Associations of Social Capital with Binge Drinking in a National Sample of Adults: The Importance of Neighborhoods and Networks

Abstract

Background.

While considerable research on adult binge drinking has focused on social influences, the potential role of social capital has been largely overlooked. This study examines the role of social capital, assessed in terms of both neighborhood and social network characteristics, in understanding adult binge drinking.

Methods.

Adults ages 30-80 were randomly drawn from the RAND American Life Panel and completed an online survey (analytic sample n=1,383). The main predictor variables were neighborhood cohesion, neighborhood order, and social network density. Associations of social capital with past month binge drinking (any, number of days) were examined, controlling for demographic characteristics.

Results.

Zero-inflated negative binominal regression analysis indicated that any binge drinking was more likely among adults who lived in highly ordered neighborhoods and who had denser social networks but was negatively associated with neighborhood cohesion. However, binge drinking was more frequent among those who lived in neighborhoods lacking order and who had sparser social networks, but had no association with neighborhood cohesion. Age was not found to moderate associations of social capital with binge drinking.

Conclusions.

Given that the associations of social capital with adult binge drinking behavior appear to differ by level of influence and type of drinking behavior, there is a need to gain a more nuanced understanding of these complex associations, including the mechanisms through which they operate.

INTRODUCTION

One in four U.S. adults age 26 and older report past month binge drinking,¹ a behavior associated with unintended injury, interpersonal violence, certain chronic diseases, alcohol use disorder, and other health risks.² Those age 35 and older consume half of the total binge drinks among U.S. adults³ and the number of drinks consumed by these binge drinkers has significantly increased in the past decade.⁴ Identifying factors influencing binge drinking in older age groups is critical to informing prevention and treatment efforts. This study examines the role of social capital in adult binge drinking.

Social capital refers to patterns of engagement, trust, and mutual obligation among individuals within social structures⁵ that represent the norms and networks that enable peoples’ cooperation and social participation.⁶ Social capital is thought to play a role in controlling deviance and promoting public welfare, and may function within the context of personal social networks and at broader levels such as neighborhoods or communities.⁷ Some studies find social capital to be a risk factor for adult binge drinking,^8,9 others find it to be protective,^10,11 and still others find evidence for both.^12,13 Mixed findings are perhaps to be expected, given multiple mechanisms through which social capital influences behavior such as direct peer influences, shaping social environments and exposures to risk, influencing the availability of support, and affecting beliefs and attitudes through social norms and monitoring mechanisms.¹⁴ Given these complexities, it is important to gain a better understanding of how different facets of social capital are associated with binge drinking behavior.

Social-ecological models emphasize the importance of examining multiple levels of influence on behavior.¹⁵ This study focuses on neighborhood and social network indicators of social capital to understand binge drinking. At the neighborhood level, social capital is often assessed in terms of cohesion (e.g., neighbors get along, look out for one another) and order (e.g., low crime, violence). There is some evidence that drinking is lower among individuals living in communities with higher cohesion and order,¹⁶ although the evidence is mixed. Community features may exert their influence on drinking through their impact on more proximal social processes involving one’s close social ties (e.g., exposure to others’ behavior, formal social control, social stress);¹⁷ as such, social network indicators of social capital are also important to consider. Network density, the extent to which our close ties know and interact with each other, is a common way of assessing this more proximal form of social capital.^18,19 However, little is known about the role of social networks in the drinking behavior of adults, as nearly all research on this topic has been conducted with college students.^20,21

This study examines associations of neighborhood and network indicators of social capital with past month binge drinking (any, frequency) in a nationally representative U.S. sample of adults age 30-80 to address three limitations in the existing literature. First, most studies have focused on young people, which may have limited applicability to drinking behavior of adults. Second, few studies have examined associations of social capital with both any binge drinking and frequency of binge episodes; as a result, more complex associations between social capital and binge drinking may be overlooked (e.g., greater social capital may increase the likelihood of any binge drinking by presenting more opportunities to engage in this often-social behavior,²² yet has a tempering effect on binge frequency due to social control received from others).²³ Finally, it is rare for both neighborhood and social network indicators of social capital to be examined within the same study; as such, it is unclear if neighborhood social capital is associated with binge drinking after controlling for more proximal network influences.

