The 58 reference contexts in paper Daniel Kim, Christopher F Baum, Michael Ganz, S.V. Subramanian, Ichiro Kawachi (2011) “The contextual effects of social capital on health: a cross-national instrumental variable analysis” / RePEc:boc:bocoec:786

  1. Start
    3677
    Prefix
    Introduction The notion of societal conditions and the social environment as fundamental causes of health and disease is not new, dating back more than a century to the works of Durkheim and Virchow
    Exact
    (Link & Phelan, 1995; Cassel, 1976; Durkheim, 1897; Virchow, 1848).
    Suffix
    Social capital, a major attribute of the social environment, has garnered scientific and government attention over the last decade as a plausible broad determinant of population health, educational outcomes, and economic growth (Kawachi, Subramanian, & Kim, 2007).
    (check this in PDF content)

  2. Start
    4124
    Prefix
    Social capital, a major attribute of the social environment, has garnered scientific and government attention over the last decade as a plausible broad determinant of population health, educational outcomes, and economic growth (Kawachi, Subramanian, & Kim, 2007). While social capital may be a property of contexts/collectives or individuals (Kawachi, Subramanian, & Kim, 2007;
    Exact
    Kim & Kawachi, 2007; Kim et al., 2006; Kawachi & Berkman, 2000), and
    Suffix
    has a strong tradition of being conceptualized as the resources embedded within an individual’s social network (Bourdieu, 1977; Lin, 2001; Flap 1991), its novelty lies at the former collective level (Kim & Kawachi, 2007; Kawachi et al., 2004), where it has been defined, according to some scholars, as the features of social organisation, including trust, civic
    (check this in PDF content)

  3. Start
    4328
    Prefix
    While social capital may be a property of contexts/collectives or individuals (Kawachi, Subramanian, & Kim, 2007; Kim & Kawachi, 2007; Kim et al., 2006; Kawachi & Berkman, 2000), and has a strong tradition of being conceptualized as the resources embedded within an individual’s social network
    Exact
    (Bourdieu, 1977; Lin, 2001; Flap 1991),
    Suffix
    its novelty lies at the former collective level (Kim & Kawachi, 2007; Kawachi et al., 2004), where it has been defined, according to some scholars, as the features of social organisation, including trust, civic participation, and reciprocity norms facilitating cooperation for mutual benefit (Kim & Kawachi, 2007; Putnam, 2000; Kawachi et al., 1997).
    (check this in PDF content)

  4. Start
    4417
    Prefix
    be a property of contexts/collectives or individuals (Kawachi, Subramanian, & Kim, 2007; Kim & Kawachi, 2007; Kim et al., 2006; Kawachi & Berkman, 2000), and has a strong tradition of being conceptualized as the resources embedded within an individual’s social network (Bourdieu, 1977; Lin, 2001; Flap 1991), its novelty lies at the former collective level
    Exact
    (Kim & Kawachi, 2007; Kawachi et al., 2004),
    Suffix
    where it has been defined, according to some scholars, as the features of social organisation, including trust, civic participation, and reciprocity norms facilitating cooperation for mutual benefit (Kim & Kawachi, 2007; Putnam, 2000; Kawachi et al., 1997).
    (check this in PDF content)

  5. Start
    4672
    Prefix
    an individual’s social network (Bourdieu, 1977; Lin, 2001; Flap 1991), its novelty lies at the former collective level (Kim & Kawachi, 2007; Kawachi et al., 2004), where it has been defined, according to some scholars, as the features of social organisation, including trust, civic participation, and reciprocity norms facilitating cooperation for mutual benefit
    Exact
    (Kim & Kawachi, 2007; Putnam, 2000; Kawachi et al., 1997).
    Suffix
    At the area or contextual level, social capital may serve as a “public good”, with positive spillover effects onto the health of members of broader society (Putnam, 2000). For example, collective action across a country may mobilise to enact health-promoting policies with potential benefits to all citizens.
    (check this in PDF content)

  6. Start
    4892
    Prefix
    , as the features of social organisation, including trust, civic participation, and reciprocity norms facilitating cooperation for mutual benefit (Kim & Kawachi, 2007; Putnam, 2000; Kawachi et al., 1997). At the area or contextual level, social capital may serve as a “public good”, with positive spillover effects onto the health of members of broader society
    Exact
    (Putnam, 2000).
    Suffix
    For example, collective action across a country may mobilise to enact health-promoting policies with potential benefits to all citizens. Comparative work by the political scientist Robert Putnam favors such a policy-related mechanism.
    (check this in PDF content)

