832, p =.034 Effect size = .21			t(326) = –2.136, p =.017 Effect size = .24			t(326) = –.996, p =.160 Effect size = .11
	Mean		SD	Mean		SD	Mean		SD
Financial resources
Definitely enough (N = 181)	28.48***		5.39	12.83***		3.30	15.65***		2.93
Maybe/maybe not enough (N = 147)	26.98		4.66	11.40		2.82	14.59		2.55
	t(326)= 4.426, p=.000 Effect size = .30			t(326)= 4.164, p=.000 Effect size = .46			t(326)= 3.472, p=.001 Effect size = .38
	Mean		SD	Mean		SD	Mean		SD
Participating in volunteer work
Not participating in volunteer work (N = 171)	25.92***		5.42	11.07***		3.14	14.85*		3.03
Participating in volunteer work (N = 157)	28.92		4.52	13.39		2.73	15.53		2.51
	t(323.042)= –5.463, p=.000 Effect size = .60			t(325.110)= –7.151, p=.000 Effect size = .79			t(322.795)= –2.217, p=.014 Effect size = .24

Note: Student’s T-test (1-tailed) for mean differences, *p < .05, **p < .01, ***p < .001. Effect sizes have been calculated using Hedges’ g.

Results of regression analysis to test criterion validity of the PSCSE are shown in Table 5. Level of education, financial resources and participation in volunteer work were included as confounders, since these variables were found to be associated with social capital as shown in our testing of construct validity. In all models, both financial resources and participation in volunteer work were significant associated with the social capital variables. Significant associations were found between scores on both frequency and diversity of social participation, on the one hand, and overall social capital, as well as bridging and bonding social capital, on the other. These results support the criterion validity of the PSCSE.

Table 5. Results of testing criterion validity of PSCSE: Regression coefficients (B) of associations theoretically concurrent relevant measures and social capital (N = 328).

dependent variable	Total Social Capital			Bridging Social Capital			Bonding Social Capital
predictor	B(SE)	95% CI		B(SE)	95% CI		B(SE)	95% CI
		LL	UL		LL	UL		LL	UL
Social participation-frequency	.47*(.05)	.38	.56	.34*(.05)	.24	.43	.49*(.05)	.40	.59
	F(4.323) = 44.33, p = .000			F(4,323) = 34.51, p=.000			F(4,323) = 31.80, p=.000
Social participation-diversity	.45*(.05)	.35	.54	.35*(.05)	.25	.44	.44*(.05)	.34	.54
	F(4,323) = 38.57, p = .000			F(4,323) = 34.37, p = .000			F(4,323) = 23.44, p = .000

Note: Regression analysis, with standardized scores of included core variables. Model: Enter, *p≤ .001. Education, financial resources and participation in volunteer work were included as confounders in all analyses.

Personal social capital, mental well-being and loneliness

Results indicated that overall personal social capital was significantly associated with emotional well-being (F = 5.65, p < .001 and B = .16, p = .004); psychological well-being (F = 9.94, p < .001 and B = .28, p < .001) and social well-being (F = 16.36, p < .001 and B = .43, p < .001). In addition, female respondents experienced higher levels of emotional and psychological well-being than male respondents did and respondents without a life partner experienced more psychological well-being than respondents with a life partner.

In Table 6 the results of three hierarchical regression analyses are presented, in which bridging and bonding social capital were both separately included, to recognise differences in their contribution to the explained variance of respectively emotional, psychological and social well-being. Bonding social capital was positively associated with all three subscales of well-being, whilst bridging social capital was significant only in the model explaining variance in social well-being.

Table 6. Results of regression analysis of association between bonding and bridging social capital and well-being (sub)scales (N = 328).

dependent variable	Emotional well-being			Psychological well-being			Social well-being
predictor	B(SE)	95% CI		B(SE)	95% CI		B(SE)	95% CI
		LL	UL		LL	UL		LL	UL
age	−.02 (.01)	−.04	.00	−.02 (.01)	−.04	.00	−.01 (.01)	−.04	.01
gender	.28* (.12)	.03	.52	.43*** (.12)	.19	.67	.13 (.12)	−.09	.36
partner	−.04 (.14)	−.32	.23	–.28* (.14)	−.55	−.02	−.23 (.31)	−.49	.02
health	.15* (.06)	.03	.27	.08 (.06)	−.04	.20	.05 (.06)	−.07	.16
bonding social capital	.22** (.06)	.09	.34	.23*** (.06)	.11	.35	.25* (.06)	.13	.36
bridging social capital	−.03 (.06)	−.15	.09	.10 (.06)	−.02	.22	.25* (.06)	.13	.36
	Adj. R^2 = .08 F(6,321)=5.71, p=.000			Adj. R^2 = .12 F(6,321)=8.62, p=.000			Adj. R^2 = .19 F(6,321)=13.61, p=.000

Note: Model: Enter, *p ≤ .05, **p ≤ .01, ***p ≤ .001. Standardised scores were included for emotional well-being, psychological well-being, social well-being, bonding social capital and bridging social capital.