METHODS

Participants and Procedures

A random sample of 2,615 adults (ages 30-80) from the RAND American Life Panel (ALP)²⁴ were invited to participate, with the intention of closing the survey once 1,700 surveys were completed. The ALP is a nationally representative Internet panel of over 5,000 U.S. adults who were age 18 or older at recruitment into the panel. ALP members are recruited into the panel via probability-based sampling methods, either sampled by random digit dial (landline and cell phone) or address-based sampling; individuals cannot otherwise volunteer to participate. A further advantage over most Internet panels is that ALP respondents do not need Internet access when they are initially recruited (laptops and internet subscriptions are provided, if needed). The survey included personal and network assessment modules, and was closed after six weeks in the field (April 29 – June 9, 2019) with 1,771 completions of both modules. The analytic sample (n=1,312) excludes individuals who reported “don’t know” for any neighborhood item (cohesion n=323; order n=54), were in the “other race” category (due to small cell size, n=89), or were missing data on binge drinking (n=9) or any background covariate (income n=4; rural n=3). Participants provided informed consent, and study materials and procedures were approved by RAND’s Human Subjects Protection Committee. Sample demographics are reported in Table 1. Of those invited to participate, respondents were more likely to be non-Hispanic white, married, higher income, and older than non-respondents. Comparing the complete sample (N=1,771) to the analytic sample (N=1,312) indicated that non-Hispanic white, married, and rural individuals were less likely to be missing information on key variables. However, the regression model results presented here indicate that these variables were generally not significantly associated with the dependent variables, suggesting that bias from missingness was not a contributor to the overall observed effects.

Table 1.

Descriptives for Main Study Variables (unweighted analytic sample, n = 1,312)

Variable	Mean (SD) / (%)
Male	44%
Non-Hispanic White	77.1%
Hispanic	14.0%
Black	8.9
Married	65%
College graduate	53%
Income (in $10,000)	8.2 (SD = 5.4)
Rural neighborhood	23.9%
Age (in years)	56 years (SD = 13.62)
Neighborhood cohesion	12.9 (SD = 2.5)
Neighborhood order	0.5 (SD = .5)
Network density	0.4 (SD = 0.2)

Note. Neighborhood cohesion and neighborhood order are correlated at r = .54; network density is correlated with these neighborhood indicators at r ≤ .07.

Measures

Past month binge drinking was assessed as the number of days a participant consumed at least four (for women) or five (for men) drinks in a row in the past 30 days.

Neighborhood social capital was assessed in terms of social cohesion (4 items; α=.91; “People in my neighborhood… are willing to help their neighbors; look out for one another; can be trusted; generally get along with each other”)²⁵ and neighborhood order (7 items; α=.87; e.g., “My neighborhood is… safe; pleasant for physical activity; clean.”).²⁶ Items were rated from 1=strongly disagree to 4=strongly agree, reverse scored as needed, and summed such that higher scores indicated greater social capital. Due to being highly skewed, a median split was used to dichotomize the neighborhood order measure.

Network social capital was assessed by asking participants to name up to 10 people they interacted most often with in the past six months and then, for each unique pair of network members, asking whether they knew and interacted with each other (“ties”). Network density was calculated as the proportion of ties among network members relative to the total number of possible ties. Higher scores indicated greater social capital.

Background covariates.

Analyses controlled for age, gender, race/ethnicity, educational attainment, and household income (in $10,000), as well as rural vs. urban neighborhood status (obtained by linking participants’ current ZIP codes to the American Community Survey).