  7. Start
    5676
    Prefix
    Leveraging social capital may therefore be a powerful means to improve population health. Furthermore, the adverse effects of income inequality have been posited to take place, at least in part, through the erosion of social capital/cohesion
    Exact
    (Kondo et al., 2009;
    Suffix
    Kawachi, I., 2000; Marmot & Wilkinson, 2001; Wilkinson & Pickett, 2009; Kim et al., 2008). At the individual level, social capital may yield beneficial private health returns to personal investments (Kawachi & Berkman, 2000)—for instance, participation in a civic group boosting one’s health through psychosocial processes such as social support (Kim & Kawachi, 2006).
    (check this in PDF content)

  8. Start
    5748
    Prefix
    Furthermore, the adverse effects of income inequality have been posited to take place, at least in part, through the erosion of social capital/cohesion (Kondo et al., 2009; Kawachi, I., 2000; Marmot & Wilkinson, 2001;
    Exact
    Wilkinson & Pickett, 2009; Kim et al., 2008).
    Suffix
    At the individual level, social capital may yield beneficial private health returns to personal investments (Kawachi & Berkman, 2000)—for instance, participation in a civic group boosting one’s health through psychosocial processes such as social support (Kim & Kawachi, 2006).
    (check this in PDF content)

  9. Start
    5904
    Prefix
    Furthermore, the adverse effects of income inequality have been posited to take place, at least in part, through the erosion of social capital/cohesion (Kondo et al., 2009; Kawachi, I., 2000; Marmot & Wilkinson, 2001; Wilkinson & Pickett, 2009; Kim et al., 2008). At the individual level, social capital may yield beneficial private health returns to personal investments
    Exact
    (Kawachi & Berkman, 2000)
    Suffix
    —for instance, participation in a civic group boosting one’s health through psychosocial processes such as social support (Kim & Kawachi, 2006). Although multiple studies have investigated the relations between social capital at a contextual level (i.e., at the level of entire countries, states, or neighbourhoods/communities) and general health and disease-specific outcomes (e.
    (check this in PDF content)

  10. Start
    6056
    Prefix
    At the individual level, social capital may yield beneficial private health returns to personal investments (Kawachi & Berkman, 2000)—for instance, participation in a civic group boosting one’s health through psychosocial processes such as social support
    Exact
    (Kim & Kawachi, 2006).
    Suffix
    Although multiple studies have investigated the relations between social capital at a contextual level (i.e., at the level of entire countries, states, or neighbourhoods/communities) and general health and disease-specific outcomes (e.g., cardiovascular disease, cancer), findings have been conflicting to date (Lynch et al., 2001; Kennelly, O’Shea, & Garvey, 2003; Mansyur et al., 2008; Hel
    (check this in PDF content)

  11. Start
    6382
    Prefix
    Although multiple studies have investigated the relations between social capital at a contextual level (i.e., at the level of entire countries, states, or neighbourhoods/communities) and general health and disease-specific outcomes (e.g., cardiovascular disease, cancer), findings have been conflicting to date
    Exact
    (Lynch et al., 2001;
    Suffix
    Kennelly, O’Shea, & Garvey, 2003; Mansyur et al., 2008; Helliwell & Putnam, 2004; Gundelach & Kreiner, 2004; Kim, Subramanian, & Kawachi. 2007; Almedom & Glandon D, 2007). A systematic literature review (Kim, Subramanian, & Kawachi, 2007) found that multilevel studies showed weaker, modest associations for contextual trust than individual-level trust, and that the former estimates be
    (check this in PDF content)

  12. Start
    6436
    Prefix
    Although multiple studies have investigated the relations between social capital at a contextual level (i.e., at the level of entire countries, states, or neighbourhoods/communities) and general health and disease-specific outcomes (e.g., cardiovascular disease, cancer), findings have been conflicting to date (Lynch et al., 2001; Kennelly, O’Shea, & Garvey, 2003;
    Exact
    Mansyur et al., 2008; Helliwell & Putnam, 2004; Gundelach & Kreiner, 2004;
    Suffix
    Kim, Subramanian, & Kawachi. 2007; Almedom & Glandon D, 2007). A systematic literature review (Kim, Subramanian, & Kawachi, 2007) found that multilevel studies showed weaker, modest associations for contextual trust than individual-level trust, and that the former estimates became attenuated to nonsignificance after controlling for individual levels of trust.
    (check this in PDF content)

  13. Start
    6632
    Prefix
    ) and general health and disease-specific outcomes (e.g., cardiovascular disease, cancer), findings have been conflicting to date (Lynch et al., 2001; Kennelly, O’Shea, & Garvey, 2003; Mansyur et al., 2008; Helliwell & Putnam, 2004; Gundelach & Kreiner, 2004; Kim, Subramanian, & Kawachi. 2007; Almedom & Glandon D, 2007). A systematic literature review (Kim, Subramanian, &
    Exact
    Kawachi, 2007)
    Suffix
    found that multilevel studies showed weaker, modest associations for contextual trust than individual-level trust, and that the former estimates became attenuated to nonsignificance after controlling for individual levels of trust.
    (check this in PDF content)

