Does Your Body Know Who You Know? Multiple Roles of Network Members’ Socioeconomic Status for Body Weight Ratings

Abstract

Combining the theory of social capital with work on three social factors, respectively, socioeconomic status (SES), gender, and lifestyle, this study examines four roles of accessed SES (network members’ SES) for body weight ratings: direct association, indirect association through lifestyle, mediating role in the relationship between SES and body weight ratings, and interaction with gender. Analyzing data from the 2004 U.S. General Social Survey, this study measures body weight ratings (visually evaluated by interviewers) and two indicators of accessed SES on the educational dimension (network members’ average education and proportion of network members with some college education or more). The results show evidence not for the direct role of accessed education but rather for its three other roles. More educated adults of both genders have access to more educated network members; those with more educated network members have a stronger athletic identity (a proximate indicator of lifestyle); and those with a more solid athletic identity have lower body weight ratings. Also, men with more educated network members have higher body weight ratings, but the opposite pattern applies to women. This study refines social capital theory and advances our understanding of network, socioeconomic, lifestyle, and gender disparities in body weight.

Keywords

network members’ status social capital lifestyle gender socioeconomic status body weight

Social (as well as genetic) factors have been well demonstrated to influence body weight (Bombak 2014; Bordo 2003; Bourdieu [1979] 1984; Gard and Wright 2005; McLaren 2007; Ross and Mirowsky 1983). Recently, the social network perspective on social determinants of body weight has stimulated much research (e.g., Christakis and Fowler 2007; Cohen-Cole and Fletcher 2008; de la Haye et al. 2011; Fletcher, Bonell, and Sorhaindo 2011; Hruschka et al. 2011; Kiernan et al. 2012; Leahey et al. 2011; Trogdon, Nonnemaker, and Pais 2008; Valente et al. 2009). Accessed socioeconomic status (SES)—network members’ (or alters’) SES—has achieved decades-long prominence in the social network literature (Fischer et al. 1977; Homans 1950; Laumann 1966; Lazarsfeld and Merton 1954). However, its relationship with body weight has been given little attention (Christakis and Fowler 2007; Moore, Daniel, Paquet, et al. 2009).

The purpose of this study is to extend the theory of social capital (Lin 1982, 2001) to the domain of body weight by combining research on accessed SES with work on SES, gender, and lifestyle (Bordo 2003; Cockerham 2005; McLaren 2007; Ross 1994). We analyze four roles of accessed SES for body weight ratings: direct association, indirect association through lifestyle, mediating role in the relationship between SES and body weight ratings, and interaction with gender. Data are drawn from the 2004 General Social Survey (GSS) in the United States, which to our knowledge includes the only nationally representative data of adults with information on body weight and accessed SES. The data allow the measurement of body weight ratings (visually evaluated by GSS interviewers) and accessed SES on the educational dimension. The findings extend social capital theory and advance our understanding of the complex social dynamics of body weight at both the individual and network levels.

Literature Review

Social Networks, Accessed SES, and Body Weight

Traced back to Emile Durkheim’s ([1897] 1951) classic study on suicide, the social network perspective has inspired a long research tradition on health consequences of various relationship-based factors (for reviews, see Berkman et al. 2000; House, Landis, and Umberson 1988; Pescosolido 2006; K. P. Smith and Christakis 2008; Song, Son, and Lin 2011; Thoits 2011; Umberson and Montez 2010). Recently, this perspective has grown into a burgeoning approach to the social distribution of body weight.

Existing social network research on body weight centers on four network factors: alters’ body weight, lifestyle, network norm, and social support. Using longitudinal community data collected in Framingham, Massachusetts, Nicholas A. Christakis and James H. Fowler (2007) first demonstrate that one (or ego) is more likely to become obese if connected to obese friends. The positive association between ego’s and alters’ weight and the positive association between ego’s and alters’ weight-related lifestyle are further confirmed in most later studies (e.g., Cohen-Cole and Fletcher 2008; de la Haye et al. 2011; Fletcher et al. 2011; Fowler and Christakis 2008; Hruschka et al. 2011; Leahey et al. 2011; Trogdon et al. 2008; Valente et al. 2009; VanderWeele 2011). Theoretically, from the social causation perspective, one major explanation for these positive associations is social network influence. In addition, network norm measured as alters’ subjective attitude toward body shape and weight control is positively related to ego’s weight loss intention and behavior (Leahey et al. 2011; Paxton et al. 1999). Furthermore, social support from alters is shown to facilitate ego’s weight loss (Kiernan et al. 2012; Marcoux, Trenkner, and Rosenstock 1990).

Accessed SES is an upstream structural attribute of personal networks. This attribute captures alters’ positions in the socioeconomic hierarchy, and constitutes the meso-level hierarchical context in which ego dwells and socializes in daily life. Accessed SES can directly influence the aforementioned four network factors (i.e., alters’ body weight, lifestyle, network norm, and social support). It has attracted voluminous research for its role in social interaction, social exchange, social comparison, status attainment, and health in the past six decades (Fischer et al. 1977; Homans 1950; Laumann 1966; Lazarsfeld and Merton 1954; for reviews see Lin 1999; Song 2013). Among diverse theoretical approaches to social capital (Bourdieu [1983] 1986; Carpiano 2006; Coleman 1990; Lin 1982, 2001; Portes 1998; Putnam 2000), one network-based approach directly theorizes the function of accessed SES and can help us understand the significance of accessed SES for body weight (Lin 1982, 2001). Other approaches to social capital are beyond the scope of this present study.

Nan Lin’s network-based approach conceptualizes social capital as resources embedded in social networks, operationalizes it as assets alters possess, and specifies it as alters’ hierarchical social positions, especially alters’ SES (Lin 1982, 2001). Two network instruments can be used to measure accessed SES: the position and name generators. The position generator asks respondents to identify contacts associated with a sample of occupational positions (Lin, Fu, and Hsung 2001). The name generator asks respondents to name contacts they discuss important matters with (Burt 1984). The network-based approach to social capital argues that accessed SES indicates nonredundant, valuable social resources beyond and above ego’s own resources, and can be accessed and mobilized to facilitate ego’s purposive actions through multiple mechanisms, including exerting influences, offering information, serving as social credentials, and reinforcing group identity and recognition. Substantial empirical research verifies the positive impact of accessed SES on status attainment across societies over the past three decades (for reviews, see Lin 1999; Portes 1998). Recently, this network-based social capital theory has been extended to health and well-being outcomes (for a review, see Song 2013). Theoretically, accessed SES protects health through seven pathways: encouraging healthy norms, providing social support, cultivating healthy lifestyle, facilitating access to health care, decreasing stress exposure, advancing social status, and reinforcing psychological resources (Erickson 2003; Song 2011; Song and Lin 2009). The positive associations between accessed SES and various health-related outcomes (e.g., physical and mental health, life satisfaction, health information search, and smoking cessation) are demonstrated across societies (Acock and Hurlbert 1993; Christakis and Fowler 2008; Haines, Beggs, and Hurlbert 2011; Moore et al. 2011; Song 2011; Song and Chang 2012; Song and Lin 2009; Verhaeghe and Tampubolon 2012).

Three studies have examined the impact of accessed SES on weight-related outcomes (Christakis and Fowler 2007; Legh-Jones and Moore 2012; Moore, Daniel, Paquet, et al. 2009). Two of them analyze data from a community sample of young and middle-aged adults (18 to 55 years old) in Montreal, Canada (Legh-Jones and Moore 2012; Moore, Daniel, Paquet, et al. 2009). They measure accessed SES on the occupational dimension using the position generator. They find that accessed occupational status is negatively associated with body weight (measured by waist circumference and body mass index) and physical inactivity. Another study uses data from a longitudinal community sample of adults aged 21 or older in Framingham, Massachusetts (Christakis and Fowler 2007). It measures accessed education using the name generator, and shows no significant association between accessed education and body weight (measured by body mass index). Note that some prior studies focus on parents or households and contribute to documenting a negative association between parental or household SES and body weight (Ball and Mishra 2006; Baum and Ruhm 2009; Langenberg et al. 2003; O’Dea and Caputi 2001; Schmeer 2010). But these studies are limited to only family relationships. They do not completely capture one’s personal networks and thus cannot fully measure accessed SES.

In sum, despite the growing attention to the association between social networks and body weight, the role of accessed SES remains underexplored. The two existing studies report mixed evidence. Spencer Moore et al. (Moore, Daniel, Paquet, et al. 2009) find a direct negative association between accessed occupation and body weight, but Christakis and Fowler (2007) report no evidence for the direct association between accessed education and body weight. These inconsistent results may be due to four major differences, respectively, in research designs, samples, societies, and measurements of accessed SES. Furthermore, the results from these three studies lack national representativeness and generalization due to their data limitations. Theoretically more important, how accessed SES works together with well-documented, individual-level social determinants of body weight (i.e., SES, gender, and lifestyle) remains unclear.