Analytical approach

We used zero-inflated negative binomial regression, which is appropriate for count data that may be overdispersed (the unweighted unconditional mean number of binge days was 0.6, with a variance of 7.8, indicating overdispersion), and zero-inflated versions of the model to address the fact that 86% of participants reported no binge drinking.²⁷ Zero-inflated negative binomial regression models have two sets of predictors: one is used in a logistic model to predict zero values (current non-bingers in this study), and the other (here we use the same set) is used in a negative binomial model that predicts counts of binges (which may be zero or some positive integer). All cases are used in both analyses, but are weighted based on the results of the logistic component of the model. For ease of interpretation, we present the logistic portion of the model in terms of predicting that the respondent is a binger (rather than an extra zero), and we provide odds ratios and incidence rate ratios as well as the regression coefficients in Table 2. Three models were estimated: Model 1 includes background and neighborhood social capital variables; Model 2 adds the network social capital variable; and Model 3 adds interactions of age with each social capital variable. Analyses include survey weights in order to match U.S. population demographic distributions,²⁴ and were conducted using Stata v15.1.

Table 2.

Results from Zero-inflated Negative Binomial Regression Models

	Model 1				Model 2
Predicting Positive Binge Status	OR	(95% CI)	b	(SE)	OR	(95% CI)	b	(SE)
Male (vs. female)	1.31	(0.36 - 4.78)	0.27	(0.66)	1.46	(0.36 - 6.00)	0.38	(0.72)
Hispanic (vs. White)	26.58	(0.48 - 1477.34)	3.28	(2.05)	83.10	(4.48 - 15413.33)	4.42	(1.49) †
Black (vs. White)	0.31	(0.04-2.73)	−1.17	(1.11)	0.22	(0.03 - 1.44)	−1.52	(0.96)
Married (vs. unmarried)	0.44	(0.09 - 2.73)	−0.83	(0.82)	0.46	(0.10 - 2.03)	−0.78	(0.76)
College graduate (vs. not)	0.14	(0.03 - 0.70)	−1.94	(0.81) *	0.23	(0.05 - 1.02)	−1.45	(0.75) *
Income	1.19	(0.96–1.47)	0.17	(0.11)	1.26	(1.04 0 1.53)	0.23	(0.10) *
Rural (vs. urban)	1.08	(0.33 - 3.58)	0.08	(0.61)	0.96	(0.31 - 2.94)	−0.04	(0.57)
Age	0.88	(0.78–0.99)	−0.13	(0.06) *	0.86	(0.76 - 0.97)	−0.15	(0.06) †
Neighborhood cohesion	0.63	(0.42–0.92)	−0.47	(0.20) *	0.53	(0.35 - 0.80)	−0.63	(0.21) †
Neighborhood order	5.53	(1.04–29.25)	1.71	(0.85) *	5.00	(0.89 – 28.07)	1.61	(0.88) #
Network density					1.04	(1.02 - 1.06)	0.04	(0.01) †
Predicting Counts of Binges	IRR	(95% CI)	b	(SE)	IRR	(95% CI)	b	(SE)
Male (vs. female)	2.48	(1.35-4.56)	0.91	(0.31) ‡	2.53	(1.41 - 4.56)	0.93	(0.30) †
Hispanic (vs. White)	0.90	(0.37 - 2.23)	−0.10	(0.46)	0.90	(0.46 - 1.76)	−0.1	(0.34)
Black (vs. White)	1.07	(0.40 - 2.86)	0.07	(0.50)	1.03	(0.41 - 2.59)	0.03	(0.47)
Married (vs. unmarried)	0.88	(0.44 - 1.74)	−0.13	(0.35)	0.67	(0.39 - 1.33)	−0.4	(0.35)
College graduate (vs. not)	1.27	(0.62 - 2.62)	0.24	(0.37)	1.05	(0.55 - 2.01)	0.05	(0.33)
Income	0.96	(0.89 - 1.04)	−0.04	(0.04)	0.94	(0.89 - 1.00)	−0.06	(0.03) #
Rural (vs. urban)	0.93	(0.45 - 1.93)	−0.07	(0.37)	1.00	(0.50 - 1.99)	.00	(0.35)
Age	1.01	(0.97 - 1.05)	0.01	(0.02)	1.01	(0.99 - 1.03)	0.01	(0.01)
Neighborhood cohesion	1.01	(0.88 - 1.16)	0.01	(0.07)	1.04	(0.92 - 1.17)	0.04	(0.06)
Neighborhood order	0.53	(0.29 - 0.97)	−0.64	(0.31) *	0.58	(0.32 - 1.05)	−0.54	(0.30) #
Network density					0.99	(0.98 - 0.99)	−0.01	(0.01) *

Note.