  14. Start
    7000
    Prefix
    Subramanian, & Kawachi, 2007) found that multilevel studies showed weaker, modest associations for contextual trust than individual-level trust, and that the former estimates became attenuated to nonsignificance after controlling for individual levels of trust. Few studies have explored associations between social capital measured at the country level and individual self-rated health
    Exact
    (Helliwell & Putnam, 2004; Mansyur et al., 2008; Poortinga, 2006;
    Suffix
    Rostila, 2007). For example, Helliwell & Putnam (2004) and Mansyur et al. (2008) examined average country levels of social trust in relation to individual self-rated health using data from 49 and 45 countries respectively in the European and World Values Surveys.
    (check this in PDF content)

  15. Start
    7105
    Prefix
    Few studies have explored associations between social capital measured at the country level and individual self-rated health (Helliwell & Putnam, 2004; Mansyur et al., 2008; Poortinga, 2006; Rostila, 2007). For example,
    Exact
    Helliwell & Putnam (2004) and Mansyur et al. (2008)
    Suffix
    examined average country levels of social trust in relation to individual self-rated health using data from 49 and 45 countries respectively in the European and World Values Surveys. Based on least squares regression models, these studies found that higher country-level social trust was associated and not associated with better self-rated health, respectively (Helliwell &
    (check this in PDF content)

  16. Start
    7538
    Prefix
    al. (2008) examined average country levels of social trust in relation to individual self-rated health using data from 49 and 45 countries respectively in the European and World Values Surveys. Based on least squares regression models, these studies found that higher country-level social trust was associated and not associated with better self-rated health, respectively
    Exact
    (Helliwell & Putnam, 2004; Mansyur et al., 2008).
    Suffix
    While the findings to date would appear to discount the utility of leveraging contextual social capital to improve population health, their interpretation is challenged by a fundamental concern which plagues this burgeoning literature: all of these studies, which have been observational in design, have relied on conventional regression estimates, and are prone to bias because the exposur
    (check this in PDF content)

  17. Start
    8451
    Prefix
    bias may occur if one estimates country-level associations of social capital with health, but fails to account for unobserved country characteristics correlated/co-varying with social capital, leading to a spurious statistical relationship. By isolating the random variation in exposures, instrumental variables can overcome such bias and can yield more valid effect estimates
    Exact
    (Wooldridge, 2008).
    Suffix
    This technique has proven useful in addressing confounding and reverse causation issues to better quantify the causal roles of other factors in medicine and public health, including obesity and neighborhood conditions (Davey Smith et al., 2009; Kamstrup et al., 2009; Fish et al., 2010).
    (check this in PDF content)

  18. Start
    8715
    Prefix
    This technique has proven useful in addressing confounding and reverse causation issues to better quantify the causal roles of other factors in medicine and public health, including obesity and neighborhood conditions (Davey Smith et al., 2009;
    Exact
    Kamstrup et al., 2009; Fish et al., 2010).
    Suffix
    Our goal was to estimate the causal association between average country levels of social trust and individual self-rated health using instrumental variables (IV), and to compare these findings with those derived using conventional methods.
    (check this in PDF content)

  19. Start
    9258
    Prefix
    Given the few studies of social capital at the country level, our analyses tested the prior claims that country-level social capital is beneficial to health. In light of past evidence of stronger associations for contextual social capital in women (versus men)
    Exact
    (Kim & Kawachi, 2006; Kim & Kawachi, 2007) and
    Suffix
    individuals with higher (versus lower) levels of social capital (Poortinga, 2006), we further explored whether countrylevel trust might relate variably to health across selected population sub-groups.
    (check this in PDF content)

  20. Start
    9375
    Prefix
    In light of past evidence of stronger associations for contextual social capital in women (versus men) (Kim & Kawachi, 2006; Kim & Kawachi, 2007) and individuals with higher (versus lower) levels of social capital
    Exact
    (Poortinga, 2006),
    Suffix
    we further explored whether countrylevel trust might relate variably to health across selected population sub-groups. Using these new effect estimates, we then quantified the numbers of deaths that might be avoided by elevating average social trust across nations.
    (check this in PDF content)

