at SciVerse ScienceDirect

Social Science & Medicine 74 (2012) 1362e1367

Contents lists available

Social Science & Medicine

journal homepage: www.elsevier .com/locate/socscimed

Network social capital, social participation, and physical inactivity in an urbanadult population

Hannah Legh-Jones a, Spencer Moore a,b,*

a School of Kinesiology and Health Studies, Queen’s University, ON, CanadabCentre de recherche du Centre Hospitalier de l’Université de Montréal, Canada

a r t i c l e i n f o

Article history:Available online 22 February 2012

Keywords:Social capitalSocial participationPhysical inactivityCanadaUrban health

* Corresponding author. School of Kinesiology aUniversity, 69 Union St. PEC Rm. 215, KingstoTel.: þ1 613 533 6000x78667; fax: þ1 613 533 2

E-mail address: mooresp@queensu.ca (S. Moore).

0277-9536/$ e see front matter � 2012 Elsevier Ltd.doi:10.1016/j.socscimed.2012.01.005

a b s t r a c t

Research on individual social capital and physical activity has tended to focus on the association amongphysical activity, generalized trust, and social participation. Less is known about the association betweennetwork social capital, i.e., the resources accessed through one’s social connections, and physical inac-tivity. Using formal network measures of social capital, this study examined which specific dimension ofnetwork capital (i.e. diversity, reach and range) was associated with physical inactivity, and whethernetwork social capital mediated the association between physical inactivity and social participation. Datacame from the 2008 Montreal (Canada) Neighbourhood Networks and Healthy Aging survey, in which2707 adults 25 years and older in 300 Montreal neighbourhoods were surveyed. Physical activity wasself-reported using the International Physical Activity Questionnaire (IPAQ). IPAQ guidelines provided thebasis for the physical inactivity cutoff. Network social capital was measured with a position generatorinstrument. Multilevel logistic methods were used to examine the association between physical inac-tivity and individual social capital dimensions, while adjusting for socio-demographic and -economicfactors. Higher network diversity was associated with a decreased likelihood of physical inactivity.Consistent with previous findings, individuals who did not participate in any formal associations weremore likely to be physically inactive compared to those with high levels of participation. Networkdiversity mediated the association between physical inactivity and participation. Generalized trust andthe network components of reach and range were not shown associated with physical inactivity. Findingshighlight the importance of social participation and network social capital and the added value ofnetwork measures in the study of social capital and physical inactivity. Population-based programstargeting physical inactivity among adults might consider ecological-level interventions that leverageassociational involvement and interpersonal relationships to improve population-level physical activity.

� 2012 Elsevier Ltd. All rights reserved.

Introduction

Within health sciences research, there has been a lack ofconsensus concerning the definition and measurement of socialcapital (Kawachi, Subramanian, & Kim, 2008). Two mainapproaches to studying social capital and health have developed:social cohesion (i.e., communitarian) and network capitalapproaches. Social cohesion approaches tend to conceptualize andmeasure social capital as the resources (e.g., trust and norms)available to social groups; network capital approaches tend to focuson the resources (e.g., social support, information channels)

nd Health Studies, Queen’sn, ON, Canada K7L 3N6.009.

All rights reserved.

embedded within an individual’s social networks (Kawachi et al.,2008). Within the network capital approach, the term socialnetwork usually refers to the pattern of a person’s social connec-tions, while network social capital refers to the amount and qualityof resources that a person might access through their socialnetworks. Recent research on social capital and health has sug-gested that the specific mechanisms by which social capital isassociated with health may differ depending on the approach ortype of measure used to operationalize the concept of social capital(Carpiano & Hystad, 2011; Moore et al., 2010, 2011). Measures oftrust, reciprocity, and group normsmay better capture psychosocialmechanisms, whereas network resource indicators may bettercapture network mechanisms linking social capital to health.Identifying the differences in how such measures are associatedwith health and health behavioural outcomes is important forincreasing our understanding of the psychosocial and network

H. Legh-Jones, S. Moore / Social Science & Medicine 74 (2012) 1362e1367 1363

mechanisms potentially linking social capital to health. Few studieshave compared network capital and social cohesion measures toassess the degree to which each is associated with physical inac-tivity. Furthermore, no studies as far as we are aware have assessedthe degree to which network capital mechanisms might mediatethe association between physical inactivity and social participation.

