Drinking and Neighborhood Contexts of Alcohol’s Harms from Others

Abstract

Aims

Negative consequences of alcohol (or secondhand effects) extend beyond drinkers to affect other people, including both known others (friends, family members, spouses/partners) and strangers. Secondhand effects of alcohol manifest across various social environments, including the places where people drink and the neighborhoods where they live. These neighborhoods are characterized by different levels of alcohol availability and degrees of residential social cohesion. Hence, social environments may confer risk or protect from harms from others’ drinking. The current study explores: (a) how drinking venues and neighborhood contexts relate to harms from other people’s drinking (both known others and strangers), and (b) whether these associations vary by gender.

Methods

Using pooled data from the National Alcohol Survey and National Alcohol’s Harms to Others Survey (N = 5425), we regressed harms from various drinking others on social environment characteristics (drinking venues, alcohol availability and social cohesion) for the full sample and separately by gender. We used the false discovery rate method to adjust for multiple testing.

Results

Overall, greater neighborhood social cohesion was associated with lower odds of harm from drinking others and, specifically, harm from drinking strangers. The effect of social cohesion was most pronounced for men.

Conclusions

Social cohesion was the most salient neighborhood factor associated with reduced alcohol-related harms from strangers. Directions for future research and policies to mitigate these harms are discussed.

Short Summary: Social cohesion protected against harm from drinking strangers in the full sample. Gender-stratified analyses showed social cohesion had the most pronounced effect for men.

The World Health Organization (WHO) endorsed a global strategy on alcohol that includes policy options and interventions to address the social consequences of drinking (World Health Organization 2010). This strategy includes an emphasis on alcohol’s harm to others, which includes the physical, mental and social harms people experience as a result of someone else’s alcohol consumption (Room et al. 2010). According to general population surveys, being insulted or humiliated and emotionally hurt are the most prevalent harms, followed by verbal abuse and arguments (Giesbrecht et al. 2010; Laslett et al. 2011; Nayak et al. 2019). More serious harms include injuries due to traffic crashes and aggression.

Physical and mental health effects of these harms include poor general health status and reduced well-being (Casswell et al. 2011), greater depression and distress (Greenfield et al. 2016), and reduced quality of life (Karriker-Jaffe et al. 2012). Although some evidence suggests harms from a stranger’s drinking may not always have negative effects (Karriker-Jaffe et al. 2012), harms from known drinkers, such as spouses/intimate partners and family members, are associated with significantly higher levels of distress among victims (Karriker-Jaffe et al. 2012).

Despite increasing attention to this public health issue, environments surrounding alcohol’s harms from various others remain relatively unexplored. There is a dearth of literature on the risky social environments contributing to these harms. A focus on broader contextual factors will elucidate circumstances under which individuals experience these harms. Thus, using a large, nationally representative survey in the USA, the current study explores the social environment, including drinking contexts and neighborhood area characteristics, that may be associated with harm from various others’ drinking experienced by men and women.

SOCIAL ENVIRONMENTS AND ALCOHOL-RELATED HARMS

Harms from others’ drinking are likely to occur in contexts that have social or physical characteristics that promote behaviors creating these harms. Routine activity theory (Cohen and Felson 2010) suggests harms might occur when offenders and possible victims interact in spaces that have reduced or absent guardianship; that is, a lack of protocols or persons situated in the setting to reduce victimization (Hollis et al. 2013). Such spaces include drinking venues and alcohol outlets such as bars and liquor stores where alcohol-related harms are likely to occur. In this study, we distinguish between two types of social environment factors that may be associated with alcohol-related harms: drinking venues and the neighborhood area.

Drinking venues

The places where people drink—public spaces, bars, parties or home settings—are associated with the frequency and quantity of consumption and, subsequently, alcohol-related harm (Hughes et al. 2011). For example, a study using several years of the US National Alcohol Survey (NAS) found those who drank heavily at bars and public venues were at greater risk of fights and drink driving than those who consumed relatively less in all settings (Nyaronga et al. 2009). Consistent with routine activity theory, contexts where heavy drinking is considered normative may foster assault perpetration and victimization by people who have been drinking (Wells et al. 2011).