^#p < .10.

^*p < .05.

^†p < .01.

^‡p < .001.

RESULTS

Weighted prevalence of past month binge drinking was 18.7%. Model 1 indicates that living in a more cohesive neighborhood was associated with lower likelihood of being a binge drinker, but was not associated with binge frequency. In contrast, living in a more ordered neighborhood was associated with higher likelihood of being a binge drinker, but less frequent binge drinking. While older age was associated with lower likelihood of binge drinking, age was unrelated to binge frequency. When network density was added in Model 2, having a denser personal network was associated with higher likelihood of being a binge drinker, but less frequent binge drinking. The associations of binge drinking with neighborhood cohesion remained unchanged, and those with neighborhood order weakened to marginal significance (p<.10) after controlling for network density. Age was not associated with either binge drinking outcome after controlling for network density, nor was age found to moderate associations of social capital with binge drinking behavior in Model 3 (all ps>.10; not presented).

DISCUSSION

Adults living in neighborhoods where people get along, help and look out for one another had a lower likelihood of any past month binge drinking compared to those living in less cohesive neighborhoods. This is consistent with prior research showing a protective effect of neighborhood cohesion on alcohol use.¹⁶ However, among those who engaged in binge drinking, neighborhood cohesiveness was unrelated to the frequency of this behavior. Our results also indicate that adults living in a highly ordered neighborhood and with a denser social network were more likely to engage in any binge drinking, but reported less frequent binge drinking episodes. In our sample of 30-80 year olds, we found no evidence that these associations are moderated by age. Together, these findings add to a growing literature indicating that social capital is neither inherently “protective” nor “risky” in terms of its influence on adult binge drinking behavior; rather, the complexity of findings reflects the different mechanisms through which social processes operate.¹³

Interpreting these results within a social control framework,²³ living in a well ordered neighborhood may increase feelings of safety and opportunities to drink socially; when coupled with having a denser social network, this may increase the chance that binge drinking will occur. However, once the transition to binge drinking has been made, these aspects of social capital may constrain the frequency of binge drinking behavior. In contrast, living in a highly cohesive neighborhood may impact social norms and constrain behavior in such a way that the transition to binge drinking is less likely to occur, even if the opportunity to drink increases. By essentially serving as an ‘off switch’ for any binge drinking, neighborhood cohesion may be less relevant than other aspects of social capital to how frequently this behavior occurs. While the mechanisms underlying these associations need further examination, our results suggest that programs to reduce adult binge drinking may be most effective if they target both structural and cognitive components of social capital⁶ at the neighborhood and network levels.

Results should be interpreted in light of certain study limitations, including the exclusive reliance on self-report and lack of objective neighborhood socioeconomic data. In addition, the Model 2 associations with neighborhood order were marginally significant and should be interpreted with caution. Finally, the cross-sectional design precludes us from drawing conclusions about temporal associations between social capital and binge drinking.

This study provides an important initial examination of the role of neighborhood and network indicators of social capital in adult binge drinking behavior. Consistent with the larger literature on social capital and alcohol use, results highlight the complexities of these associations and the need for better understanding the mechanisms underlying them. Studies that use a more fine-grained approach to data collection, such as ecological momentary assessment,²⁸ may be particularly useful by illuminating how different aspects of social capital influence affective responses, drinking motivations, and binge drinking behavior in real time.

HIGHLIGHTS.

• Neighborhood and network social capital are both relevant to adult binge drinking

• Social capital is associated with both more and less binge drinking

• Associations differ depending on type of social capital and type of binge behavior

• Age was not found to moderate associations of social capital with binge drinking