Physical inactivity is defined as a state inwhich body movementis minimal (Dietz, 1996). Persons are classified as being physicallyinactive when they fall below recommended levels of moderateand vigorous energy expenditure. For example, physical activityguidelines posted by the Canadian Society for Exercise Physiologyrecommend that adults 18e64 years accumulate at least 150 min ofmoderate-to-vigorous intensity aerobic activity per week in boutsof 10 or more minutes to achieve health benefits (Canadian Societyfor Exercise Physiology, 2011). Physical inactivity has been shown toincrease the risk of obesity, coronary heart disease, type 2 diabetes,and certain types of cancers (Guthold, Ono, Strong, Chatterji, &Somnath, 2008), and is considered a modifiable risk factor forthese diseases (Warburton, Nicol, & Bredin, 2006).

Published studies of individual social capital and physical (in)activity have shown fairly consistent relationships. Higher levels ofsocial capital have tended to be associated with higher levels ofphysical activity or a lower likelihood of physical inactivity. Studiesusing social trust to reflect social capital have shown that highcompared to low levels of trust tend to reduce the risk of physicalinactivity (OR: 0.58; 95% CIs: 0.42e0.79) (Ueshima et al., 2010). Inaddition, Poortinga (2006) reported that high levels of trustincreased walking (OR: 1.20; 95% CIs: 1.07e1.34) and sports-relatedphysical activity (OR: 1.20; 95% CIs: 1.06e1.36). Individuals whohave higher generalized trust in others tend to be more physicallyactive. Studies that have examined social participation, i.e.,a person’s level of engagement in formal and informal groups, andphysical (in)activity have also been prominent in the literature.Lindstrom et al. showed that low social participationwas associatedwith an increased risk of low-leisure time physical activity in bothmen (OR: 2.20; 95% CIs: 1.90e2.70) and women (OR: 2.20; 95% CIs:1.80e2.50) (Lindstrom, Hanson, & Ostergren, 2001). In anotherstudy, Lindstrom, Moghaddassi, and Merlo (2003) reported thatindividuals with low social participation had much higher odds ofleisure time physical inactivity compared to individuals with highsocial participation (OR: 3.59; 95% CIs: 2.95e4.35). Poortinga(2006) also showed that individuals with medium and high levelsof civic participation had an increased likelihood of walking, sports-related physical activity, and overall activity levels. Similarly,Greiner, Li, Kawachi, and Hunt (2004) found that individuals whoreported more community involvement were more likely to bephysically active that those who reported low communityinvolvement (OR: 1.62; 95% CIs: 1.31e2.01).

Current research on social capital and physical (in)activity hasbeen based mainly on measures of social trust and participation.Social trust has been identified as belonging mainly to a category ofpsychosocial mechanisms (Abbot & Freeth, 2008). Social partici-pation has been recognized as a type of behavioural/structuralaspect of social capital that facilitates the development of one’ssocial networks and sense of social integration (Swaroop &Morenoff, 2006; Szreter & Woolcock, 2004). Besides psychosocialand behavioral mechanisms, social capital may also benefit healththrough mechanisms related to norms and attitudes and socialnetworks (Kawachi, Kennedy, & Glass, 1999). Network measures ofsocial capital potentially have higher content validity when itcomes to capturing people’s social connections and resourceaccessibility, and have been shown to have a different associationwith obesity and self-reported health than trust and participation(Moore, Daniel, Paquet, Dube, & Gauvin, 2009; Moore et al., 2011).Hence, one of the limitations in current research on social capital

and physical (in)activity is knowledge of the degree to whichnetwork social capital is associated with physical (in)activity.Conceptually, it is critical to recognize the diverse dimensions bywhich social capital may be associated with physical inactivity sothat effective public health interventions that target social capitalto improve health can be designed.

Researchers have suggested that one pathway that may linksocial participation to health is through the enhancement of one’ssocial relationships and social capital (Szreter & Woolcock, 2004).In other words, social participation provides opportunities forindividuals to gain access to resources to which they might nototherwise have had access (Szreter & Woolcock, 2004). In thissense, network social capital may in fact mediate the associationbetween physical inactivity and social participation. Yet, to ourknowledge, few studies have actually had the type of networksocial capital data necessary to assess the degree to which networkmechanisms might mediate the association between physicalinactivity and social participation.