Harms in specific drinking venues also vary by gender (Fillmore 1985; Kaplan et al. 2017). Fillmore (1985) noted alcohol-related violence was more common in the home for women and in bars and in public for men. A more recent study found that, for women, drinking in public was significantly associated with a greater likelihood of fights compared to drinking at home, whether their own or a friend’s or family member’s (Wells et al. 2005). For men, a similar drinking frequency was more likely to result in alcohol-related aggression among those who drank in public relative to private home settings (Wells et al. 2005). In this study, we expect people who engage in more frequent drinking in bars and public places will be more likely to report harms from drinking others.

Neighborhood alcohol availability

Harms from others’ drinking also may be associated with neighborhood area alcohol availability. Although overall alcohol outlet density is related to higher alcohol consumption and related damages such as traffic crashes, violence and suicide (Popova et al. 2009), findings from studies specifying alcohol outlet types vary (Gmel et al. 2016). Specifically, there may be differences in alcohol’s harms to others when on-premise establishments (e.g., bars and restaurants) are differentiated from off-premise outlets (e.g., liquor and convenience stores).

Extant work suggests violence has a stronger relationship with densities of off-premise establishments than with densities of bars (Gruenewald et al. 2006). Speaking to routine activities, off-premise outlets such as liquor and convenience stores may invite public drinking, and they typically have low overall guardianship (Hollis et al. 2013). As a result of reduced regulation around off-premise outlets, patrons may loiter outside to consume alcohol (Graham 2006). Therefore, the likelihood of being harmed by someone else’s drinking may be higher in areas with greater off-premise outlet density. On-premise outlets such as bars and nightclubs invite multiple drinkers to a single setting, in which drinking-related norms reduce inhibitions, alter socially appropriate behaviors, and enable violence (Wells et al. 2011). Guardians such as security personnel and other safety measures are in place within many of these on-premise environments, however, and these mitigate harms compared to off-premise outlets (Pridemore and Grubesic 2013). The current study explores associations between different types of alcohol outlets and harms from both known and unknown drinkers.

Neighborhood social cohesion

Certain neighborhood social processes may protect against harms caused by other people’s drinking. Through collective efficacy, or the ability of individuals within a community to work together to solve problems and maintain social order, secondhand alcohol-related harms may be mitigated (Greenfield and Jones 1993). Weitzman and Chen (2005) reported greater social capital was significantly related to lower risk of harms from others’ drinking. Although no research to date has examined neighborhood social cohesion in studies of alcohol-related harms, evidence suggests social cohesion may protect against such harms, as it is negatively related to violence (Sampson et al. 1997) and promotes social control of deviant behaviors (Collins et al. 2014). Cohesion also fosters a desire to adhere to community norms that enhance safety (Pratt and Cullen 2005), inhibiting behaviors such as heavy alcohol consumption (Gerich 2014) and discouraging public harms (Hirschfield and Bowers 1997). Thus, individuals should be less likely to experience harm from drinking others within a socially cohesive neighborhood.

Social cohesion commonly operates among known residents (neighbors and friends). Therefore, in neighborhood areas where there is high social cohesion, harms from known others’ drinking may be reduced. Effects may not extend to control of drinking strangers, as established community norms are likely to be unfamiliar to outsiders who do not reside in the area. It is possible, however, that residents in cohesive neighborhoods might intervene in public spaces to reduce harms caused by drinking strangers. Further, effects of social cohesion may be weaker in private settings, such as with a partner or family member who drinks heavily. We consider social cohesion in the examination of harms experienced from drinking strangers and known others.

CURRENT STUDY

The current study builds on prior research by examining how the social environment relates to secondhand harms due to various others’ drinking, regardless of one’s own drinking status. Study aims are to: (a) identify types of drinking venues that are related to increased risk of harms from drinking strangers and from known others; (b) investigate whether neighborhood area characteristics (alcohol availability, social cohesion) are related to harms from drinking others; and (c) explore whether associations between these social environmental contexts and harms from others’ drinking vary by gender.