SES and Body Weight

SES is, in general, positively associated with perceived body weight but negatively associated with actual body weight in developed countries (Chang and Christakis 2003; Sobal and Stunkard 1989). In a recent literature review on the association between SES and obesity/overweight (McLaren 2007), more than 65 percent of all report associations involving education are negative, 2 percent positive, and 33 percent nonsignificant or curvilinear (N = 48). The corresponding numbers for occupation and income are, respectively, 59, 4, and 37 percent (N = 27), and 31, 8, and 61 percent (N = 62). The fact that education plays a more consistent negative role than occupation and income may be due to its more direct relevance to body weight. Education can be more directly positively related to the internalization of social expectations on body weight, health literacy, weight concern, sense of control, and weight-related lifestyle earlier on in people’s lives (McLaren 2007; Ross and Wu 1995; Trost et al. 2002). Occupation may affect body weight through its physical requirement, working environment, and social status. The role of income may come in last in the form of financial and time investment in weight management.

Gender and Body Weight

Gender affects body weight mainly due to gendered social norms on body weight. The slenderness norm is more strictly imposed on women than on men (Bordo 2003; Hesse-Biber 2007). Women are more likely to evaluate themselves as overweight, suffer from body dissatisfaction, perceive weight stigma and discrimination, and engage in weight management practices (Chang and Christakis 2003; Schafer and Ferraro 2011; Warin et al. 2008). Women but not men pay socioeconomic, marital status, and social penalty due to excess body weight (Conley and Glauber 2007; Crosnoe, Frank, and Mueller 2008). In contrast, men face two gender norms: breadwinning and masculinity (Courtenay 2000; McLaren 2007). The norm of gendered division of labor labels men as breadwinners who should aim at socioeconomic resources. Men are also more expected to choose physical labor careers. At the same time, a larger body size remains an important symbol of masculine identity and social dominance for men.

Note that SES interacts with gender (Ross 1994). Thinness is a more distinctive symbol of socioeconomic positions for women than men (Bourdieu [1979] 1984). As a recent literature review on the association between SES and obesity/overweight reports (McLaren 2007), among developed societies, more than 95 percent of all reported significant associations for women are negative and less than 5 percent positive (N = 480), while the corresponding numbers for men are, respectively, 80 and 20 percent (N = 262).

Lifestyle and Body Weight

Weight-related lifestyle reflects collective patterns of weight-related behaviors, including dietary habits, and light, moderate, or vigorous physical activities and exercise such as walking, running, basketball, aerobics, tennis, swimming, biking, and lifting heavy objects (Grzywacz and Marks 2001; Ross 1994; Wardle and Griffith 2001). Weight-related behaviors benefit weight management and control and are negatively associated with body weight (Bauman 2004; Ross 1994). As described later, although collecting no data on weight-related lifestyle, the 2004 GSS has information on athletic identity. Athletic identity refers to “one’s perception of himself/herself as a person who participates in exercise, sports, and physical activity or as one who does not, and the active efforts made to verify this self-view” (Anderson 2004:39). It predicts and can serve as a proximate indicator of weight-related lifestyle. Self-rated measures of athletic identity have proven to be positively associated with exercise status, exercise frequency, physical and athletic activities, and sports participation (Anderson 2004; Anderson et al. 2009; Lamont-Mills and Christensen 2006; Schutte and McNeil 2015). In the study of Cheryl B. Anderson (2004), for example, the significant positive coefficients for the correlations between four latent factors of athletic identity (i.e., appearance, importance, competence, and encouragement) and a latent factor of physical activity with two indicators (i.e., stage of exercise behavior and exercise frequency per week) are strong and range from .56 to .90.

Hypotheses: Four Roles of Accessed SES for Body Weight Ratings

Based on available data and prior work on accessed SES, SES, weight-related lifestyle, and gender, this study theorizes four roles of accessed SES in the social distribution of body weight ratings (see Figure 1): (1) direct association (Figure 1a), (2) indirect association through lifestyle (Figure 1b), (3) mediating role in the relationship between SES and body weight ratings (Figure 1c), and (4) interaction with gender (Figure 1d).

Figure 1.

The conceptual model of accessed SES, SES, lifestyle, gender, and body weight ratings.

Direct and Indirect Associations

The network-based theory of social capital argues that accessed SES represents valuable social resources (Lin 1982, 2001). As such, accessed SES can be associated with body weight for three reasons: network norm, social support, and lifestyle. First, alters’ SES can shape network norm (Song and Chang 2012). Network norm can regulate people’s attitude toward and belief about body weight. Higher-SES people are more weight conscious and more concerned about body image and weight management (Bourdieu [1979] 1984; Chang and Christakis 2003; McLaren 2007; O’Dea and Caputi 2001; Schafer and Ferraro 2011; Wardle and Griffith 2001; S. J. Williams 1995). If surrounded by higher-SES alters, egos are exposed to a stronger network norm of weight control and more conscious of body weight (Brewis, Hruschka, and Wutich 2011; Hruschka et al. 2011; Leahey et al. 2011; Paxton et al. 1999). Second, alters’ SES can determine the quantity and quality of social support available from alters (Lin and Ao 2008; Song et al. 2011). Weight management requires the investment of “economic and cultural means” (Bourdieu [1979] 1984). Higher-SES contacts may be more able to offer egos diverse forms of assistance (e.g., informational, instrumental, emotional, appraisal, and financial aid) with weight management or loss (Kiernan et al. 2012; Marcoux et al. 1990; Trost et al. 2002). Third, alters’ SES can help cultivate lifestyle. Lifestyle is an objective product of socioeconomic standing (Bourdieu [1979] 1984; Cockerham 2005; Weber 1978). Higher-SES people are more likely or able to practice and manifest a weight-control lifestyle (Grzywacz and Marks 2001; Ross 1994; Trost et al. 2002; Wardle and Griffith 2001). If connected to higher-SES alters, egos are more likely to observe and imitate those weight-control behaviors (Paxton et al. 1999). Therefore, the direct-association or social resources hypothesis states the following:

Hypothesis 1 (H1): Accessed SES is negatively associated with body weight ratings (see Figure 1a).

Athletic identity strongly predicts athletic and physical activities, and can serve as a proximate indicator of weight-related lifestyle (Anderson 2004). Egos with higher-status alters can be more likely to be athletically and physically active and develop athletic identity, and, thus, be less likely to be overweight (Legh-Jones and Moore 2012; Ross 1994; Yang, Telama, and Laakso 1996). The indirect-association or lifestyle hypothesis states the following:

Hypothesis 2 (H2): Accessed SES is negatively associated with ego’s body weight ratings indirectly through its positive association with ego’s athletic identity (see Figure 1b).

Mediating Roles

Social capital indicated by accessed SES is an endogenous social factor. It can act as a mediating pathway, linking its social antecedents with instrumental and expressive outcomes (Lin 2001; Song 2011). This present study focuses on one of the most documented social precursors of accessed SES as well as body weight: ego’s SES or personal capital (i.e., resources ego possesses and controls; Lin 1999; McLaren 2007; Sobal and Stunkard 1989). Ego’s SES is convertible to accessed SES through three possible mechanisms (Bourdieu [1983] 1986; Lin 2001; Song 2011). First, in comparison with lower-SES egos, higher-SES egos may have more structural opportunities to encounter contacts who occupy higher socioeconomic positions in society. Second, higher-SES egos possess more resources, and may be more able to invest in social networking and reach higher-SES alters. Third, higher-SES egos may be more attractive to higher-SES alters due to the homophily principle (i.e., people tend to associate with like others; McPherson, Smith-Lovin, and Cook 2001). Therefore, based on the direct- and indirect-association hypotheses (H1, H2), two mediating-role or capital convertibility hypotheses state the following:

Hypothesis 3 (H3): Accessed SES by itself (H3a) or together with athletic identity (H3b) negatively mediate the relationship between ego’s SES and body weight ratings (see Figure 1c).

Interaction With Gender

The impact of accessed SES on body weight can be moderated by gender norms on body weight and division of labor. We propose two alternative hypotheses on this moderating process by drawing on prior work on the interaction effect between SES and gender on body weight. As reviewed before, there are two gender-specific patterns: (1) more consistent evidence for a negative SES-obesity association among women than men, and (2) relatively stronger evidence for a positive SES-obesity association among men than women (McLaren 2007).

The first pattern manifests in the more strict imposition of the slenderness norm on women, in particular higher-SES women (Bordo 2003; Hesse-Biber 2007). Higher-SES women are more dissatisfied with weight and more likely and able to practice a weight-control lifestyle than lower-SES counterparts (McLaren and Kuh 2004; Warin et al. 2008). When connected with higher-SES alters, in contrast with men, women have to conform to a more coercive network norm of slenderness, seek and receive more social support, practice more weight-control behaviors, and mobilize more of their alters’ resources for weight management. Thus, the first interaction-with-gender (or ideal slender women) hypothesis states the following:

Hypothesis 4a: Accessed SES is negatively associated with body weight ratings to a greater degree for women than for men (see Figure 1d).