  21. Start
    10316
    Prefix
    These are repeated cross-sectional surveys (1981-1984, 1990-1993, 1995-1997, and 1999-2004) of nationally representative samples (ranging in size between 500 and 2000 individuals, conducted through stratified random sampling) of the general population aged •18 years across countries on all six inhabited continents
    Exact
    (World Values Survey Group, 2009).
    Suffix
    Outcome variable Individual self-rated health was measured on a five-point ordinal scale, ranging from “very poor” to “very good” health. In prospective studies, global self-rated health has independently predicted morbidity and mortality, with a graded relationship between successive categorical ratings of health and probability of mortality (Idler & Benyamini, 1997; D
    (check this in PDF content)

  22. Start
    10709
    Prefix
    In prospective studies, global self-rated health has independently predicted morbidity and mortality, with a graded relationship between successive categorical ratings of health and probability of mortality
    Exact
    (Idler & Benyamini, 1997; DeSalvo et al., 2005).
    Suffix
    Self-rated health was treated as a continuous measure in all models. Predictor variable Social trust was measured using the following item which has been incorporated into past major social surveys (General Social Surveys, 2007; The Roper Center for Public Opinion Research, 2002), and has exhibited more robust associations with health outcomes than other social capital measures e.g.,
    (check this in PDF content)

  23. Start
    12263
    Prefix
    Country-level covariates consisted of real Gross Domestic Product (GDP) per capita (Heston, Summers, & Aten, 2009), logarithm of total population size (Census Bureau, 2009), total health expenditures per capita (World Health Organisation), ethnic heterogeneity (reflecting the probability of two randomly selected persons belonging to disparate ethnic and language groups)
    Exact
    (Alesina et al., 2003); and
    Suffix
    the Gini coefficient (a measure of income inequality, with values ranging from 0 for perfect equality to 1 for perfect inequality) (World Institute for Development Economics Research, 2005). As previously described, we used standardised criteria for the Gini coefficient (Kim et al., 2008).
    (check this in PDF content)

  24. Start
    12422
    Prefix
    Census Bureau, 2009), total health expenditures per capita (World Health Organisation), ethnic heterogeneity (reflecting the probability of two randomly selected persons belonging to disparate ethnic and language groups) (Alesina et al., 2003); and the Gini coefficient (a measure of income inequality, with values ranging from 0 for perfect equality to 1 for perfect inequality)
    Exact
    (World Institute for Development Economics Research, 2005).
    Suffix
    As previously described, we used standardised criteria for the Gini coefficient (Kim et al., 2008). Due to political and economic changes and/or instabilities in eastern bloc countries and Germany, Hungary, and Poland in the early 1990s, we excluded Gini coefficients for these countries prior to 1995.
    (check this in PDF content)

  25. Start
    12575
    Prefix
    selected persons belonging to disparate ethnic and language groups) (Alesina et al., 2003); and the Gini coefficient (a measure of income inequality, with values ranging from 0 for perfect equality to 1 for perfect inequality) (World Institute for Development Economics Research, 2005). As previously described, we used standardised criteria for the Gini coefficient
    Exact
    (Kim et al., 2008).
    Suffix
    Due to political and economic changes and/or instabilities in eastern bloc countries and Germany, Hungary, and Poland in the early 1990s, we excluded Gini coefficients for these countries prior to 1995.
    (check this in PDF content)

  26. Start
    13427
    Prefix
    To be ‘relevant’, such instruments must be correlated with the endogenous exposure, in order to capture adequate variation in it. To be valid or ‘exogenous’, the instruments must have no direct effect on the outcome, to pick up only random variation in the endogenous exposure
    Exact
    (Wooldridge, 2008; Martens et al., 2006;
    Suffix
    Angrist & Krueger, 2001). Hence, for our study, each instrumental variable had to be closely associated with country-level social trust, and to be unassociated with self-rated health except indirectly through its association with social trust.
    (check this in PDF content)

  27. Start
    14138
    Prefix
    Both theoretical grounding and empirical evidence support each of these instruments as a determinant of social trust. Public institutions arguably play pivotal roles in facilitating interpersonal trust
    Exact
    (Levi, 1996; Yamagishi & Yamagishi, 1994).
    Suffix
    Corruption in public institutions may diminish public trust because the costs of trust may increase when there is a greater perceived risk of being cheated (You, 2005). In a cross-national study, higher perceived country corruption predicted lower individual trust (You, 2005).
    (check this in PDF content)

  28. Start
    14358
    Prefix
    Public institutions arguably play pivotal roles in facilitating interpersonal trust (Levi, 1996; Yamagishi & Yamagishi, 1994). Corruption in public institutions may diminish public trust because the costs of trust may increase when there is a greater perceived risk of being cheated
    Exact
    (You, 2005).
    Suffix
    In a cross-national study, higher perceived country corruption predicted lower individual trust (You, 2005). Using structural equation models, the standardised coefficient estimate for one’s level of confidence in government (conceptually inversely related to the level of corruption) as a predictor of individual trust was six times larger than the estimated relation in the reverse di
    (check this in PDF content)