The following study uses a network approach to examine theassociation between social capital and physical inactivity. First, thisstudy will examine the association between the three differentdimensions (i.e., diversity, reach and range) of network capital andphysical inactivity to assess which dimension if any is associatedwith physical inactivity. Based on previous findings of networksocial capital and health (Carpiano & Hystad, 2011; Moore et al.,2009, 2011), it is hypothesized that as network capital increases,the likelihood of physical inactivity decreases, with the diversitydimension being potentially the dimension most strongly associ-ated with physical inactivity. Secondly, this study will compare theassociation of physical inactivity with network social capital, trustand participation to assess the relative significance of each inexplaining individual differences in physical inactivity in thissample of adults. Given the lack of studies in this area, there are noa-priori hypotheses about possible differences in the associationsamong the different social capital measures and physical inactivity.Finally, the study will assess the degree to which significantnetwork social capital dimensions mediate or partially mediate theassociation between physical inactivity and social participation.Based on previous studies that have proposed that networkmechanisms mediate the influence of social participation onphysical inactivity, it is hypothesized that network capital will actas a mediating variable in this study.

Method

Sample

Data came from the 2008 Montreal Neighbourhood Networksand Healthy Aging Study (MoNNET-HA). The MoNNET-HA studyused a two-stage cluster sampling design. In stage one, using 2001Canada Census data, Montreal Metropolitan Area (MMA) censustracts (N¼ 862) were stratified into low, medium, and high-incometertiles. From each tertile, one hundred census tracts wereselected (n¼ 300). In stage two of the sampling design, potentialrespondents in each census tract were stratified into three agegroups, 25e44 years, 45e64 years, and 65 years and older.Within each age group and census tract three respondentswere randomly selected for a total of nine respondents per tract. Inseven tracts, four participants were selected for a total samplesize of 2707. To be eligible for the study, participants had to be (1)non-institutionalized, (2) reside at their current address for atleast a year, and (3) be able to complete the questionnaire ineither French or English. Random digit dialling of listedtelephone numbers was used to select households for participation.The questionnaire was administered using a computer-assisted

H. Legh-Jones, S. Moore / Social Science & Medicine 74 (2012) 1362e13671364

telephone interviewing system. The study was approved by theCentre de recherche de l’Universite de Montreal (CRCHUM)Committee of Scientific Evaluation and Research Ethics.

The telephone questionnaire was completed between mid-Juneand early August 2008. The MoNNET-HA response rate was calcu-lated according to American Association for Public Opinion Researchstandard definitions at 38.7%. Chi-square analyses assessing therepresentativeness of the MoNNET-HA sample compared to 2006MMA Canada Census data showed that the MoNNET-HA sampleover-represented (1) older adults (by sampling design), (2) indi-viduals with an income less than 50,000 per year, (3) persons wholived in their places of residence formore thanfiveyears, (4) females,and (5) those with more than a high school degree.

Measures

Main outcome. The main outcome of interest for this study wasphysical inactivity. This was measured using an adapted version ofthe International Physical Activity Questionnaire (IPAQ), andcalculated using IPAQ analysis algorithms and recommendedcutoffs (IPAQ Committee, 2005). The IPAQ incorporates questionsabout the total volume of physical activity and the number of daysin a week that such activity was conducted to calculate the energycosts of activity as the metabolic equivalent of task (MET). Vigorousactivities, converted at 8.0 MET, are activities that take hard phys-ical effort, such as heavy lifting, digging, aerobics, or fast bicycling.Moderate activities (4.0 MET) are activities that make breathingsomewhat harder than normal, such as carrying light loads, bicy-cling at a regular pace, or doubles tennis. Walking activities(3.3 MET) included walking done at work or at home, walking totravel from place to place, and any other walking done solely forrecreation, sport, exercise or leisure.

For each type of activity respondents were asked, “During thelast 7 days, on how many days did you do this type of activity?”Respondents were also asked “howmuch time did you spend doingthis activity on one of those days?” An activity had to have beencarried out for at least 10 minutes to be recorded as such. For eachactivity type (vigorous, moderate, and walking), the activity’s METvalue, the number of days per week, and number of minutes perday were multiplied together. To calculate total MET for eachperson, the values for each activity were then added together. IPAQguidelines were used to classify respondents into high, moderate orlow/inactive physical activity levels. High physical activity wasdefined as at least 3 days of vigorous activity on at least 3 daysaccumulating at least 1500MET-min/week, or 7 ormore days of anycombination of walking, moderate or vigorous activity achievinga minimum of at least 3000 MET-min/week. To be moderatelyactive, one of three criteria had to be met: (1) three or more days ofvigorous activity of at least 20 min per day, (2) five or more days ofmoderate activity or walking of at least 30 min per day, or (3) five ormore days of any combination of walking, moderate or vigorousactivities achieving a minimum of at least 600 MET-min/week.Individuals were classified as being physically inactive if they didnot achieve the criteria for moderate or high physical activity (IPAQResearch Committee, 2005). For this analysis, the physical activityvariable was dichotomized into high/moderate versus low/inactive.