The second pattern suggests the coexistence of two gender norms among men: breadwinning and masculinity (Courtenay 2000; McLaren 2007). If surrounded by higher-SES alters, men may be exposed to a stronger network norm of breadwinner masculinity and be less concerned about weight management and risks of excess weight. Based on the dominant negative SES-obesity association among women (McLaren 2007), the second interaction-with-gender (or men as masculine breadwinners) hypothesis states the following:

Hypothesis 4b: Accessed SES is associated with body weight ratings positively for men but negatively for women (see Figure 1d).

Data and Method

Data

Data were drawn from the 2004 U.S. GSS. The GSS is a repeated cross-sectional survey of a nationally representative sample of noninstitutionalized U.S. adults, which has been conducted by the National Opinion Research Center since 1972. Using a full-probability sample design, the GSS interviewed respondents face to face through computer-assisted personal interviewing. The GSS regularly includes topical modules apart from core items, but does not administer all modules to all respondents. The 2004 GSS includes two modules among others: (1) social networks, and (2) genes and the environment. The first module allows the measurement of accessed SES, and the second one has information on athletic identity and body weight ratings.

The 2004 GSS has a response rate of 70.4 percent with a sample size of 2,812. The subsample for this study included only those respondents who were administered both the social networks module and the genes and the environment module (N = 1,273). The subsample is nationally representative, although its size is smaller than the full sample size (Song and Chang 2012). The comparison of all used variables’ characteristics in the full GSS sample and the subsample found no significant difference. As in previous studies on ego-centered networks (Marsden 1987; Song and Chang 2012), we had to exclude social isolates naming no contact (N = 276), and the subsample size dropped to 997 after that exclusion. Supplemental analyses showed no significant correlations of social isolation (1 = naming no contacts, 0 = naming one or more contacts) with athletic identity and body weight. The listwise deletion of cases with missing values on the variables of interest can further incur a loss of 22 percent of the subsample. We employed the multiple multivariate imputation method in Stata to correct missing-data bias and imputed missing values in the variables of interest based on 10 imputations (Royston 2005). Multiple imputation assumes that data are missing at random (MAR). Our sensitivity analysis through applying the pattern-mixture model approach suggested that departures from the MAR assumption did not affect our conclusions (Allison 2002). Each of these 10 imputed data sets included 974 respondents. Occupational prestige was not imputed for respondents who never worked as long as one year (N = 23). Table 1 shows the summary of unweighted raw sample characteristics. Also, the 2004 GSS adopted a nonrespondent, subsampling design, and we weighed its data using one sampling weight variable, WTSS. The sampling weight variable, WTSS, “takes into consideration a) the sub-sampling of non-respondents, and b) the number of adults in the household” (Smith et al. 2011:3103).

Table 1.

Dependent and Independent Variable Distributions by Gender (Unweighted Raw Sample).

	M (SD)/Percent (N)
	Full sample	Men	Women
Variables	(N = 974)	(N = 424)	(N = 550)
Dependent variable
Body weight ratings***
Below average	7% (67)	4% (15)	10% (52)
Average	68% (629)	74% (295)	64% (334)
Somewhat above average	19% (177)	17% (69)	21% (108)
Considerably above average	5% (49)	5% (20)	6% (29)
Independent variables
Athletic identity (being an athletic person is)***
Not a good description at all	12% (117)	6% (27)	17% (90)
Not a very good description	29% (274)	23% (94)	33% (180)
A fair description	31% (300)	32% (135)	30% (165)
A good description	17% (171)	24% (101)	12% (70)
A very good description	11% (110)	16% (67)	8% (43)
Accessed socioeconomic status
Alters’ average education*	5.37 (1.50)	5.48 (1.52)	5.29 (1.48)
Proportion of alters with some college education or more^†	60%	63%	58%
Age (years)*	45.14 (16.06)	46.03 (16.00)	44.26 (16.38)
Gender (1 = male)	44% (424)
Race/ethnicity
White	82% (798)	83% (350)	82% (448)
Minority (Black and other race/ethnicity)	18% (176)	17% (74)	18% (102)
Marital status (1 = married)*	56% (542)	58% (251)	53% (291)
Employment status (1 = employed full time)***	54% (525)	62% (265)	46% (260)
Socioeconomic status
Education (years)^†	14.01 (2.84)	14.14 (3.04)	13.86 (2.70)
Occupational prestige	46.46 (14.40)	46.96 (14.30)	46.08 (14.48)
Annual family income (dollars)***	56,625 (44,466)	61,289 (43,805)	53,030 (44,678)
Network size (number of alters)*	2.71 (1.40)	2.63 (1.42)	2.77 (1.38)
Role relationship (proportion of alters who were family members)*	58%	55%	60%

Note. T test (two-tailed tests) and chi square by gender:

†

p ≤ .10. *p ≤ .05. **p ≤ .01. ***p ≤ .001.

Dependent Variable

Body weight ratings

The 2004 GSS interviewers visually rated respondents’ weight without asking (“If respondent is pregnant, consider her weight aside from being pregnant.”) on a broad 4-point scale (1 = below average; 2 = average; 3 = somewhat above average; 4 = considerably above average). We have some reasonable confidence with interviewers’ evaluation for two reasons: (1) interviewers are experienced and well-trained and assess respondents’ body weight using only four broad categories rather than a detailed scale, and (2) supplemental analyses find no associations between interviewers’ social attributes (i.e., age, gender, race/ethnicity, and years of service) and their evaluation. There is evidence that well-trained personnel can accurately or adequately estimate actual body weight in general (Arsalani-Zadeh et al. 2010; Lim et al. 2013).

Independent Variables

Accessed SES

The 2004 GSS collected information on ego-centered networks using the name generator (Burt 1984). The original wording for the first question is: “From time to time, most people discuss important matters with other people. Looking back over the last six months, who are the people with whom you discussed matters important to you? Just tell me their first names or initials.” The first five names (alters) were recorded. Respondents who named one or more alters were further asked the highest educational level of each alter (1 = 0–6 years; 2 = 7–9 years; 3 = 10–12 years; 4 = high school graduate; 5 = some college; 6 = associate degree; 7 = bachelor’s degree; 8 = graduate or professional degree). Similar to previous studies using the name generator (Acock and Hurlbert 1993; Haines et al. 2011; Song and Chang 2012), we calculated two indicators of accessed SES on the educational dimension: alters’ average education and proportion of alters with some college education or more. We conducted analyses separately using the two indicators for two reasons. First, the two indicators were highly correlated with each other (.858, p < .001). They were analyzed separately to avoid multicollinearity. Second, prior studies and their measurement-specific results suggest the possible distinction between different indicators of accessed SES (Song 2015b; Song and Chang 2012).

Athletic identity

The 2004 GSS asked one question on athletic identity: “please indicate how well the description (‘an athletic person’) applies to you.” This item was ordinal rated on a 5-point scale (1 = a very good description, 2 = a good description, 3 = a fair description, 4 = not a very good description, 5 = not a good description at all). We reversed the order of these five responses so that the higher the score, the more athletic respondents perceived themselves.

Gender was a dummy variable (0 = female, 1 = male). Three socioeconomic indicators were years of education, occupational prestige coded through the NORC/GSS Occupational Prestige scores (Nakao and Treas 1990), and annual family income in constant dollars. We applied a logarithmic transformation to normalize the distribution of annual family income.

All analyses controlled for three demographic factors, one indicator of employment status, and two network structure factors. Demographic factors included age in years, race/ethnicity (0 = minority, including black and other races/ethnicities; 1 = white), and marital status (0 = unmarried, 1 = married). Employment status was a binary variable (0 = not employed full time, 1 = employed full time). Two network structure factors—network size (the number of alters) and role relationship (proportion of alters who were family members)—were controlled for a robust test of the impact of accessed SES. Supplemental analysis found no significant interaction effect between role relationship and accessed SES in the prediction of body weight.

Analytic Strategy

This study first conducted bivariate analysis of the associations between accessed education and body weight ratings, athletic identity, and SES. Then, it estimated path analysis models using the Mplus program to simultaneously examine the first three hypotheses (Muthén and Muthén 1998–2015). It used the robust mean- and variance-adjusted weighted least squares estimator because of the presence of both continuous and ordinal dependent variables. We estimated a path analysis model with three equations, respectively, for three dependent variables: accessed education (Y₁), athletic identity (Y₂), and body weight ratings (Y₃). The first equation was an ordinary least squares regression of two continuous indicators of accessed education on control variables (X₁). The second equation was an ordinal logistic regression of athletic identity on accessed education and control variables. The third equation was an ordinal logistic regression of body weight ratings on athletic identity, accessed education, and control variables. Two approaches—the Sobel test and the bootstrapping method—were used to evaluate the indirect and mediating roles of accessed education (Bollen and Stine 1992; Sobel 1982). Note that no statistical methods are available for running path analysis of dyadic observations with clustering on the ego. To examine the interaction between accessed education and gender, we added the product term of gender with mean-centered accessed education (X₂) to equation 3 (see equation 4). Significant coefficients of product terms indicate the presence of interaction (Cohen and Cohen 1983). Supplemental dyadic analysis found similar results.