  29. Start
    14469
    Prefix
    Corruption in public institutions may diminish public trust because the costs of trust may increase when there is a greater perceived risk of being cheated (You, 2005). In a cross-national study, higher perceived country corruption predicted lower individual trust
    Exact
    (You, 2005).
    Suffix
    Using structural equation models, the standardised coefficient estimate for one’s level of confidence in government (conceptually inversely related to the level of corruption) as a predictor of individual trust was six times larger than the estimated relation in the reverse direction (Brehm & Rahn, 1997).
    (check this in PDF content)

  30. Start
    14761
    Prefix
    Using structural equation models, the standardised coefficient estimate for one’s level of confidence in government (conceptually inversely related to the level of corruption) as a predictor of individual trust was six times larger than the estimated relation in the reverse direction
    Exact
    (Brehm & Rahn, 1997).
    Suffix
    This evidence is in keeping with one’s confidence in government (and by relation, perceived lack of corruption) raising one’s level of trust, as opposed to an association in the opposite direction i.e., one’s level of trust leading to greater confidence in government.
    (check this in PDF content)

  31. Start
    15287
    Prefix
    Higher population densities plausibly create public distrust because “dense” social environments may incite individuals to draw inward out of the need for privacy. Empirical evidence supports this relation
    Exact
    (Brueckner & Largey, 2006;
    Suffix
    Pew Research Center, 2007; Collier, 1998). For instance, in behavioral experiments in US cities, lower population density has been associated with acts of helpfulness (a concept related to reciprocity and social trust) shown by local residents to strangers (Levine et al., 1994).
    (check this in PDF content)

  32. Start
    15340
    Prefix
    Higher population densities plausibly create public distrust because “dense” social environments may incite individuals to draw inward out of the need for privacy. Empirical evidence supports this relation (Brueckner & Largey, 2006; Pew Research Center, 2007;
    Exact
    Collier, 1998).
    Suffix
    For instance, in behavioral experiments in US cities, lower population density has been associated with acts of helpfulness (a concept related to reciprocity and social trust) shown by local residents to strangers (Levine et al., 1994).
    (check this in PDF content)

  33. Start
    15581
    Prefix
    For instance, in behavioral experiments in US cities, lower population density has been associated with acts of helpfulness (a concept related to reciprocity and social trust) shown by local residents to strangers
    Exact
    (Levine et al., 1994).
    Suffix
    Religious fractionalization, corresponding to the degree of heterogeneity in religious group affiliations, has been independently linked to a lower risk of civil conflict (Collier & Hoeffler, 2004).
    (check this in PDF content)

  34. Start
    15786
    Prefix
    US cities, lower population density has been associated with acts of helpfulness (a concept related to reciprocity and social trust) shown by local residents to strangers (Levine et al., 1994). Religious fractionalization, corresponding to the degree of heterogeneity in religious group affiliations, has been independently linked to a lower risk of civil conflict
    Exact
    (Collier & Hoeffler, 2004).
    Suffix
    One possible mechanism for this relation is that in more fractionalised societies, the heterogeneity may serve to foster stronger forms of trust and social capital of the bridging kind (i.e., stronger ties between individuals who differ by religious affiliation) (Kawachi et al., 2004; Kim, Subramanian, & Kawachi, 2006), and to facilitate greater tolerance of religious differences.
    (check this in PDF content)

  35. Start
    16079
    Prefix
    One possible mechanism for this relation is that in more fractionalised societies, the heterogeneity may serve to foster stronger forms of trust and social capital of the bridging kind (i.e., stronger ties between individuals who differ by religious affiliation)
    Exact
    (Kawachi et al., 2004;
    Suffix
    Kim, Subramanian, & Kawachi, 2006), and to facilitate greater tolerance of religious differences. Country population density data for 1990 were taken from the International Database of the US Census Bureau (Census Bureau, 2009).
    (check this in PDF content)

  36. Start
    16121
    Prefix
    One possible mechanism for this relation is that in more fractionalised societies, the heterogeneity may serve to foster stronger forms of trust and social capital of the bridging kind (i.e., stronger ties between individuals who differ by religious affiliation) (Kawachi et al., 2004; Kim, Subramanian, &
    Exact
    Kawachi, 2006), and to
    Suffix
    facilitate greater tolerance of religious differences. Country population density data for 1990 were taken from the International Database of the US Census Bureau (Census Bureau, 2009). Corruption was measured using the 1999 Corruption Perceptions Index (CPI) (Transparency International, 2009) for perceived country corruption among public officials/politicians in the previous three years ac
    (check this in PDF content)