Study design

We used pooled cross-sectional data from the 2015 US NAS and 2015 US National Alcohol’s Harm to Others Survey (NAHTOS). Both used the same dual-frame landline and mobile telephone sampling strategy and included oversamples of Black/African American and Hispanic respondents. In the landline sample, one respondent from each household was randomly selected, while in the mobile telephone sample, the individual answering was selected. The same measures relevant to the study questions were collected across both surveys. The overall cooperation rates (proportion who participated after being confirmed eligible) for the NAS and NAHTOS were 45.4% and 67.3%, respectively. The Institutional Review Boards of the authors’ agencies approved all protocols. Respondent home addresses were geocoded and linked with contextual data at the ZIP code level. A majority (68.7%) gave an address for geocoding, and the remaining respondents were matched to their ZIP code centroid. Data were weighted to represent the US general population using the 1-year 2013 American Community Survey (ACS) Public Use Microdata files. Analysis weights adjusted for sampling strategy and differential nonresponse.

Study sample

The analytic sample (N = 5425) was limited to respondents with ZIP code data who were under 65 years old, as high-risk drinking and secondhand alcohol’s harm from others are more prevalent among younger adults. In the weighted sample, about half (51.1%) were women, and the mean age was 44 years (SD = 13.26). The majority were non-Hispanic White (62.9%), followed by 16.5% Hispanic or Latino, 12.7% Black/African American, and 7.9% other race/ethnicity (including 2% missing race/ethnicity).

MEASURES

Outcome variables

Alcohol’s harms from others

Respondents reported on eight harms caused by a drinker in the past 12 months: Being harassed, bothered or insulted; feeling threatened or afraid; having had property vandalized; being in a traffic accident; being harmed physically; being pushed, hit or assaulted; family or marital problems; and financial problems due to someone else’s drinking. Harms were attributed to strangers (9.5%), friends or coworkers (6.6%) (hereafter referred to as ‘friends’), family members (5.7%), and spouses/intimate partners (4.0%) (‘partners’). Respondents may have attributed harms to more than one source. Overall, 21.7% of respondents reported having experienced any harm from a drinker.

Predictor variables

Drinking venues

Respondents indicated the frequency they spent drinking in four different venues, with responses ranging from ‘every day/nearly every day’ to ‘never’ (including those who were not drinkers). The final indicators were dichotomized into drinking at least once a month (vs. less often) in each venue: at in public places with friends (3.2%), bars (21.1%) and parties (16.3%). Respondents may be coded for frequent drinking in multiple venues.

Neighborhood area alcohol availability

Alcohol outlet data were drawn from ZIP code business patterns data (United States Census Bureau 2014b), which included the number of establishments in each ZIP code classified by business type. Outlet densities were generated using the count of alcohol outlets per square mile. The two alcohol outlet density variables included bars (M = 0.78 SD = 3.56) and off-premise outlets (liquor and convenience stores) (M = 1.90, SD = 3.62). Convenience stores were included as likely alcohol outlets only in states where alcohol sales were legal in this type of store; state alcohol control policies were coded using information from the Alcohol Policy Information System (National Institute on Alcohol Abuse and Alcoholism 2008) and National Alcohol Beverage Control Association handbooks (National Alcohol Beverage Control Association 2010). A national validation study showed composite business patterns data correspond well (r = 0.89) with alcohol outlet counts from official state records (Matthews et al. 2011).

Neighborhood social cohesion

Three questions assessed perceived connectedness and solidarity among residents. Respondents were asked about their agreement with statements on whether neighbors are willing to help one another, their neighborhood is close-knit, and neighbors are trustworthy. Responses used a 4-point Likert scale ranging from strongly disagree (1) to strongly agree (4); reliability was very good (Cronbach’s α = 0.83; M = 2.97, SD = 0.67).

Covariates

Self-reported covariates included gender, marital status (unmarried, which included being separated, divorced, widowed or never married [41.7%] vs. the reference group of being married or living with a partner), income (categorized as ‘up to $20,000’, ‘$20,001–$60,000’, ‘$60,001–$100,000’, and ‘$100,001 or more’, with a category for missing income [9.9%]), no college degree (69.7%; vs. the reference group of those with a college degree or higher), race/ethnicity (with White being the referent), and age.

Respondents’ own alcohol use was captured using drinking volume consumed per week (M = 3.31 drinks, SD = 9.70). This was based on a graduated quantity-frequency approach (Greenfield 2000). Respondents first reported on the maximum number of beverages consumed in the prior 12 months, categorized into ranges of 1–2, 3–4, 5–7, 8–11 and 12+ drinks per day. For each of the quantity ranges, a categorical frequency at each level of consumption was captured on a 7-point Likert-type scale coded as the category midpoint in days. Responses ranged from ‘never’ (coded as 0 days) to ‘every day or nearly every day’ (360 days). The quantity and frequency midpoint values were used to calculate the total yearly consumption, which was then divided by 52, resulting in the average weekly volume, in drinks.