Y_{1} = f (X_{1}) .

Y_{2} = f (X_{1} {+ Y}_{1}) .

Y_{3} = f (X_{1} {+Y}_{1} {+ Y}_{2}) .

Y_{3} = f (X_{1} {+ Y}_{1} {+ Y}_{2} {+ X}_{2}) .

Results

Table 1 shows the summary of unweighted raw sample characteristics. Most of the sample (68 percent) were rated as average body weight, with women significantly more often rated as both below average and above average than men. The majority of the sample (59 percent) identified with being an athletic person, with men having significantly stronger athletic identity than women.

Accessed Education and Body Weight Ratings

Our bivariate analysis showed that at the p value of .001, alters’ average education was correlated negatively with body weight ratings (−.108) and positively with four other key variables: athletic identity (.135), years of education (.554), occupational prestige (.404), and annual family income (.314). At the p value of .001, the proportion of alters with some college education or more was correlated negatively with body weight ratings (−.114) and positively with athletic identity (.140), years of education (.476), occupational prestige (.315), and annual family income (.254). We also conducted separate bivariate analysis for men and women and found similar results with one exception. The negative correlations between the two indicators of accessed education and body weight ratings were significant only for women.

Next, we estimated two path analysis models to simultaneously examine the direct, indirect, and mediating roles of accessed education for body weight ratings (see equations 1 –3, and Table 2). The results support two of the four hypotheses (H2, H3b). Net of control variables including ego’s own education, alters’ average education was not directly associated with body weight ratings (see equation 3 and model 1). But it was positively associated with athletic identity (.077, p < .01), which in turn had a negative association with body weight ratings (−.179, p < .001; see equation 2). As results from both the Sobel test and the bootstrapping method indicated, alters’ average education was indirectly associated with body weight ratings through athletic identity in a negative direction (−.014, p < .05). Adults connected with on average more educated contacts perceived themselves more athletic, and furthermore, the more athletic adults had lower body weight ratings. Furthermore, all three indicators of SES—education (.213, p < .001), occupational prestige (.012, p < .01), and annual family income (.198, p < .001)—had positive associations with alters’ average education, but only education was associated with body weight (−.055, p < .01; see equation 1). As results from both the Sobel test and the bootstrapping method showed, alters’ average education played a significant negative mediating role (−.003, p < .05), together with athletic identity, in the relationship between education and body weight ratings. More educated adults had on average more educated contacts, and those tied to, on average, more educated contacts were more athletic and had lower body weight ratings. Supplemental analyses found no evidence for the indirect effects of other factors apart from that of education on body weight ratings through accessed SES and athletic identity.

Table 2.

Parameter Estimates of the Path Analysis Models of Accessed Socioeconomic Status, Athletic Identity, Body Weight Ratings, and Control Variables (N = 974).

	Path analysis model 1			Path analysis model 2
Independent Variables	Average education of alters	Athletic identity	Body weight ratings	Proportion of alters with some college education or more	Athletic identity	Body weight ratings
Age	.007**	−.010***	.005^†	.000	−.009***	.006*
Age	(0.003)	(0.002)	(0.003)	(0.001)	(0.002)	(0.003)
Gender (1 = male)	.065	.546***	.122	.017	.547***	.136^†
Gender (1 = male)	(0.086)	(0.071)	(0.087)	(0.025)	(0.072)	(0.083)
Race/ethnicity (1 = White)	−.032	−.103	−.059	.018	−.106	−.045
Race/ethnicity (1 = White)	(0.117)	(0.097)	(0.110)	(0.033)	(0.098)	(0.110)
Marital status (1 = married)	.083	−.127	−.070	.024	−.126	−.065
Marital status (1 = married)	(0.100)	(0.087)	(0.098)	(0.030)	(0.088)	(0.096)
Education (years)	.213***	.035^†	−.055**	.052***	.039*	−.055**
Education (years)	(0.023)	(0.020)	(0.019)	(0.006)	(0.019)	(0.020)
Employed full time	−.057	−.001	.252**	−.026	.003	.245**
Employed full time	(0.089)	(0.074)	(0.087)	(0.025)	(0.074)	(0.086)
Occupational prestige	.012**	−.001	.005	.002*	−.001	.004
Occupational prestige	(0.004)	(0.003)	(0.003)	(.001)	(0.003)	(0.004)
Annual family income (log)	.198***	.021	−.037	.047**	.023	−.040
Annual family income (log)	(0.059)	(0.046)	(0.052)	(0.017)	(0.046)	(0.051)
Network size (number of alters)	.046	.028	.029	.007	.032	.030
Network size (number of alters)	(0.033)	(0.026)	(0.034)	(0.009)	(0.026)	(0.031)
Role relationship (proportion of alters as family members)	−.544***	.063	.006	−.108**	.058	.005
Role relationship (proportion of alters as family members)	(0.120)	(0.097)	(0.118)	(0.035)	(0.098)	(0.120)
Accessed socioeconomic status
Average education of alters		.077**	−.047
Average education of alters		(0.033)	(0.035)
Proportion of alters with some college education or more					.254*	−.239^†
Proportion of alters with some college education or more					(0.106)	(0.122)
Athletic identity			−.179***			−.175***
Athletic identity			(0.045)			(0.045)
Intercept	−.479			−.717***
Intercept	(0.562)			(0.165)
Intercept 1		−.460	−2.432***		−.608	−2.291***
Intercept 1		(0.449)	(0.495)		(0.462)	(0.477)
Intercept 2		.550	−.235		.401	−.091
Intercept 2		(0.449)	(0.490)		(0.463)	(0.467)
Intercept 3		1.374**	.763		1.225**	.907^†
Intercept 3		(0.452)	(0.486)		(0.465)	(0.466)
Intercept 4		2.154***			2.006***
Intercept 4		(0.459)			(0.472)
R²/Pseudo R²	.353	.119	.088	.248	.119	.094

Note. Numbers in parentheses are standard errors.

†

p ≤ .10. *p ≤ .05. **p ≤ .01. ***p ≤ .001 (two-tailed test).

The results in model 2 are similar to those in model 1. The second indicator of accessed education (proportion of alters with some college education or more) was negatively associated with body weight ratings at a marginal significance level (−.239, p < .10; see equation 3 and model 2). But it had a positive association with athletic identity (.254, p < .05; see equation 2), which, in turn, was negatively associated with body weight ratings (−.175, p < .001). As results from both the Sobel test and the bootstrapping method indicated, its negative indirect association with body weight ratings through athletic identity was significant (−.044, p < .05). Furthermore, all three indicators of SES—education (.052, p < .001), occupational prestige (.002, p < .05), and family income (.047, p < .01)—had positive associations with this indicator of accessed education, but only education was associated with body weight ratings (−.055, p < .01; see equation 1). As results from both the Sobel test and the bootstrapping method showed, this indicator of accessed education together with athletic identity negatively mediated (−.002, p < .05) the association between education and body weight ratings.

Interaction Between Accessed Education and Gender

Finally, we examined the interaction between accessed education and gender by entering product terms of gender with two mean-centered indicators of accessed education, respectively, into the ordinal logistic regression model of body weight ratings (see equation 4 and Table 3). Results support one of the two interaction-with-gender hypotheses (H4b). Both indicators of accessed education positively interacted with being male (.226, p < .05; 1.238, p < .01; see models 1 and 2). They were associated with body weight ratings positively for men (.041, .243) but negatively for women (−.185, −.995). With more educated network members, men had higher body weight ratings, but women had lower body weight ratings. Supplemental analysis created a binary variable to indicate above-average body weight ratings (0 = below average or average, 1 = somewhat or considerably above average), and found similar results with one exception. Alters’ average education interacted with gender at a marginal significance level (p < .10).

Table 3.

Ordinal Logistic Regressions of Body Weight Ratings on Accessed Socioeconomic Status, Athletic Identity, Control Variables, and Interaction Terms (N = 974).