Results of the mediation analyses indicated that the associations between bonding social capital and the three subscales of well-being were mediated by loneliness (Table 7; corrected for bridging social capital). Only the direct association between bonding social capital and social well-being was significant (B = .14, p = .043), which indicates a partial mediation for this association. In this model, bridging social capital was also a significant factor (B = .25, p < .001). The association between bridging social capital and social well-being was not found to be mediated by loneliness.

Table 7. Results of mediation analysis (PROCESS) with loneliness as mediating variable in the association between bonding social capital and subscales of well-being (N = 328).

dependent variable	Emotional well-being			Psychological well-being			Social well-being
predictor	B(SE)	95%CI		B(SE)	95% CI		B(SE)	95% CI
		LL	UL		LL	UL		LL	UL
Bonding social capital	.06 (.07)	−.08	.19	.06 (.07)	−.07	.19	.14* (.07)	.00	.27
Loneliness	−.29*** (.06)	−.41	−.17	−.31*** (.06)	−.43	−.19	−.21*** (.06)	−.32	−.09
Bridging social capital	−.02 (.06)	−.14	.19	.11 (.06)	−.00	.23	.25*** (.06)	.14	.37
Age	−.03* (.01)	−.05	−.01	−.03* (.01)	−.05	−.01	−.02 (.01)	−.04	.00
Gender	.20 (.12)	−.04	.43	.34** (.12)	.12	.57	.08 (.11)	−.15	.30
Partner	−.15 (.14)	−.42	.12	−.40** (.13)	−.66	−.14	−.31* (.13)	−.57	.05
Health	.13* (.06)	.02	.25	.06 (.06)	.05	.18	.03 (.06)	−.08	.15
Mediating variable: Loneliness
	B (SE)	95%CI		B (SE)	95%CI		B (SE)	95%CI
		LL	UL		LL	UL		LL	UL
Total	.22 (.06)	.09	.34	.23 (.06)	.11	.35	.25 (.06)	.13	.36
Indirect path	.16 (.05)	.07	.25	.17 (.04)	.09	.25	.11 (.03)	.05	.18
Sobel test	z = 4.41***			z = .4.66***			z = 3.33***

Note: PROCESS Mediation analysis (model 4) with standardised scores of included core variables. *p ≤ .05, **p ≤ .01, ***p ≤ .001. Number of bootstrap samples: 5000.

Discussion

The aim of the current study was to add to the limited empirical research of possible benefits of personal social capital for the well-being of elderly. The previously validated personal social capital scale for a general population (Chen et al., 2009) was translated and adjusted into the personal social capital scale for elderly (PSCSE). The intended two-factor structure as well as the construct and criterion validity were examined, before the assumed associations between personal bonding and bridging social capital on the one hand and social, emotional and psychological well-being, on the other, were explored.

The two-factor structure of the PSCSE was confirmed by our exploratory and confirmatory factorial analysis, although one item loaded on both bonding and bridging social capital, resulting in a sufficient CFI-score, but a less sufficient GFI-score (Hu & Bentler, 1999), which we will refer to later in this section. Subsequently, construct validity was supported by our findings indicating that, in line with previous research, individuals with higher education, more financial resources or participating in volunteer work possessed more social capital than individuals with lower education or less financial resources (Han et al., 2015; Zhang et al., 2011) or individuals that were not involved in any volunteer work (Theurer & Wister, 2010; Zhang et al., 2009). Although bridging social capital in particular is found to be of economic value (i.e. Zhang et al., 2011; Beugelsdijk & Smulders, 2003), in our sample financially secure respondents also scored higher on bonding social capital. Finally, criterion validity was supported by significant associations between bonding and bridging social capital and both frequency and diversity of social participation (Engbers et al., 2017; Hooghe & Vanhoutte, 2011).