Main exposure

Network Capital. Network capital was measured using a positiongenerator. The position generator measures individuals’ socialcapital by assessing a person’s ties to others working in specifictypes of occupations. Participants were asked to indicate whetherthey know someone on a first name basis who holds a certainoccupation in society. Ten occupations were selected from a listing

of 90 that had been ranked according to gender-neutral job prestigescores within Canada. The list was divided into octiles, ranging inlow and high prestige scores. One occupation from each octile wasrandomly selected, with two additional occupations included.Selected occupations were randomly listed in the position gener-ator (Highschool teacher, carpenter, musician/artist, taxi driver,physician, janitor, registered nurse, welder, accountant, andreceptionist). Scores for reachability, diversity, and range werecalculated and each dimension was considered separately ratherthan as a single measure.

Lin et al. (2001) argue that the position generator captures threedimensions of social capital: (1) reachability, (2) diversity, and (3)range. The three indices of social capital can be calculated using theprestige scores associated with each occupation accessed. Reach-ability represents the hierarchical dimension of social capital, and issimply the most prestigious occupation that a person can reachthrough their social ties. Diversity represents network size asreflected in the number of different occupations accessed. Range isthe difference between the highest and lowest prestige jobaccessed, and reflects the potential degree to which networks“reach out” into the social structure (Haines, Beggs, & Hurlbert,2011).

Generalized trust was assessed using the U.S General SocialSurvey question “Generally speaking, would you say that mostpeople can be trusted or that you can’t be too careful?” Responseswere recorded as: (1) most people can be trusted, (2) can’t be toocareful, (3) depends, (4) most people cannot be trusted. High trustwas classified as answering “most people can be trusted;” low trustwas classified as answering, “can’t be too careful”, “depends”, or“most people cannot be trusted.”

Social participation. To measure social participation, participantswere asked if they had been active as a member or officer in, first ofall, a neighbourhood group or association, and, secondly, any othergroup or association. Responses were recorded as “yes”, “no” “don’tknow” or “no response.” Social participation was analysed asa categorical variable consisting of (1) no social participation, (2)low participation, i.e., involvement in either a neighbourhood orother voluntary association or group, and (3) high participation, i.e.,involvement in both a neighbourhood and other voluntary associ-ation or group.

Covariates. Gender, household income category, educationalattainment, age category, and self-reported health were consideredin this analysis. Participants self-identified as male or female. Agewas stratified into six age categories: 25e34, 35e44, 45e54,55e64, 65e74, and 75 or older. Educational attainment was thehighest level of education that a respondent had completed as ofthe survey date. Respondents could choose one of seven levels: Nodegree, certificate, or diploma; secondary (high) school diploma orequivalent; trades certificate or diploma; College certificate ordiploma below Bachelor’s degree level; University certificate ordiploma at Bachelor’s level; Master’s degree; earned Doctoratedegree; or no response. Individual household income was assessedby asking respondents to identify one of five total householdincome categories: less that $28,000; $28,000e$49,000;$50,000e$74, 000; $75,000e$ 100,000; and above $100,000. Giventhe importance of overall health status for physical activity and itsrole as a potential confounder of the association between socialcapital and physical inactivity, our models also adjusted for self-reported health (SRH) status. SRH was classified as excellent, verygood, good, fair, or poor.

Statistical analysis procedures

Multilevel logistic regression analysis was used to assess theassociations among physical inactivity, generalized trust, social

Table 1Descriptive characteristics of the MoNNET-HA and physical inactivity sample,n¼ 2672.

Variable Percent

Physical activity levelPhysically inactive 17.10

Age group25e34 14.6335e44 17.5845e54 20.1355e64 16.2965e74 20.8775þ 10.49

GenderMale 35.35Female 64.65

Educational attainmentNo degree 11.91High school/Trade 29.19College 20.70University 38.20

Household income<28,000 22.3528,000e49,000 27.6350,000e74,000 23.5175,000e100,000 12.54>100,000 13.98

Self-reported healthExcellent 20.57Very good 34.22Good 31.22Fair 10.65Poor 3.33

Generalized trustHigh trust 42.50Low trust 57.50

ParticipationNo 63.47Low 25.79High 10.75

Network social capital Mean (standard deviation)Diversity 4.28 (2.36)Range 37.12 (20.69)Upper reachability 78.72 (24.05)

H. Legh-Jones, S. Moore / Social Science & Medicine 74 (2012) 1362e1367 1365

participation, and network capital. Analyses showed the intra-classcorrelation coefficient of physical inactivity to be 0.01. Multilevelanalyses were used nevertheless to account for the clusteredsampling design, i.e., individuals nested within tracts, of theMoNNET-HA study. Using the SAS 9.2 glimmix procedure, threesequential multilevel logistic models were fitted. The first modelexamines the association of physical inactivity with the individualsocio-demographic, -economic, and SRH covariates. Model twoadds generalized trust and social participation to assess theirassociation with physical inactivity. The final model introduces thethree dimensions of network social capital into the equation toassess their associations with physical inactivity and compare themwith those of trust and participation. Observations were excluded ifthey were missing information on any study variables.