	Body weight ratings
Independent Variables	Model 1	Model 2
Age	.012*	.012*
Age	(0.005)	(0.005)
Gender (1 = male)	.161	.187
Gender (1 = male)	(0.180)	(0.174)
Race/ethnicity (1 = White)	−.084	−.064
Race/ethnicity (1 = White)	(0.241)	(0.236)
Marital status (1 = married)	−.128	−.136
Marital status (1 = married)	(0.194)	(0.190)
Education (years)	−.112**	−.112**
Education (years)	(0.041)	(0.042)
Employed full time	.467*	.455*
Employed full time	(0.186)	(0.183)
Occupational prestige	.009	.009
Occupational prestige	(0.007)	(0.007)
Family income (log)	−.063	−.062
Family income (log)	(0.111)	(0.107)
Network size (number of alters)	.038	.039
Network size (number of alters)	(0.065)	(0.062)
Role relationship (proportion of alters as family members)	−.006	−.015
Role relationship (proportion of alters as family members)	(0.249)	(0.247)
Athletic identity	−.305***	−.303***
Athletic identity	(0.080)	(0.081)
Accessed socioeconomic status
Average education of alters	−.185^†
Average education of alters	(0.093)
Proportion of alters with some college education or more		−.995**
Proportion of alters with some college education or more		(0.349)
Average education of alters × Male	.226*
Average education of alters × Male	(0.109)
Proportion of alters with some college education or more × Male		1.238**
		(0.456)
Intercept 1	−4.577***	−4.522***
Intercept 1	(1.138)	(1.062)
Intercept 2	−.750	−.666
Intercept 2	(1.119)	(1.024)
Intercept 3	1.135	1.232
Intercept 3	(1.120)	(1.032)
Pseudo R²	.040	.046

Note. Numbers in parentheses are standard errors.

†

p ≤ .10. *p ≤ .05. **p ≤ .01. ***p ≤ .001 (two-tailed test).

Conclusion and Discussion

Analyzing nationally representative data from the 2004 U.S. GSS, this study systematically examines four roles of accessed education in the social distribution of body weight ratings. Accessed education is associated with body weight ratings indirectly through athletic identity. Also, it mediates, together with athletic identity, the association between education and body weight ratings. Furthermore, the association between accessed education and body weight ratings is positive for men but negative for women. Findings in this study advance our theoretical understanding of network, socioeconomic, lifestyle, and gender disparities in body weight.

This study contributes to the existing literature in five ways, theoretically and methodologically. First, its findings support the extension of social capital theory (Lin 1982, 2001) into body weight through the pathway of lifestyle, and suggest the theoretical utility of integrating social capital theory with classic stratification theories on lifestyles (Bourdieu [1979] 1984; Cockerham 2005; Weber 1978). Although showing no evidence for the direct association between accessed education and body weight ratings, this study finds that accessed education is indirectly associated with body weight ratings through athletic identity in a negative direction. Individuals’ body weight can be associated with not only their own education, but also their network members’ education. Net of their own SES or personal capital, network size, role relationship, and other sociodemographic characteristics, adults living in a higher-SES network context perceive themselves as more athletic, and more athletic adults have lower body weight ratings. Also, lifestyle scholars tend to emphasize the decisive role of personal capital in the formation of lifestyles (Bourdieu [1979] 1984; Cockerham 2005; Weber 1978). This study sheds light on the independent association between social capital and athletic lifestyle net of personal capital. In particular, note that alters’ average education is positively associated with athletic lifestyle more significantly and strongly than ego’s own education. As these findings imply, policy interventions on lifestyle and body weight should not just target individuals but take into consideration socioeconomic attributes of social contacts. For a more complete picture of the role of accessed SES and the extension of social capital theory, future research needs to directly measure lifestyle, and examine other possible mechanisms linking accessed SES to body weight.

Two prior studies have mixed results on the direct effect of accessed SES. Christakis and Fowler (2007), as in our study, report that accessed education is not directly associated with body weight. Moore and colleagues (Moore, Daniel, Paquet, et al. 2009) find that accessed occupation is directly and negatively associated with body weight. Four major differences—respectively, in research designs, samples, societies, and measurements—between these three studies may account for their inconsistent results. Christakis and Fowler examine longitudinal community data of adults in Framingham, Massachusetts. Moore and colleagues analyze community data of young and middle-aged adults in Montreal, Canada. Our study examines nationally representative GSS data in the United States. Also, accessed SES is captured on the educational dimension through the name generator in the Framingham study and our study but measured on the occupational dimension through the position generator in the Montreal study. Furthermore, body weight is measured objectively in the two prior studies but subjectively in our study. Although results here cannot be considered conclusive and cannot reconcile the discrepancy in prior work, together with prior findings, they imply that accessed occupation may be more strongly related to body weight than accessed education. To address such inconsistency and achieve generalizability, future efforts need to collect nationally representative data on multidimensional measurement of accessed SES and diverse indicators of body weight.

Second, this study demonstrates the convertibility argument on the relationship between personal and social capital (Bourdieu [1983] 1986; Lin 2001; Song 2011). It expands our understanding of social precursors of social capital, and extends our knowledge of network mechanisms for persistent socioeconomic differences in body weight. As this study finds, accessed education together with athletic identity mediates the relationship between education and body weight ratings. Consistent with prior research demonstrating education as the most significant socioeconomic determinant of body weight (McLaren 2007), education is the only significant socioeconomic predictor of body weight ratings in this study. Although the other two indicators of SES—occupational prestige and annual family income—are also positively related to accessed SES, education is the only variable whose association with body weight ratings is mediated by accessed education and athletic identity. Results here indicate that a higher-education network context together with athletic lifestyle may help convert educational advantage to lower body weight, and that network attributes can be endogenous to individuals’ own structural locations. Also, these findings support the growing critique of healthism and neoliberalism (Broom 2008; Campos 2004; G. H. Williams 2003). Body weight and health lifestyle are not simply matters of personal responsibility or outcomes of individuals’ own choices. They are related to multiple layers of structural factors, such as education at the individual and network levels. To achieve a more comprehensive framework of the multilevel social production of body weight, future research should pay more attention to the mediating roles of network-level factors in linking individual-level social positions to body weight. Note that results vary by the dimension of accessed SES (Christakis and Fowler 2007; Moore, Daniel, Paquet et al. 2009b). Future research should collect data on all three dimensions of accessed SES (e.g., occupation, income, or wealth) and compare their mediating roles in the SES-weight relationship.

Third, this study finds the association between accessed education and body weight ratings to be in opposite directions for men and women, and adds to our comprehension of network mechanisms for the gendered construction of body weight. Dwelling in a higher-education network context, men are likely to have higher body weight ratings, while the opposite pattern applies to women. Social capital embedded at meso-level social networks bridges the micro and macro levels of society (Coleman 1990; Lin 2001). A higher-education network context can help produce and reproduce the gendered social norm of body weight—the slenderness ideal for women and the masculine breadwinner role for men—at the network level. The gendered network norm on body weight can be one network explanation for the higher prevalence of overweight or obesity among men than women (Hedley et al. 2004). Similar to some prior evidence on the puzzling, positive SES-weight relationship among men (McLaren 2007), the positive association between accessed education and body weight ratings among men calls for future research on direct examination of possible explanations.

Furthermore, our gender-specific findings refine Lin’s network-based theory of social capital (Lin 1982, 2001). Our study challenges the validity of one central assumption in that theory, that is, the social resources assumption. The social resource assumption conceives of accessed SES as a valuable nonredundant social resource. Accessed SES is hypothesized to have only a positive and protective function in generating instrumental (e.g., wealth, power, and reputation) and expressive (e.g., health and life satisfaction) returns. Its positive function has been the focus of, and demonstrated by, most of the relevant empirical studies (for reviews, see Lin 1999; Portes 1998; Song 2013). But our study finds that the role of accessed education for body weight ratings can be positive and negative, depending on gender. Accessed education can influence body weight through enhancing the body weight norm at the network level. But its function operates in opposite directions for the two gender groups because the body weight norm per se has gendered expectations. Our study adds to several recent studies that critically question the generalizability of the social resource assumption in social capital theory (Moore, Daniel, Gauvin, and Dubé 2009; Song 2014, 2015a, 2015b; Verhaeghe et al. 2012). To further develop the theory of social capital as accessed SES, future research should pay more attention to how the role of accessed SES can be contingent on other social factors.

Beyond its substantive theoretical significance, this present study has methodological implications for the measurement of accessed SES. Previous studies using the name generator measure two indicators of accessed SES on the educational dimension: alters’ average education and proportion of alters with certain educational levels or more (Acock and Hurlbert 1993; Haines et al. 2011; Song and Chang 2012). As this study shows, the use of the two indicators produces consistent results. Therefore, not only each alter but also higher-education alters matter for body weight. Some prior studies on health and well-being outcomes find measurement-specific results (Song 2015b; Song and Chang 2012). Their results vary by indicators of accessed SES. Taken together, these findings suggest that whether different indicators of accessed SES exert consistent effects is contingent on outcomes. Note that the name generator is not as efficient as the position generator in capturing the full range of accessed SES (Song and Lin 2009; Van der Gaag, Snijders, and Flap 2008). Future research should employ the position generator to more fully capture the impact of accessed SES.