Examination of our hypotheses shed light on possible differences in the importance of either bonding social capital or bridging social capital with regard to mental well-being of older adults. Results indicate that social bonding capital is positively associated with all three subscales of mental well-being, while bridging social capital is merely associated with social well-being. This finding supports the assumption that bonding social capital may be potentially more important for mental well-being of elderly than bridging social capital. This follows the earlier mentioned socioemotional selectivity theory (Carstensen et al., 2003) arguing that people tend to be more selective at an older age, as a strategy to maintain their well-being, and prefer social relationships with trusted others, that are emotionally important to them (Carstensen et al., 2006).

The hypothesis that this association is mediated by the experience of loneliness was confirmed as well, adding to recent findings of loneliness as a mediator in associations between concepts related to social capital and mental well-being of older adults (Chen & Feeley, 2014; Santini et al., 2016). Although the cross-sectional design of this study does not allow for conclusions of causality, this finding can be interpreted as support of the assumption that bonding social capital may help prevent feelings of loneliness and by doing this, help maintain or even enhance mental well-being.

Finally, the negative association between having a life partner and psychological well-being needs some attention, since this is not in line with previous research (Hooghe & Vanhoutte, 2011). Although our data does not provide other information for further explanation, at this age chances are that someone's life partner may be less capable of physical and/or mental activities, due to aging or illness, which has been found to have an impact on someone's quality of life (Rees, O'Boyle, & MacDonagh, 2001).

Critical notes and further research

We hope to have contributed to the development of a valid and reliable instrument to measure personal social capital of elderly and to a further understanding of its potential value for well-being in older age. However, some critical remarks should be made. In this study more male than female respondents were included. Although gender was not found to be a significant factor explaining variance in social capital, former research suggests that in a general population, men may possess more social capital than women (Skrabski, Kopp, & Kawachi, 2004; Lindstrom, 2005). Although this might be different for an elderly population, the possible influence of gender in the relationship between social capital and mental well-being, should be taken account for in further research, especially as in our sample the female respondents reported higher levels of well-being than our male respondents did.

With our method of data collection, we could not obtain accurate information about the total sampling frame and survey response rate. Secondly using an online data collection method has excluded elderly who have not stepped into the digital world yet. Furthermore, the sample consisted merely of respondents that live independently in their own home. Elderly, living in a care or retirement home, or other facilities where they receive some kind of care, were not yet included. Considering moving to such a facility may have quite an impact on an individual’s social environment (Chipps & Jarvis, 2016), further research of samples varying in housing circumstances, can both add to our understanding of the value of social capital and may result in further refinements of our scale.

Although our results show high internal consistency scores, contrary to the original scale by Chen et al. (2009), the item addressing the economic value of the relationships with family, friends and acquaintances, seemed less relevant for measuring bonding social capital of older adults. As in later stages of life the emotional value of relationships become more important (Carstensen et al., 2000; Carstensen et al., 2003), there might be less acknowledgement of the economic value of closer social relationships in this age group. To assure that our analyses were reliable, we conducted the exact same analyses without this item (χ² (26, n = 326) = 115.900, p < .001, GFI = .93, CFI = .95, and SRMR = 0.04) and found comparable results, except for a not significant difference in bonding social capital between respondents that participated in volunteer work and those that did not. In the current study it was therefor decided to keep the item. However, based on our results, both preservation and removal of the item is defensible in future research in elderly populations.

Finally, as also indicated earlier in this section, our cross-sectional design does not allow for any conclusions regarding the direction of our found associations, and any inference about cause and effect, based on the current findings, remains therefore speculative.

Conclusion

Well-being of the rapidly growing population of elderly and factors that may add to a positive aging process, are increasingly gaining attention in research literature. Our results indicate that personal social capital and bonding social capital in particular, may well be among these factors, especially considering the assumption that aging is accompanied with an increasing need for social emotional resources (Carstensen et al., 2006). This need may be strengthened by older age-related events such as an increase of physical inabilities, decreased mobility or loss of loved ones, as well as findings that social networks and participation tend to decrease with age (Forte, 2009), probably due to the very same events. Development of interventions focusing on protecting or reinforcing bonding social capital in older age, can help maintain mental well-being and facilitate processes of positive aging. One area that seems promising to this end, is that of social media applications, as these can overcome some of the challenges which elderly are facing (i.e. decreased mobility and social participation). Although the research literature on the value of information and communication technologies for enhancing social capital of elderly is currently limited (Barbosa Neves, Fonseca, Amaro & Pasqualotti, 2018), these continuously developing technologies can be expected to offer a wide variety of possibilities (Chen & Li, 2017; Utz & Muscanell, 2015), considering they become increasingly intertwined with our lives.

Disclosure statement

No potential conflict of interest was reported by the authors.