  37. Start
    16399
    Prefix
    Country population density data for 1990 were taken from the International Database of the US Census Bureau (Census Bureau, 2009). Corruption was measured using the 1999 Corruption Perceptions Index (CPI)
    Exact
    (Transparency International, 2009)
    Suffix
    for perceived country corruption among public officials/politicians in the previous three years according to mult iple international sources; values ranged from 0 (very corrupt) to 10 (very honest).
    (check this in PDF content)

  38. Start
    16910
    Prefix
    Religious fractionalisation was measured using an index reflecting the probability of two randomly selected persons belonging to different religious groups. This index was previously constructed by collapsing country religious affiliation data from the early 1980s
    Exact
    (Barrett, 1982)
    Suffix
    into nine categories (Collier): Catholic, Protestant, Muslim, Jew, Hindu, Buddhist, Eastern Religion (other than Buddhist), and no affiliation. All instrumental variables were modeled as continuous.
    (check this in PDF content)

  39. Start
    18935
    Prefix
    S1*Instrument1i+S2*Instrument2i + S3*Covariate1i + S4*Covariate2i + ei In the second stage using OLS (Equation 2 below), self-rated health (SRHi) was regressed on the predicted value of country social trust (Trusti) and covariates (as in Equation 1). Our reported IV estimate for country trust is algebraically equivalent to the coefficient on country trust in the second stage regression
    Exact
    (Wooldridge, 2008; Baum, 2006).
    Suffix
    (2) SRHi = E0 + E1*Trusti + E2*Covariate1i + E3*Covariate2i + ui The first set of models was based on Country Set A and jo int ly applied the Corruption Perceptions Index and logarithm of population density as instruments for country trust.
    (check this in PDF content)

  40. Start
    19923
    Prefix
    Consequently, for the interaction models (in which interactions between country-level trust and gender or individual trust were tested separately), we incorporated additional instruments: the interactions between gender/individual trust and the original instrumental variables
    Exact
    (Wooldridge, 2008).
    Suffix
    These additional instruments would be correlated with the interaction between country-level trust and gender/individual trust (relevance), and were hypothesized to be associated with self-rated health only through their associations with these interactions (validity/exogeneity).
    (check this in PDF content)

  41. Start
    21123
    Prefix
    We evaluated both the relevance and validity/exogeneity of the instrumental variables. The Kleibergen-Paap rank LM test was used to assess instrument relevance under the null hypothesis that the instruments were uncorrelated with country-level trust
    Exact
    (Kleibergen & Paap, 2006;
    Suffix
    Baum, Schaffer, & Stillman, 2007). Hansen's J test examined for instrument validity/exogeneity under the null hypothesis that the instruments were jointly exogenous (Baum et al., 2007; Hayashi, 2000).
    (check this in PDF content)

  42. Start
    21326
    Prefix
    The Kleibergen-Paap rank LM test was used to assess instrument relevance under the null hypothesis that the instruments were uncorrelated with country-level trust (Kleibergen & Paap, 2006; Baum, Schaffer, & Stillman, 2007). Hansen's J test examined for instrument validity/exogeneity under the null hypothesis that the instruments were jointly exogenous
    Exact
    (Baum et al., 2007; Hayashi, 2000).
    Suffix
    The C statistic evaluated the exogeneity of each instrument individually. The Durbin-Wu Hausman endogeneity test was used to test the endogeneity of country trust (Baum et al., 2007; Hayashi, 2000).
    (check this in PDF content)

  43. Start
    21528
    Prefix
    Hansen's J test examined for instrument validity/exogeneity under the null hypothesis that the instruments were jointly exogenous (Baum et al., 2007; Hayashi, 2000). The C statistic evaluated the exogeneity of each instrument individually. The Durbin-Wu Hausman endogeneity test was used to test the endogeneity of country trust
    Exact
    (Baum et al., 2007; Hayashi, 2000).
    Suffix
    In analyses that used either the logarithm of country population density or corruption/religious fractionalisation index as an instrument while adding the other variable as a covariate, no direct association with self-rated health was seen for any variable when modeled as a covariate (p>0.65 for the association with the logarithm of country population density, corrupt
    (check this in PDF content)