To examine whether network social capital mediated theassociation between social participation and physical inactivity,multilevel regression and the three-step analysis procedures rec-ommended by Krull and MacKinnon were followed (Krull &Mackinnon, 1999). First, the association between no social partici-pation and network diversity was estimated with network diversitytreated as the outcome variable. The estimated association betweennetwork diversity and no social participationwas represented usinga. Second, the association among physical inactivity, social partici-pation, and network diversity was estimated with physical inac-tivity as the outcome variable. The estimated association betweenphysical inactivity and network diversity was represented using b.The product of the two estimated coefficients, ab, provides anestimate of the mediated effect of social participation. The ab pointestimate of the mediated effect can be considered the product oftwo random variables, and first-order Taylor series expansion canbe used to provide estimates of the standard error of the mediatedeffect (Krull &Mackinnon,1999). The ratio of the ab estimate and itsstandard error were used to calculate z-scores, Wald statistics, and95% confidence intervals to test the null hypothesis that the abestimate of the mediation effect was zero. All estimates wereadjusted for the participants’ socio-demographic and -economiccharacteristics. The mediated effect estimate ab, its 95% confidenceintervals, and significance level are reported in text.

Results

Table 1 provides a summary of the socio-economic, -demo-graphic, and social capital characteristics of the MoNNET-HAphysical inactivity sample. Due to missing variable informationfor some participants, analyses were conducted with a final samplesize of 2672 Montreal adults. Table 2 reports the adjusted oddsratios for the three fitted models. AmongMoNNET-HA participants,17.10% reported being physically inactive. Approximately 11% ofrespondents reported participating in a neighbourhood as well asan outside voluntary organization or group over the last five years,while 26% reported participating in either a neighbourhood oroutside group in that period. For generalized trust, 42.5% ofparticipants reported high trust.

In terms of the association of socio-demographic and -economiccharacteristics with physical inactivity, analyses showed thatyounger age groups compared to those older than 75 years old andmen (OR: 0.73; 95% CIs: 0.58e0.93) compared to women were lesslikely to be physically inactive. In terms of educational attainment,persons with no high school degree were more likely to physicallyinactive compared to university graduates. With higher levels ofself-reported health, participants were less likely to be physicallyinactive. With regard to the social capital variables, the studyshowed that compared to those with high participation, thosewith no participation were more likely to be physically inactive(OR: 1.64; 95% CIs: 1.06e2.54). Individuals with higher network

diversity were shown less likely to be physically inactive (OR: 0.87;95% CIs: 0.80e0.95). Generalized trust was not found associatedwith physical inactivity. The mediation analysis showed thatnetwork diversity partially mediated the association betweenphysical inactivity and no social participation (ab¼ 0.10; 95% CIs:0.04e0.17; p< 0.01).

Discussion

To assess the importance of individual network social capital forphysical inactivity, this study examined the association amongphysical inactivity, three dimensions of network social capital,social participation and trust. In contrast to previous research, ourstudy used formal network measures of social capital to assesswhether specific dimensions of network social capital were asso-ciated with physical inactivity. The network diversity dimension ofsocial capital was found to be associated with physical inactivity. Inaddition, our study also showed that network social capital medi-ated the association between social participation and physicalinactivity.

Results of this study add to the growing literature on the rela-tionship of social capital and physical inactivity in several ways.

Table 2Adjusted odds ratios and 95% confidence intervals from multilevel logistic regres-sion analyses, n¼ 2672.