This present study is only a starting point for systematically exploring multiple roles of accessed SES for body weight. It has four data limitations that call for caution in the interpretation of its findings and stimulate further investigation. First, the 2004 GSS data are cross-sectional. They do not allow us to reconcile the persistent debate over social causation versus social selection. The social causation perspective states that SES, accessed SES, and lifestyle affects body weight, while the social selection perspective argues vice versa. Our study proposes hypotheses from the social causation perspective for one methodological reason. The key explanatory variable—accessed education—was measured retrospectively over the last six months prior to the survey. Body weight ratings were measured at the survey time and serve more appropriately as the outcome variable. But due to the cross-sectional nature of the data, our study is in no way to causally confirm the social causation perspective, deny the social selection perspective, or reject the coexistence of these two perspectives. Social selection explanations are possible for us to interpret our significant findings here. People who are overweight or obese may be less likely to be physically active and develop athletic identity, either because of their own intentional withdrawal or discriminatory social exclusion by others (Trost et al. 2002). People may choose to adopt a more active and athletic lifestyle in the hope of meeting higher-SES contacts. People may make efforts to get included into higher-SES networks for the purpose of status attainment and mobility (Conley and Glauber 2007). Women may perform rigid weight control and management to get connected with higher-SES circles (Crosnoe et al. 2008). Men may identify themselves with masculine breadwinners and pay less attention to body weight for the goal of entering a higher-SES network. In addition, the negative association between education and body weight ratings may be partly due to weight stigma (Bombak 2014; Crosnoe and Muller 2004). For the purpose of causality tests, longitudinal data are needed.

Second, the 2004 GSS rests on interviewers’ visual categorization of respondents’ body weight. Ideally, body weight can be objectively measured as a continuous variable (e.g., body mass index [BMI]). GSS categorizes respondents’ body weight into four ordered groups. Categorical data lose finer information. Statistical analysis of categorical data is less powerful than that of continuous data (Agresti 2010). Despite our reasonable confidence with GSS interviewers’ visual evaluation, there can be measurement errors including those due to interviewers’ gender bias. Future research should use objective continuous measurement of body weight to replicate hypotheses in this study. Third, in this study, athletic identity serves as a subjective indirect indicator of lifestyle. Future work should measure lifestyle directly and objectively. Finally, the 2004 GSS uses the name generator to capture ego-centered networks. Ego’s report of alters’ education may be biased. The GSS does not probe egos to report alters’ body weight, the entry of which into statistical models can help test the robustness of the effect of accessed SES. The name generator tends to map networks characterized by strong ties (i.e., high-intimacy relationships; Marsden 1987). The position generator captures networks less constrained by strong ties (Lin et al. 2001). Future research should collect data on accessed SES simultaneously measured through both generators and examine whether results vary by these two network instruments.

Despite its limitations, the 2004 GSS provides the only nationally representative data of adults with information on body weight ratings and accessed SES. This study develops the theory of social capital as network members’ resources through combining it with prior work on three social determinants of body weight (i.e., SES, gender, and lifestyle). This study advances our knowledge of the interplay between network- and individual-level social determinants of body weight. It represents a major advance by shedding light on three nondirect functions of accessed education. Accessed education has a positive association with athletic identity that is negatively related to body weight ratings. It can work together with athletic identity as one mechanism for educational disparity in body weight ratings. It can enhance the operation of gendered body weight norm within personal networks.

Footnotes

Acknowledgements

An earlier version of this article was presented at the annual meeting of American Sociological Association in New York in 2013. The authors are indebted to Paul DiMaggio, Tom Smith, and Jibum Kim for their help with the data, and to Nan Lin, James Moody, Katharine Donato, Tony N. Brown, Patrick Royston, Yulia Marchenko, Linda K. Muthén, Ian White, Mike Brick, Jae-Kwang Kim, and anonymous reviewers for their helpful suggestions. The authors thank Katherine Tracy Everhart, Erin Bergner, Erika Leslie, and Kristen C. Forbes for their research assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Lijun Song is an associate professor in the Department of Sociology and Center for Medicine, Health, and Society at Vanderbilt University. Her major research interests include social networks, medical sociology, and social stratification. Her publications have appeared in journals such as Social Forces, Journal of Health and Social Behavior, Society and Mental Health, Social Science & Medicine, Social Networks, American Behavioral Scientist, and Research in the Sociology of Work.

Philip J. Pettis is a Ph.D. student in the Department of Sociology at Vanderbilt University. His research focuses on medical sociology, social networks, gay men’s health, body image, and sexuality.

Bhumika Piya is a Ph.D. candidate in the Department of Sociology at Vanderbilt University. Her research interests include international migration, environmental demography, and health disparities. Her dissertation examines the relationship between environmental stress and health of women and children in Bangladesh.

References

Acock

Alan C.

Hurlbert

Jeanne S.

1993. “Social Networks, Marital Status, and Well-being.” Social Networks 15(3):309–34.

Agresti

Alan

. 2010. Analysis of Ordinal Categorical Data. Vol. 656. 2nd ed. Hoboken, NJ: John Wiley and Sons.

Allison

Paul D.

2002. Missing Data. Iowa City, IA: Sage Publication.

Anderson

Cheryl B.

2004. “Athletic Identity and Its Relation to Exercise Behavior: Scale Development and Initial Validation.” Journal of Sport & Exercise Psychology 26(1):39–56.

Anderson

Cheryl B.

Mâsse

Louise C.

Zhang

Hong

Coleman

Karen J.

Chang

Shine

. 2009. “Contribution of Athletic Identity to Child and Adolescent Physical Activity.” American Journal of Preventive Medicine 37(3):220–26.

Arsalani-Zadeh

Reza

Macfie

Dave

McNaughton

Lee

Ramsey

Claire

Andrews

Mark

. 2010. “Accuracy of Estimating Body Weight in Immobile, Emergency Orthopaedic Admissions.” Nutrition & Dietetics 67(1):50–51.

Ball

Mishra

2006. “Whose Socioeconomic Status Influences a Women’s Obesity Risk: Her Mother’s, Her Father’s or Her Own?” International Journal of Epidemiology 35(1):131–38.

Baum

Charles L.

Ruhm

Christopher J.

2009. “Age, Socioeconomic Status and Obesity Growth.” Journal of Health Economics 28(3):635–48.

Bauman

Adrian E.

2004. “Updating the Evidence That Physical Activity Is Good for Health: An Epidemiological Review 2000–2003.” Journal of Science and Medicine in Sport 7(1):6–19.

10.

Berkman

Lisa F.

Glass

Thomas

Brissette

Ian

Seeman

Teresa E.

2000. “From Social Integration to Health: Durkheim in the New Millennium.” Social Science & Medicine 51(6):843–57.

11.

Bollen

Kenneth A.

Stine

Robert A.

1992. “Bootstrapping Goodness-of-fit Measures in Structural Equation Models.” Sociological Methods & Research 21(2):205–29.

12.

Bombak

Andrea E.

2014. “The ‘Obesity Epidemic’: Evolving Science, Unchanging Etiology.” Sociology Compass 8(5):509–24.

13.

Bordo

Susan

. 2003. Unbearable Weight: Feminism, Western Culture, and the Body. 10th anniversary ed. Berkeley, CA: University of California Press.

14.

Bourdieu

Pierre.

[1979] 1984. Distinction: A Social Critique of the Judgement of Taste. Translated by Nice

Cambridge, MA: Harvard University Press.

15.

Bourdieu

Pierre

. [1983] 1986. “The Forms of Capital.” Pp. 241–58 in Handbook of Theory and Research for the Sociology of Education, edited by Richardson

J. G.

Westport, CT: Greenwood Press.

16.

Brewis

Alexandra A.

Hruschka

Daniel J.

Wutich

Amber

. 2011. “Vulnerability to Fat-stigma in Women’s Everyday Relationships.” Social Science & Medicine 73(4):491–97.

17.

Broom

Dorothy

. 2008. “Hazardous Good Intentions? Unintended Consequences of the Project of Prevention.” Health Sociology Review 17(2):129–40.

18.

Burt

Ronald S.

1984. “Network Items and the General Social Survey.” Social Networks 6(4):293–339.

19.

Campos

Paul

. 2004. The Obesity Myth: Why America’s Obsession with Weight Is Hazardous to Your Health. New York: Gotham Books.

20.

Carpiano

Richard M.

2006. “Toward a Neighborhood Resource-based Theory of Social Capital for Health: Can Bourdieu and Sociology Help?” Social Science & Medicine 62(1):165–75.

21.

Chang

Virginia W.

Christakis

Nicholas A.

2003. “Self-perception of Weight Appropriateness in the United States.” American Journal of Preventive Medicine 24(4):332–39.

22.

Christakis

Nicholas A.

Fowler

James H.

2007. “The Spread of Obesity in a Large Social Network over 32 Years.” New England Journal of Medicine 357(4):370–79.

23.

Christakis

Nicholas A.

Fowler

James H.

2008. “The Collective Dynamics of Smoking in a Large Social Network.” The New England Journal of Medicine 358(21):2249–58.

24.

Cockerham

William C.

2005. “Health Lifestyle Theory and the Convergence of Agency and Structure.” Journal of Health and Social Behavior 46(1):51–67.

