Table 3 Empirical evidence by types of studies and countries: Multilevel studies

Sampson, Raudenbush & Earls, 1997 [107]

Cross sectional data came from 1995 Project on Human Development in Chicago Neighborhoods, 8782 individuals in 343 Neighbourhood clusters in Chicago.

Level 1: Individual (micro)

Level 2: Neighborhoods (Meso).

Collective efficacy, defined as social cohesion among neighbors combined with their willingness to intervene on behalf of the common good.

'Aggregated' social capital (individual level responses aggregated to neighborhood level).

Violent crime and homicide in Chicago, the USA.

Collective efficacy negatively associated with neighbourhood variations in violent crime and homicide.

Variance components both within neighborhoods (0.320) and between neighbourhoods (0.026) for collective efficacy estimated and ICC is 7.51%.

Kawachi, Kennedy, & Glass, 1999 [7]

Cross-sectional data among 167,259 respondents came from the Centers for Disease Control Behavioral Risk Factor Surveillance Surveys.

Level 1: Individual (micro)

Level 2: States (Macro).

Using three GSS measures of civic trust, reciprocity (helpfulness of others) and civic engagement (membership in group) and based on these indices states were characterized as high, medium and low social capital

'Aggregated' social capital (individual level responses aggregated to state level).

Self-rated health between US states.

Person living in a state with low levels of social capital had an increased probability of lower self-rated health than someone living in an area of higher social capital.

Variance component for both levels and/or ICC was not reported.

Subramanian, Kawachi & Kennedy, 2001 [108]

Cross-sectional data used from the 1993–94 Behavioral Risk Factor Surveillance System and the 1986–90 General Social Surveys.

Level 1: Individual (micro)

Level 2: States (Macro).

Operationalized as the percent of residents in each state responding that 'other people would try to take advantage of you if they could (mistrust).

'Aggregated' social capital (individual level responses aggregated to state level).

Self-rated health between US states

After controlling for income-inequality and overall income a significant effect of social capital was observed.

Variance component for both levels and/or ICC was not reported.

Subramanian, Kim, & Kawachi, 2002 [14]

Cross-sectional data among 21,456 individuals nested within 40 US communities included in the 2000 Social Capital Community Benchmark Survey.

Level 1: Individual (micro)

Level 2: States (Macro).

Perceptions of individual trust were derived by summing individual responses on (1) general interpersonal trust and (2) degrees of trustworthiness of neighbors, co workers, fellow congregants, store employees where the individual shops, and local police. At the community level, a contextual social trust variable was aggregated from individual responses to questions on interpersonal trust.

'Aggregated' social capital (individual level responses aggregated to state level).

Self-rated health between US states.

High community levels of social trust and self-rated health are positively associated, a significant cross-level interaction effect between community and individual trust also observed.

Variance component for both levels and/or ICC was not reported.

Browning & Cagney, 2002 [109]

Cross sectional data came from 1994 Project on Human Development in Chicago Neighborhoods, 1991–2000 Metropolitan Community Information Center-Metro Survey; 2218 individuals in 333 Neighbourhood clusters in Chicago.

Level 1: Individual (micro)

Level 2: Neighborhoods (Meso).

Collective efficacy such as reciprocity, density of local networking, social cohesion, informal social control used for conceptualizing social capital.

'Aggregated' social capital (individual level responses aggregated to neighborhood level).

Self-rated physical health between Chicago Neighborhoods, the USA.

Higher levels of neighbourhood collective efficacy associated with better self-rated overall health.

Variance component for both levels and/or ICC was not reported.

Wen, Browning & Cagney, 2003 [110]

A cross-sectional data employed from 1990 Decennial Census; the 1994–95 Project on Human Development in Chicago Neighborhoods-Community Survey and the 1991–2000 Metropolitan Chicago Information Center Metro Survey for 8782 individuals in 343 neighborhoods clusters in Chicago.

Level 1: Individual (micro)

Level 2: Neighborhood (Meso).

Collective efficacy such as reciprocity, density of local networking, social cohesion, informal social control used for conceptualizing social capital.

'Aggregated' social capital (individual level responses aggregated to neighborhood level).

Self-rated health in Chicago neighborhoods in the USA.

Neighbourhood social capital associated with better individual self-rated health.

Variance component for both levels and/or ICC was not reported

Franzini & Spears, 2003 [111]

A cross-sectional study based on Texas, USA, in 1991. Using the 1990 US census of total 61,557 heart disease deaths in Texas in 1991 recorded, 54,640 (89%) were linked to the census information by geocoding and the individual's addresses were geocoded to12,344 block-groups, 3788 tracts, and 247 counties in Texas.

Level 1: Individual (micro)

Level 2: Block-group level (Meso)

Level 3: Tract Level (Meso)

Level 4: County (Macro).

Social capital as one of the indicators of social context was operationalized by homeownership (percent of owner-occupied housing units) at the tract and county level and the crime index (defined as number of serious crimes known to police per 100,000 population) at the county level.

Contextual social capital

Premature mortality from heart disease. Years of potential life lost were computed as the 1990 life expectancy in Texas at age when death occurred.

Individual level characteristics were major predictors. Social context at the block-group, tract, and county level plays an important role in explaining years of life lost to heart disease.

Block-group level wealth, tract level own group ethnic density, and county level social capital, had significant effect on years of life lost to heart disease in Texas.