Variable Model 1 Model 2 Model 3

Age group25e34 0.37 (0.24, 0.59) 0.36 (0.23, 0.56) 0.37 (0.24, 0.59)35e44 0.51 (0.34, 0.77) 0.50 (0.33, 0.76) 0.52 (0.34, 0.79)45e54 0.53 (0.36, 0.79) 0.53 (0.36, 0.78) 0.55 (0.37, 0.81)55e64 0.55 (0.37, 0.81) 0.54 (0.36, 0.80) 0.58 (0.39, 0.86)65e74 0.78 (0.56, 1.10) 0.80 (0.57, 1.12) 0.81 (0.58, 1.15)75þ 1.00 1.00 1.00GenderMale 0.73 (0.58, 0.92) 0.73 (0.58, 0.92) 0.73 (0.58, 0.93)Female 1.00 1.00 1.00Educational attainmentNo degree 1.82 (1.28, 2.61) 1.65 (1.15, 2.36) 1.47 (1.01, 2.13)High school/trade 1.15 (0.86, 1.55) 1.06 (0.78, 1.43) 0.98 (0.73, 1.34)College 1.41 (1.03, 1.93) 1.34 (0.98, 1.83) 1.29 (0.94, 1.77)University 1.0 1.00 1.00Household income<28,000 1.44 (0.89, 2.33) 1.36 (0.84, 2.20) 1.21 (0.75, 1.97)28,000e49,000 1.17 (0.74, 1.84) 1.12 (0.71, 1.77) 1.07 (0.68, 1.68)50,000e74,000 1.08 (0.68, 1.70) 1.05 (0.66, 1.64) 1.03 (0.65, 1.62)75,000e100,000 0.89 (0.53, 1.51) 0.88 (0.52, 1.49) 0.85 (0.51, 1.45)>100,000 1.00 1.00 1.00Self-reported healthExcellent 0.19 (0.11, 0.33) 0.20 (0.11, 0.34) 0.19 (0.11, 0.34)Very Good 0.30 (0.18, 0.50) 0.31 (0.19, 0.51) 0.31 (0.19, 0.52)Good 0.50 (0.31, 0.82) 0.50 (0.31, 0.82) 0.49 (0.30, 0.80)Fair 0.56 (0.33, 0.95) 0.56 (0.33, 0.95) 0.55 (0.32, 0.93)Poor 1.0 1.0 1.00Generalized trustLow trust 1.10 (0.87, 1.39) 1.07 (0.85, 1.36)High trust 1.00 1.00ParticipationNo 1.89 (1.23, 2.89) 1.64 (1.06, 2.54)Low 1.49 (0.94, 2.35) 1.42 (0.90, 2.26)High 1.00 1.00Network capital dimensionsDiversity 0.87 (0.80, 0.95)Reach 0.99 (0.99, 1.00)Range 1.01 (1.00, 1.02)

H. Legh-Jones, S. Moore / Social Science & Medicine 74 (2012) 1362e13671366

First, the current study did not show an association betweenphysical inactivity and individual level generalized trust. Thisdiscrepancy may be due to the use of generalized trust compared tomore particular measures of trust such as neighbourhood trust orinstitutional trust. Differences between the findings of this studyand those of other studies may also reflect differences in socialattitudes about trust in the community or culture in which thestudy was performed (Abbott & Freeth, 2008). Second, similar tostudies that have been conducted in Sweden, the United States,Australia, and Japan, this study showed an association betweenphysical inactivity and social participation. The mediation analysesshowed that social network pathways partially explain the associ-ation between social participation and physical inactivity. Yet,social participation remained associated with physical inactivity,suggesting that social participation may act along a number ofpathways to reduce physical inactivity. Social participation may bedirectly associated with physical inactivity in that participationmay involve individuals joining running clubs or other types ofsports associations. Indirectly, social participation may increaseone’s access to information about physical activity opportunities orthe importance of physical activity for health. Social participationmay also decrease the likelihood of physical inactivity by creatingopportunities to meet others and develop one’s network capital.

Third, findings from this study suggest that individuals withgreater network diversity are less likely to be physically inactivecompared to those with lesser diversity. Social networks maythemselves be associatedwith physical inactivity through a number

of potential mechanisms. Greater network diversity can expanda person’s range of informational sources and level of social support,and can provide greater access to materials and resources thatreduce the chances of physical inactivity. Physical inactivity may bediscouraged through the provisioning of emotional, instrumental,appraisal, or informational support that encourages people to beactive (Berkman, Glass, Brissette, & Seeman, 2000). Individuals whoreceive positive support for physical activity are more likely toengage in regular activity than those individuals who lack support(Cerin & Leslie, 2008; Giles-Corti & Donovan, 2002). Other studieshave suggested that social networks may also exert social influencethrough shared norms and attitudes (Smith & Christakis, 2008). Inthis regard, networkdiversitymayserve to breakupdensenetworksin which there are strong social norms encouraging sedentarybehaviours or physical inactivity. However, additional data howeverare needed to evaluate this potential mechanism.