25.

Cohen

Jacob

Cohen

Patricia

. 1983. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum.

26.

Cohen-Cole

Ethan

Fletcher

Jason M.

2008. “Is Obesity Contagious? Social Networks vs. Environmental Factors in the Obesity Epidemic.” Journal of Health Economics 27(5):1382–87.

27.

Coleman

James S.

1990. Foundations of Social Theory. Cambridge, MA: Belknap Press of Harvard University Press.

28.

Conley

Dalton

Glauber

Rebecca

. 2007. “Gender, Body Mass, and Socioeconomic Status: New Evidence from the PSID.” Pp. 253–75 in Advances in Health Economics and Health Services Research 17: The Economics of Obesity, edited by Bolin

Cawley

Oxford, England: Elsevier.

29.

Courtenay

Will H

. 2000. “Constructions of Masculinity and Their Influence on Men’s Well-being: A Theory of Gender and Health.” Social Science & Medicine 50(10):1385–401.

30.

Crosnoe

Robert

Frank

Kenneth

Mueller

Anna Strassmann

. 2008. “Gender, Body Size and Social Relations in American High Schools.” Social Forces 86(3):1189–216.

31.

Crosnoe

Robert

Muller

Chandra

. 2004. “Body Mass Index, Academic Achievement, and School Context: Examining the Educational Experiences of Adolescents at Risk of Obesity.” Journal of Health and Social Behavior 45(4):393–407.

32.

de la Haye

Kayla

Robins

Garry

Mohr

Philip

Wilson

Carlene

. 2011. “Homophily and Contagion as Explanations for Weight Similarities among Adolescent Friends.” Journal of Adolescent Health 49(4):421–27.

33.

Durkheim

Emile

. [1897] 1951. Suicide: A Study in Sociology. Translated by Spaulding

John

Simpson

George

. New York: Free Press.

34.

Erickson

Bonnie H.

2003. “Social Networks: The Value of Variety.” Contexts 2(1):25–31.

35.

Fischer

Claude

Jackson

Robert Max

Steuve

C. Ann

Gerson

Kathleen

Jones

Lynne McCallister

Baldassare

Mark

. 1977. Networks and Places. New York: Free Press.

36.

Fletcher

Adam

Bonell

Chris

Sorhaindo

Annik

. 2011. “You Are What Your Friends Eat: Systematic Review of Social Network Analyses of Young People’s Eating Behaviours and Bodyweight.” Journal of Epidemiology & Community Health 65(6):548–55.

37.

Fowler

J. H.

Christakis

N. A.

2008. “Estimating Peer Effects on Health in Social Networks: A Response to Cohen-Cole and Fletcher; and Trogdon, Nonnemaker, and Pais.” Journal of Health Economics 27(5):1400–405.

38.

Gard

Michael

Wright

Jan

. 2005. The Obesity Epidemic: Science, Morality and Ideology. Abingdon, London: Routledge.

39.

Grzywacz

Joseph G.

Marks

Nadine F.

2001. “Social Inequalities and Exercise during Adulthood: Toward an Ecological Perspective.” Journal of Health and Social Behavior 42(2):202–20.

40.

Haines

Valerie

Beggs

John J.

Hurlbert

Jeanne S.

2011. “Neighborhood Disadvantage, Network Social Capital, and Depressive Symptoms.” Journal of Health and Social Behavior 52(1):58–73.

41.

Hedley

A. Allison

Ogden

Cynthia L.

Johnson

Clifford L.

Carroll

Margaret D.

Curtin

Lester R.

Flegal

Katherine M.

2004. “Prevalence of Overweight and Obesity among US Children, Adolescents, and Adults, 1999–2002.” Journal of the American Medical Association 291(23):2847–50.

42.

Hesse-Biber

Sharlene N.

2007. The Cult of Thinness. 2nd ed. New York: Oxford University Press.

43.

Homans

George C.

1950. The Human Group. New York: Harcourt Brace & Company.

44.

House

James S.

Landis

Karl R.

Umberson

Debra

. 1988. “Social Relationships and Health.” Science 241(4865):540–45.

45.

Hruschka

Daniel J.

Brewis

Alexandra A.

Wutich

Amber

Morin

Benjamin

. 2011. “Shared Norms and Their Explanation for the Social Clustering of Obesity.” American Journal of Public Health 101(S1):S295–S300.

46.

Kiernan

Michaela

Moore

Susan D.

Schoffman

Danielle E.

Lee

Katherine

King

Abby C.

Taylor

C. Barr

Kiernan

Nancy E.

Perri

Michael G.

2012. “Social Support for Healthy Behaviors: Scale Psychometrics and Prediction of Weight Loss among Women in a Behavioral Program.” Obesity 20(4):756–64.

47.

Lamont-Mills

Andrea

Christensen

Steven A.

2006. “Athletic Identity and Its Relationship to Sport Participation Levels.” Journal of Science and Medicine in Sport 9(6):472–78.

48.

Langenberg

Claudia

Hardy

Rebecca

Kuh

Diana

Brunner

Eric

Wadsworth

Michael

. 2003. “Central and Total Obesity in Middle Aged Men and Women in Relation to Lifetime Socioeconomic Status: Evidence from a National Birth Cohort.” Journal of Epidemiology & Community Health 57(10):816–22.

49.

Laumann

Edward O.

1966. Prestige and Association in an Urban Community. Indianapolis, IN: Bobbs-Merrill.

50.

Lazarsfeld

Paul F.

Merton

Robert K.

1954. “Friendship as Social Process: A Substantive and Methodological Analysis.” Pp. 18–66 in Freedom and Control in Modern Society, edited by Berger

Abel

Page

C. H.

New York: Octagon Books.

51.

Leahey

Tricia M.

LaRose

Jessica Gokee

Fava

Joseph L.

Wing

Rena R.

2011. “Social Influences Are Associated with BMI and Weight Loss Intentions in Young Adults.” Obesity 19(6):1157–62.

52.

Legh-Jones

Hannah

Moore

Spencer

. 2012. “Network Social Capital, Social Participation, and Physical Inactivity in an Urban Adult Population.” Social Science & Medicine 74(9):1362–67.

53.

Lim

Anthony C.

Kaufman

Bradley J.

O’Connor

John

Cunningham

Sandra J.

2013. “Accuracy of Weight Estimates in Pediatric Patients by Prehospital Emergency Medical Services Personnel.” The American Journal of Emergency Medicine 31(7):1108–12.

54.

Lin

Nan

. 1982. “Social Resources and Instrumental Action.” Pp. 131–45 in Social Structure and Network Analysis, edited by Marsden

P. V.

Lin

Beverly Hills, CA: Sage Publications.

55.

Lin

Nan

. 1999. “Social Networks and Status Attainment.” Annual Review of Sociology 25(4):467–88.

56.

Lin

Nan

. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge, England: Cambridge University Press.

57.

Lin

Nan

Dan

. 2008. “The Invisible Hand of Social Capital: An Exploratory Study.” Pp. 107–32 in Social Capital: An International Research Program, edited by Lin

Erickson

New York: Oxford University Press.

58.

Lin

Nan

Yang-Chih

Hsung

Ray-May

. 2001. “The Position Generator: A Measurement Technique for Investigations of Social Capital.” Pp. 57–81 in Social Capital: Theory and Research, edited by Lin

Cook

Burt

R. S.

New York: Aldine de Gruyter.

59.

Marcoux

Beth C.

Trenkner

Leslie L.

Rosenstock

Irwin M.

1990. “Social Networks and Social Support in Weight Loss.” Patient Education & Counseling 15(3):229–38.

60.

Marsden

Peter V.

1987. “Core Discussion Networks of Americans.” American Sociological Review 52(1):122–31.

61.

McLaren

Lindsay

. 2007. “Socioeconomic Status and Obesity.” Epidemiologic Reviews 29(1):29–48.

62.

McLaren

Lindsay

Kuh

Diana

. 2004. “Women’s Body Dissatisfaction, Social Class, and Social Mobility.” Social Science & Medicine 58(9):1575–84.

63.

McPherson

Miller

Smith-Lovin

Lynn

Cook

James M.

2001. “Birds of a Feather: Homophily in Social Networks.” Annual Review of Sociology 27:415–44.

64.

Moore

Spencer

Bockenholt

Ulf

Daniel

Mark

Frohlich

Katherine

Kestens

Yan

Richard

Lucie

. 2011. “Social Capital and Core Network Ties: A Validation Study of Individual-level Social Capital Measures and Their Association with Extra- and Intra-neighborhood Ties, and Self-rated Health.” Health & Place 17(2):536–44.

65.

Moore

Spencer

Daniel

Mark

Gauvin

Lise

Dubé

Laurette

. 2009. “Not All Social Capital Is Good Capital.” Health & Place 15(4):1071–77.

66.