  44. Start
    23309
    Prefix
    trusts others by 10 percentage points (in countries reporting 30-40% average country levels of social trust in the latest survey wave) or 20 percentage points (in countries with ”--\b-FRXQWU\-WUXVW\f<-Drawing on findings from a recent meta-analysis (of 1.92 times higher risk of mortality associated with moving across a four-point scale of self-rated health)
    Exact
    (DeSalvo et al., 2005),
    Suffix
    we estimated the corresponding reduction in individual risk of mortalit y. The population attributable risk (PAR) for mortality was then calculated using the formula PAR = Pe (RRe-1) / 1 + Pe (RRe-1), where Pe represents the prevalence of exposure (= 100%, assuming the entire general population is “exposed” to varying levels of social trust (Kondo et al., 2009; Wilkinson & P
    (check this in PDF content)

  45. Start
    23687
    Prefix
    The population attributable risk (PAR) for mortality was then calculated using the formula PAR = Pe (RRe-1) / 1 + Pe (RRe-1), where Pe represents the prevalence of exposure (= 100%, assuming the entire general population is “exposed” to varying levels of social trust
    Exact
    (Kondo et al., 2009;
    Suffix
    Wilkinson & Pickett. 2009)), and RRe is the relative risk of mortality with exposure. Last, we used the latest available country age-specific mortality data from the WHO (World Health Organisation, 2006) to estimate the annual number of deaths that might be avoided by raising country trust among those aged 15-74 years in nations with trust levels ”--\b.
    (check this in PDF content)

  46. Start
    29315
    Prefix
    both conventional and IV analyses suggest that the effects of country social trust on health may be stronger in women than men, and that among women, the health of trusting individuals may particularly benefit. Comparisons with prior studies Our OLS findings showed mixed agreement with those from past cross-national studies of country social trust and individual self-rated health
    Exact
    (Mansyur et al., 2008; Poortinga, 2006;
    Suffix
    Rostila, 2007; Helliwell & Putnam, 2004;). In a multilevel linear regression analysis of 70,493 respondents in 45 countries participating in the second and third waves of the Values Surveys, Mansyur et al. (2008) determined that country trust was positively though non-significantly related to better individual self-rated health (IRU- --\b- KLJKHU- WUXVW-- ȕ 0.054, P=0.32).
    (check this in PDF content)

  47. Start
    29374
    Prefix
    Comparisons with prior studies Our OLS findings showed mixed agreement with those from past cross-national studies of country social trust and individual self-rated health (Mansyur et al., 2008; Poortinga, 2006; Rostila, 2007;
    Exact
    Helliwell & Putnam, 2004;).
    Suffix
    In a multilevel linear regression analysis of 70,493 respondents in 45 countries participating in the second and third waves of the Values Surveys, Mansyur et al. (2008) determined that country trust was positively though non-significantly related to better individual self-rated health (IRU- --\b- KLJKHU- WUXVW-- ȕ 0.054, P=0.32).
    (check this in PDF content)

  48. Start
    29554
    Prefix
    Our OLS findings showed mixed agreement with those from past cross-national studies of country social trust and individual self-rated health (Mansyur et al., 2008; Poortinga, 2006; Rostila, 2007; Helliwell & Putnam, 2004;). In a multilevel linear regression analysis of 70,493 respondents in 45 countries participating in the second and third waves of the Values Surveys,
    Exact
    Mansyur et al. (2008)
    Suffix
    determined that country trust was positively though non-significantly related to better individual self-rated health (IRU- --\b- KLJKHU- WUXVW-- ȕ 0.054, P=0.32). Likewise, two multilevel studies of nations in the European Social Survey found positive, non-significant associations between higher country social trust and good or very good individual self-rated health (Poortinga,
    (check this in PDF content)

  49. Start
    29956
    Prefix
    Likewise, two multilevel studies of nations in the European Social Survey found positive, non-significant associations between higher country social trust and good or very good individual self-rated health
    Exact
    (Poortinga, 2006;
    Suffix
    Rostila, 2007). By contrast, in an OLS analysis of 83,520 individuals in 49 countries from the first three waves, Helliwell & Putnam (2004) found that higher average levels of country trust were significantly associated with better individual self-rated health (for 10% higher trust: ȕ 0.104, P<0.01).
    (check this in PDF content)

  50. Start
    30093
    Prefix
    Likewise, two multilevel studies of nations in the European Social Survey found positive, non-significant associations between higher country social trust and good or very good individual self-rated health (Poortinga, 2006; Rostila, 2007). By contrast, in an OLS analysis of 83,520 individuals in 49 countries from the first three waves,
    Exact
    Helliwell & Putnam (2004)
    Suffix
    found that higher average levels of country trust were significantly associated with better individual self-rated health (for 10% higher trust: ȕ 0.104, P<0.01). Because these estimates were markedly larger than our estimates, in a supplemental analysis, we restricted our data to the vast majority of shared countries and first three waves.
    (check this in PDF content)