The use of network capital measures in the study of socialcapital and health has been limited. Previous studies of socialcapital and physical (in)activity have tended to rely on measuresof social participation and trust. Formal network capitalmeasures capture dimensions of network diversity as well as ofthe hierarchical aspects of resource accessibility that comes withsocial capital. Formal network social capital measures haveshown significant associations with obesity and self-reportedhealth in previous studies and different samples (Carpiano &Hystad, 2011; Moore et al., 2010, 2011). To this growing litera-ture, we have shown the importance of network diversity forpotentially helping reduce physical inactivity among urban adultpopulations.

Finally, while this study found no downsides to having highsocial capital in terms of network diversity when it comes tophysical inactivity, it is important to note that there is a generaltendency for individuals to associate with those of similar charac-teristics to one’s own, i.e., the homophily principle. It is likely thatindividuals who associate with others who are physically inactiveare more likely to be physically inactive themselves. As such,further exploration into the physical activity behaviours of one’snetwork members or the types of social activities in which indi-viduals participate is worthwhile.

Limitations

There are a number of limitations that need to be considered forthis study. First, although cross-sectional survey data arecommonly used in research on social capital and physical activity,causal inferences should be avoided. Physically inactive individualsmay face barriers to social participation and developing their socialnetworks that would reduce a person’s social capital. As such, it isalso important to note that the mediation analysis is cross-sectional. Mediation analyses using cross-sectional data are notunusual but findings must be interpreted with recognition of thecross-sectional nature of the data. Longitudinal data on socialcapital and physical inactivity would help discern the directionalityof the social capital and physical inactivity relationship, as well asguide further research on the mechanisms behind social capital.Second, the MoNNET-HA physical inactivity information is self-reported. While self-reported physical activity is common inlarger population studies, there are a number of potential limita-tions. Self-reported physical activity can introduce social desir-ability or recall bias, which can potentially lead to over-estimatingphysical activity. Studies have suggested that adults tend to over-estimate their physical activity levels (Sallis & Saelens, 2000).Nevertheless, despite the recognized limitations of self-reportedphysical activity, the IPAQ has been tested for validity and reli-ability across developed and developing countries (Craig et al.,

H. Legh-Jones, S. Moore / Social Science & Medicine 74 (2012) 1362e1367 1367

2003). Finally, although social participation was shown associatedwith physical inactivity, our measures of social participation arelimited. More extensive information on the type or intensity ofsocial participation would have enhanced our ability to examinethe potential pathways by which social participation is associatedwith physical inactivity.

Conclusion

This study demonstrated a significant association amongnetwork capital diversity, social participation andphysical inactivityamong Montreal adults. The significance of network diversitysuggests that there are important components of social capital thatare not captured using conventional measures of trust and partici-pation. This study recommends further development of networkapproaches to the study of physical inactivity and social capital.Greater understanding of how social networks are associated withphysical inactivity can be used to inform physical activity programs.Such programs might target associational involvement and peernetworks as ameans to improve individual levels of physical activity.

Acknowledgements

S.M. holds a New Investigator Award from the Canadian Insti-tutes of Health Research (CIHR) e Institute of Aging. Data collectionand analysis was supported by a team grant from the CanadianInstitutes of Health Research (Grant no. MOP-84584).

References

Abbott, S., & Freeth, D. (2008). Social capital and health-starting to make sense ofthe role of generalized trust and reciprocity. Journal of Health Psychology, 13(7),874e883.

Berkman, L. F., Glass, T., Brissette, I., & Seeman, T. E. (2000). From social integrationto health: Durkheim in the new millenium. Social Science & Medicine, 51(6),843e857.

Canadian Society for Exercise Physiology. (2011). Canadian physical activity guide-lines. http://www.csep.ca/CMFiles/Guidelines/CPAGuideline_Report_JAN2011.pdf Last accessed December 1, 2011.

Carpiano, R., & Hystad, P. (2011). “Sense of community belonging” in health surveys:what social capital is it measuring? Health & Place, 17(2), 606e617.

Cerin, E., & Leslie, E. (2008). How socio-economic status contributes to participationin leisure-time physical activity. Social Science & Medicine, 66(12), 2596e2609.

Craig, C. L., Marshall, A. L., Sjostrom, M., Bauman, A. E., Booth, M. L., Ainsworth, B. E.,et al. (2003). International physical activity questionnaire: 12-country reliabilityand validity. Medicine & Science in Sports & Exercise, 35(8), 1381e1395.