Variance component for both levels and/or ICC was not reported.

Veenstra, 2005 [112]

A cross-sectional study data came from two original data sets, one pertaining to features of 25 communities in British Columbia, Canada and the other to characteristics of individuals living in them. Individual responses (N = 1435) collected from a mailed survey of randomly selected residents aged 18 and higher during the summer and fall of 2002. A random selection of households was drawn from the most current telephone listings using a systematic random sampling technique, and a survey questionnaire was then administered by post in a five-stage process.

Level 1: Individual (micro)

Level 2: Community (Meso).

Individual-level social capital was operationalized through individuals' perception about social and political trust and participation in voluntary associations. To measure attributes of communities the study determined (i) the number of public spaces per capita (sports, recreational, casual and social, cultural, religious, school and hall spaces in particular), (ii) the number of voluntary organizations per capita (sports and athletics, community, minorities, arts and culture, business, political, health and social services, religious and other organizations in particular), and (iii) average levels of community and political trust (aggregates of the trust scales).

contextual social capital

Physical health-long-term illness, health problem or handicap that limits daily activities or the work. Mental health was assessed emotional well-being. Self-rated health (including both physical & mental health).

Household income and political trust were particularly important predictors of long-term illness, but community social capital were mostly irrelevant in this instance The strongest predictors of fair/poor health were age and political trust, followed by income and community level variables were not significantly related to self-rated health.

Only the measure of depressive symptoms had variability that could be reasonably attributed to the community and a mere 2.1% of variability (ICC) could be attributed. The other two measures of health, i.e., the presence of a long-term illness and self-rated health status, were predicted by individual-level factors only.

Drukker, Kaplan, Feron & van Os, 2003 [113]

A longitudinal cohort study of 7236 children and their families in the city of Maastricht 36 neighbourhoods, in the Netherlands.

Level 1: Individual (micro)

Level 2: Neighborhood (Meso).

Social capital was measured using two collective efficacy scales: informal social control, and social cohesion and trust.

'Aggregated' social capital (individual level responses aggregated to neighborhood level).

Children's general health and satisfaction and the mental health and behaviour

Social capital non-specifically associated children's general health and satisfaction. The mental health and behaviour dimensions were more specifically associated with degree of informal social control in the neighborhood.

Variance component for both levels and/or ICC was not reported.

Mohan, Twigg, Barnard & Jones, 2005 [114]

A follow-up study based on English sample of 7578 individuals followed from1984/85 to 2001 modelled individual and ecological data simultaneously and data come from the Health and Lifestyle Survey(HALS)

Level 1: Individual (micro)

Level 2: Electoral wards (Meso).

Used area measurer of social capital on a range of indicators (drawn from various surveys) such as- participation in voluntary activities (from GHS); political activity, social activity, election participation, altruistic activity etc (from BHPS); friendly community and 'community sprit' (from SHE).

'Aggregated' social capital (individual level responses aggregated to electoral wards level).

The probability of individual mortality.

Not found conclusive evidence in support of social capital as a contextual construct which has an influence on health.

Variance component for both levels and/or ICC was not reported.

Lindström, Moghaddassi, & Merlo, 2004 [15]

A cross sectional study data came from the public health survey in Malmö, 1994. A total of 3,602 individuals aged 20–80 years living in 75 Neighbourhoods were considered.

Level 1: Individual (micro)

Level 2: Neighborhood (Meso).

The social participation was used as a proxy for social capital at the individual level. Individual- social participation defined as how actively the person takes part in the activities of formal and informal groups as well as other activities in society during the past 12 months. Items were summed and were classified as having low social participation (score was three or less activities out of 13 items).

'Aggregated' social capital (individual level responses aggregated to neighborhood level).

The influence of neighbourhood and individual factors on self-reported health in the neighborhoods of city of Malmö, Sweden.

The neighborhood level social capital is associated with self-reported health.

The neighborhood variance in self-reported health was mainly influenced by individual factors with 0.0% ICC.

Drukker, Buka, Kaplan Mckenzie & van Os, 2005 [115]

A cross-sectional study based on data from (1) the Project on Human Development in Chicago Neighborhoods (PHDCN), USA and (2) the Maastricht Quality of Life study (MQoL), the Netherlands. For the PHDCN, 874 census tracts were combined to create 343 "neighborhood clusters" (NCs) consisting of approximately 8000 inhabitants each. NC Maastricht consists of 36 residential neighborhoods, housing between 300 and 8500 inhabitants, and all these neighborhoods were selected for the MQoL. Both the PHDCN and the MQoL consisted of a family cohort study as well as a community survey.

Level 1: Individual (micro)

Level 2: Neighborhood cluster/residential neighborhood (Meso).

Subjective neighborhood social capital used and operationalized by perception about informal social control (ISC) and social cohesion and trust (SC&T) that developed by Sampson and colleagues [107] and construct scales consist of 5 items each and respondents answered these on a 5-point Likert scale.

'Aggregated' social capital (individual level responses aggregated to the neighborhood level).

Children's (age11–12) perceived health measured in 5-item Likert type scale.

Chicago had lower levels of SC&T while Maastricht had lower levels of ISC. Higher levels of ISC and SC&T were associated with higher levels of children's perceived health, in both Maastricht and the Chicago Hispanic sub-sample, but not in the Chicago non-Hispanic samples.

Variance component for both levels and/or ICC was not reported.