Moore

Spencer

Daniel

Mark

Paquet

Catherine

Dub

Laurette

Gauvin

Lise

. 2009. “Association of Individual Network Social Capital with Abdominal Adiposity, Overweight and Obesity.” Journal of Public Health 31(1):175–83.

67.

Muthén

Linda K.

Muthén

Bengt O.

1998–2015. Mplus User’s Guide. 7th ed. Los Angeles, CA: Muthén & Muthén .

68.

Nakao

Keiko

Treas

Judith

. 1990. Computing 1989 Occupational Prestige Scores. Chicago, IL: NORC.

69.

O’Dea

Jennifer A.

Caputi

Peter

. 2001. “Association between Socioeconomic Status, Weight, Age and Gender, and the Body Image and Weight Control Practices of 6- to 19-year-old Children and Adolescents.” Health Education Research 16(5):521–32.

70.

Paxton

Susan J.

Schutz

Helena K.

Wertheim

Elenor H.

Muir

Sharry L.

1999. “Friendship Clique and Peer Influences on Body Image Concerns, Dietary Restraint, Extreme Weight-loss Behaviors, and Binge Eating in Adolescent Girls.” The Journal of Abnormal Psychology 108(2):255–66.

71.

Pescosolido

Bernice A.

2006. “Of Pride and Prejudice: The Role of Sociology and Social Networks in Integrating the Health Sciences.” Journal of Health and Social Behavior 47(3):189–208.

72.

Portes

Alejandro

. 1998. “Social Capital: Its Origins and Applications in Modern Sociology.” Annual Review of Sociology 24:1–24.

73.

Putnam

Robert D.

2000. Bowling Alone. New York: Simon & Schuster.

74.

Ross

Catherine E.

1994. “Overweight and Depression.” Journal of Health and Social Behavior 35(1):63–79.

75.

Ross

Catherine E.

Mirowsky

John

. 1983. “Social Epidemiology of Overweight: A Substantive and Methodological Investigation.” Journal of Health and Social Behavior 24(3):288–98.

76.

Ross

Catherine E.

Chia-ling

. 1995. “The Links between Education and Health.” American Sociological Review 60(5):719–45.

77.

Royston

Patrick

. 2005. “Multiple Imputation of Missing Values: Update of ICE.” The Stata Journal 5(4):527–36.

78.

Schafer

Markus H.

Ferraro

Kenneth F.

2011. “The Stigma of Obesity: Does Perceived Weight Discrimination Affect Identity and Physical Health?” Social Psychology Quarterly 74(1):76–97.

79.

Schmeer

Kammi K.

2010. “Household Income during Childhood and Young Adult Weight Status: Evidence from a Nutrition Transition Setting.” Journal of Health and Social Behavior 51(1):79–91.

80.

Schutte

Nicola S.

McNeil

Dominic G.

2015. “Athletic Identity Mediates between Exercise Motivation and Beneficial Outcomes.” Journal of Sport Behavior 38(2):234–52.

81.

Smith

Kirsten P.

Christakis

Nicholas A.

2008. “Social Networks and Health.” Annual Review of Sociology 34:405–29.

82.

Smith

Tom W.

Marsden

Peter V.

Hout

Michael

Kim

Jibum

. 2011. General Social Surveys, 1972–2010. Principal Investigator, Smith

Tom W.

Co-Principal Investigator, Marsden

Peter V.

Hout

Michael

. Sponsored by National Science Foundation. Edited by NORC. Chicago, IL: National Opinion Research Center; Storrs, CT: The Roper Center for Public Opinion Research, University of Connecticut.

83.

Sobal

Jeffery

Stunkard

Albert J.

1989. “Socioeconomic Status and Obesity: A Review of the Literature.” Psychological Bulletin 105(2):260–75.

84.

Sobel

Michael E.

1982. “Asymptotic Intervals for Indirect Effects in Structural Equations Models.” Pp. 290–312 in Sociological Methodology, edited by Leinhart

San Francisco, CA: Jossey-Bass.

85.

Song

Lijun

. 2011. “Social Capital and Psychological Distress.” Journal of Health and Social Behavior 52(4):478–92.

86.

Song

Lijun

. 2013. “Social Capital and Health.” Pp. 233–57 in Medical Sociology on the Move: New Directions in Theory, edited by Cockerham

William Carl

. Dordrecht, The Netherlands: Springer.

87.

Song

Lijun

. 2014. “Bright and Dark Sides of Who You Know in the Evaluation of Well-being: Social Capital and Life Satisfaction across Three Societies.” Pp. 259–78 in Social Capital and Its Institutional Contingency: A Study of the United States, Taiwan and China, edited by Lin

Nan

Yang-Chih

Chen

Chih-Jou

. London: Routledge.

88.

Song

Lijun

. 2015a. “Does Knowing People in Authority Protect or Hurt? Authoritative Contacts and Depression in Urban China.” American Behavioral Scientist 59(9):1173–88.

89.

Song

Lijun

. 2015b. “Does Who You Know in the Positional Hierarchy Protect or Hurt? Social Capital, Comparative Reference Group, and Depression in Two Societies.” Social Science & Medicine 136–137:117–27.

90.

Song

Lijun

Chang

Tian-Yun

. 2012. “Do Resources of Network Members Help in Help Seeking? Social Capital and Health Information Search.” Social Networks 34(4):658–69.

91.

Song

Lijun

Lin

Nan

. 2009. “Social Capital and Health Inequality: Evidence from Taiwan.” Journal of Health and Social Behavior 50(2):149–63.

92.

Song

Lijun

Son

Joonmo

Lin

Nan

. 2011. “Social Support.” Pp. 116–28 in The Sage Handbook of Social Network Analysis, edited by Scott

John

Carrington

Peter J.

London: Sage Publications.

93.

Thoits

Peggy A.

2011. “Mechanisms Linking Social Ties and Support to Physical and Mental Health.” Journal of Health and Social Behavior 52(2):145–61.

94.

Trogdon

Justin G.

Nonnemaker

James

Pais

Joanne

. 2008. “Peer Effects in Adolescent Overweight.” Journal of Health Economics 27(5):1388–99.

95.

Trost

Stewart G.

Owen

Neville

Bauman

Adrian E.

Sallis

James F.

Brown

Wendy

. 2002. “Correlates of Adults’ Participation in Physical Activity: Review and Update.” Medicine & Science in Sports & Exercise 34(12):1996–2001.

96.

Umberson

Debra

Montez

Jennifer Karas

. 2010. “Social Relationships and Health: A Flashpoint for Health Policy.” Journal of Health and Social Behavior 51(1 suppl):S54–S66.

97.

Valente

Thomas W.

Fujimoto

Kayo

Chou

Chih-Ping

Spruijt-Metz

Donna

. 2009. “Adolescent Affiliations and Adiposity: A Social Network Analysis of Friendships and Obesity.” Journal of Adolescent Health 45(2):202–204.

98.

Van der Gaag

Martin P. J.

Snijders

Tom A. B.

Flap

Henk D.

2008. “Position Generator Measures and Their Relationship to Other Social Capital Measures.” Pp. 27–48 in Social Capital: An International Research Program, edited by Lin

Erickson

New York: Oxford University Press.

99.

VanderWeele

Tyler J.

2011. “Sensitivity Analysis for Contagion Effects in Social Networks.” Sociological Methods & Research 40(2):240–55.

100.

Verhaeghe

Pieter-Paul

Pattyn

Elise

Bracke

Piet

Verhaeghe

Mieke

Van De Putte

Bart

. 2012. “The Association between Network Social Capital and Self-rated Health: Pouring Old Wine in New Bottles?” Health & Place 18(2):358–65.

101.

Verhaeghe

Pieter-Paul

Tampubolon

Gindo

. 2012. “Individual Social Capital, Neighbourhood Deprivation, and Self-rated Health in England.” Social Science & Medicine 75(2):349–57.

102.

Wardle

Jane

Griffith

2001. “Socioeconomic Status and Weight Control Practices in British Adults.” Journal of Epidemiology & Community Health 55(3):185–90.

103.

Warin

Megan

Turner

Karen

Moore

Vivienne

Davies

Michael

. 2008. “Bodies, Mothers and Identities: Rethinking Obesity and the BMI.” Sociology of Health & Illness 30(1):97–111.

104.

Weber

Max

. 1978. Economy and Society. Edited by Roth

Wittich

Berkeley, CA: University of California Press.

105.

Williams

Gareth H.

2003. “The Determinants of Health: Structure, Context and Agency.” Sociology of Health & Illness 25(3):131–54.

106.

Williams

Simon J.

1995. “Theorising Class, Health and Lifestyles: Can Bourdieu Help Us?” Sociology of Health & Illness 17(5):577–604.

107.

Yang

Xiaolin

Telama

Risto

Laakso

Lauri

. 1996. “Parents’ Physical Activity, Socioeconomic Status and Education as Predictors of Physical Activity and Sport among Children and Youths - A 12-Year Follow-up Study.” International Review for the Sociology of Sport 31(3):273–91.