Dietz, W. H. (1996). The role of lifestyle in health: the epidemiology and conse-quences of inactivity. Proceedings of the Nutrition Society, 55, 829e840.

Giles-Corti, B., & Donovan, R. J. (2002). The relative influence of individual, socialand physical environment determinants of physical activity. Social Science &Medicine, 54(12), 1793e1812.

Greiner, K. A., Li, C., Kawachi, I., & Hunt, D. C. (2004). The relationship of socialparticipation and community rating to health and health behaviors in areaswith high and low popultation density. Social Science & Medicine, 59(11),2303e2312.

Guthold, R., Ono, T., Strong, K. L., Chatterji, S., & Somnath, A. (2008). Worldwidevariability in physical inactivity: a 51-country survey. American Journal ofPreventive Medicine, 34(6), 486e494.

Haines, V., Beggs, J., & Hurlbert, J. (2011). Neighborhood disadvantage, networksocial capital, and depressive symptoms. Journal of Health and Social Behavior,52(1), 58e73.

International Physical Activity Questionnaire (IPAQ) Research Committee. (2005).Guidelines for data processing and analysis of the international physical activityquestionnaire (IPAQ) e short and long form. Retrieved from. https://sites.google.com/site/theipaq/scoring-protocol.

Kawachi, I., Kennedy, B. P., & Glass, R. (1999). Social capital and self-rated health:a contextual analysis. American Journal of Public Health, 89(8), 1187e1193.

Kawachi, I., Subramanian, S., & Kim, D. (2008). Social capital and health: a decade ofprogress and beyond. In I. Kawachi, S. Subramanian, & D. Kim (Eds.), Socialcapital and health (pp. 1e26). New York: Springer Science þ Business Media,LCC.

Krull, J., & Mackinnon, D. P. (1999). Multilevel mediation modeling in group-basedintervention studies. Evaluation Review, 23(4), 418e444.

Lin, N., Fu, Y.-c., & Hsung, R.-M. (2001). The Position generator: measurementtechniques for investigations of social capital. In N. Lin, K. Cook, & R. S. Burt(Eds.), Social capital theory and research (pp. 57e81). New York: Aldine DeGruyter.

Lindstrom, M., Hanson, B. S., & Ostergren, P.-O. (2001). Socioeconomic differencesin leisure-time physical activity: the role of social participation and socialcapital in shaping health related behaviour. Social Science & Medicine, 52(3),441e451.

Lindstrom, M., Moghaddassi, M., & Merlo, J. (2003). Social capital and leisure timephysical activity: a population based multilevel analysis in Malmo, Sweden.Journal of Epidemiology and Community Health, 57(1), 23e28.

Moore, S., Daniel, M., Paquet, C., Dube, L., & Gauvin, L. (2009). Association of indi-vidual network social capital with abdominal adiposity, overweight andobesity. Journal of Public Health, 31(1), 175e183.

Moore, S., Gauvin, L., Daniel, M., Kestens, Y., Bockenholt, U., Dubé, L., et al. (2010).Associations among park visitation, age, social participation, and age compo-sition in Montreal. Leisure Sciences, 32(4), 318e336.

Moore, S., Bockenholt, U., Daniel, M., Frochlich, K., Kestens, Y., & Richard, L. (2011).Social capital and core network ties: a validation study of individual-level socialcapital measures and their association with extra- and intra-neighbourhoodties, and self-rated health. Health & Place, 17(2), 536e544.

Poortinga, W. (2006). Perceptions of the environment, physcial activity, and obesity.Social Science & Medicine, 63(11), 2835e2846.

Sallis, J. F., & Saelens, B. E. (2000). Assessment of physical activity by self-report:status, limitations, and future directions. Research Quarterly for Exercise andSport, 71(2), 1e14.

Smith, K. P., & Christakis, N. A. (2008). Social networks and health. Annual Review ofSociology, 35, 405e429.

Swaroop, S., & Morenoff, J. D. (2006). Building community: the neighbourhoodcontext of social organization. Social Forces, 84(3), 1665e1695.

Szreter, S., & Woolcock, M. (2004). Health by association? Social capital, socialtheory, and the political economy of public health. International Journal ofEpidemiology, 33(4), 650e667.

Ueshima, K., Fujiwara, T., Takao, S., Suzuki, E., Iwasw, T., Doi, H., et al. (2010). Doessocial capital promote physical activity? A population-based study in Japan.PLoS ONE, 5(8), e12135.

Warburton, D. E., Nicol, C. W., & Bredin, S. S. (2006). Health benefits of physicalactivity: the evidence. Canadian Medical Association Journal, 174(6), 801e809.