  51. Start
    32202
    Prefix
    Population health benefits and approaches Seminal work by the epidemiologist Geoffrey Rose showed that a small favorable shift in a population’s exposure level may reap small individual benefits yet produce large public health gains—a key concept in public health known as the “prevention paradox”
    Exact
    (Rose, 1985).
    Suffix
    Based on our IV analysis results, raising the percentage of a country’s population that trusts others by 20 percentage points (equivalent to the difference in average trust levels separating Austria and Canada in the mid-1980s) could roughly improve individual self-rated health by 0.25 points on a five-point scale.
    (check this in PDF content)

  52. Start
    32567
    Prefix
    Based on our IV analysis results, raising the percentage of a country’s population that trusts others by 20 percentage points (equivalent to the difference in average trust levels separating Austria and Canada in the mid-1980s) could roughly improve individual self-rated health by 0.25 points on a five-point scale. Findings from a recent meta-analysis
    Exact
    (DeSalvo et al., 2005)
    Suffix
    suggest that this change would translate into an approximately 5.6% lower individual risk of mortality. In turn, this finding would signify a population attributable fraction of 5.3%. The Appendix Table shows the estimated annual number of preventable deaths by raising country trust.
    (check this in PDF content)

  53. Start
    34468
    Prefix
    Interactions Our findings suggest that health may more closely tied to country-level trust among women than among men. A similar pattern has been observed in other studies in the social capital literature
    Exact
    (Kim & Kawachi, 2006; Kim & Kawachi, 2007),
    Suffix
    as well as studies of neighbourhood socioeconomic associations with health behaviours (Wang et al., 2007) and chronic disease outcomes (Diez Roux et al., 2001; Sundquist, Malmstrom, & Johannson, 2004; Kim et al., 2010).
    (check this in PDF content)

  54. Start
    34609
    Prefix
    A similar pattern has been observed in other studies in the social capital literature (Kim & Kawachi, 2006; Kim & Kawachi, 2007), as well as studies of neighbourhood socioeconomic associations with health behaviours
    Exact
    (Wang et al., 2007) and
    Suffix
    chronic disease outcomes (Diez Roux et al., 2001; Sundquist, Malmstrom, & Johannson, 2004; Kim et al., 2010). While plausible explanations for these gender differences exist (for example, a higher responsiveness in women due to gender-related perceptions of social/socioeconomic environments) (Kim et al., 2010), this pattern was less consistently found in our IV anal
    (check this in PDF content)

  55. Start
    34728
    Prefix
    A similar pattern has been observed in other studies in the social capital literature (Kim & Kawachi, 2006; Kim & Kawachi, 2007), as well as studies of neighbourhood socioeconomic associations with health behaviours (Wang et al., 2007) and chronic disease outcomes (Diez Roux et al., 2001; Sundquist, Malmstrom, & Johannson, 2004;
    Exact
    Kim et al., 2010).
    Suffix
    While plausible explanations for these gender differences exist (for example, a higher responsiveness in women due to gender-related perceptions of social/socioeconomic environments) (Kim et al., 2010), this pattern was less consistently found in our IV analysis, and warrants confirmation in other studies.
    (check this in PDF content)

  56. Start
    34955
    Prefix
    While plausible explanations for these gender differences exist (for example, a higher responsiveness in women due to gender-related perceptions of social/socioeconomic environments)
    Exact
    (Kim et al., 2010),
    Suffix
    this pattern was less consistently found in our IV analysis, and warrants confirmation in other studies. The apparent cross-level interact ion in women between trust at the country level and at the individual level is also in keeping with past findings (Poortinga, 2006).
    (check this in PDF content)

  57. Start
    35225
    Prefix
    in women due to gender-related perceptions of social/socioeconomic environments) (Kim et al., 2010), this pattern was less consistently found in our IV analysis, and warrants confirmation in other studies. The apparent cross-level interact ion in women between trust at the country level and at the individual level is also in keeping with past findings
    Exact
    (Poortinga, 2006).
    Suffix
    Underlying mechanisms for this apparent synergism should be explored. Study limitations Our study had several key limitations. First, we did not account for aggregate social trust at more proximal, sub-national levels (e.g., the neighbourhood level), which may have contributed to residual confounding.
    (check this in PDF content)

  58. Start
    37596
    Prefix
    At the same time, the stronger associations with the use of instrumental variables could partly be due to the instrumental variables’ ability to reduce measurement error in country-level trust
    Exact
    (Wooldridge, 2008;
    Suffix
    Angrist & Krueger, 2001). Conclusions In a large cross-national study, our findings, including those using instrumental variables to reduce endogeneity bias, support the presence of beneficial effects of higher country-level social trust on individual self-rated health.
    (check this in PDF content)