High neighborhood area disadvantage

From the 2014 ACS 5-year estimates (United States Census Bureau 2014a), we used proportions of residents receiving public assistance, residents in poverty, adults without a high school diploma, and adults who were unemployed (excluding those not in the labor force) to create an index for concentrated disadvantage (Cronbach’s α = 0.84, M = 0.17, SD = 0.09). High disadvantage was designated for respondents living in ZIP codes for which the index was greater than one standard deviation above the mean; 8.8% of respondents lived in a highly disadvantaged area.

Metropolitan location

Rural–urban continuum codes provided information on whether respondents resided in a metropolitan area (83.9%) compared to all other county types.

Analyses

We used weighted binary logistic regression to assess whether drinking venues and neighborhood areas were associated with harms from others’ drinking. First, we examined gender differences in key predictors and outcomes given documented differences in drinking prevalence (Grant et al. 2017) and harm (Casswell et al. 2011; Laslett et al. 2011; Nayak et al. 2019) between men and women. Then, we estimated multivariate models with all respondents, followed by gender-stratified models. In models predicting harm from partners, we limited the analytic sample to respondents who reported having a spouse or partner at the time of the survey (n = 2919). There were no differences across the two surveys on the distributions of the key predictors or the outcome variables. Multilevel modeling was not warranted given lack of respondent clustering across ZIP codes.

To account for tests of multiple hypotheses on the same sample and the possibility of increased likelihood of false positives, we applied the false discovery rate (FDR) adjustment to generate q-values following model estimations. There were 30 tests total (six independent variables of interest and five outcomes using a given analytic sample). To calculate the q-values, we first ranked the significant P-values from smallest to largest and divided by 30 tests. This product was then multiplied by the alpha of 0.05 to create the threshold the P-value must meet in order to reject the null hypothesis (Benjamini and Hochberg 1995).

RESULTS

Bivariate analyses

Table 1 presents weighted bivariate comparisons of reported harms from drinking others and key predictors for men and women. There were significant gender differences in the prevalence of harm from a drinking stranger, family member and partner: Men reported higher rates of harm from strangers (10.9% vs. 8.2% for women, P < 0.05), whereas women reported higher rates of harm from family members (6.9% vs. 4.5% for men, P < 0.05) and partners (5.1% vs. 2.1% for men, P < 0.01). With respect to the social environment, women reported lower frequencies of drinking in all venues. There were no statistically significant gender differences in residential area alcohol outlet density or self-reported social cohesion.

Table 1.

Harm from various others in the past 12 months and key predictors by full sample and gender differences, <65 years old¹

	Full sample (N = 5425)	Men (n = 2285)	Women (n = 3190)	P
	%	%	%
Harm from others
Any harm (of 8)	21.68	21.12	22.22	ns
Stranger	9.50	10.87	8.18	*
Friend	6.60	7.13	6.08	ns
Family	5.73	4.50	6.91	*
Partner (n = 2919)	3.62	2.06	5.08	**
Frequent drinking in venue
Public	3.20	2.17	1.03	**
Bars	21.14	10.81	8.51	**
Parties	16.32	8.01	6.79	*
Neighborhood characteristics	Mean (SD)	Mean (SD)	Mean (SD)
Bar density	0.78 (3.56)	0.75 (3.19)	0.80 (3.80)
Off-premise density	1.90 (3.62)	1.81 (3.28)	1.95 (3.85)
Social cohesion	2.97 (0.67)	2.97 (0.65)	2.98 (0.68)

Note:  ¹Reported percentages are weighted; ns are unweighted.

^*** P < 0.001, ^**P < 0.01, ^*P < 0.05, ^†P < 0.10.

Multivariate analyses

Our first research question concerns whether drinking venues are related to any harm from others’ drinking, particularly alcohol’s harms from drinking strangers. Although odds of harm were slightly elevated for those who drank frequently in bars (any harm: OR = 1.39, P < 0.05; harm by strangers: OR = 1.49, P < 0.05), parties (any harm: OR = 1.44, P < 0.001; harm by friends: OR = 1.64, P < 0.05) and public settings (harm by friends: OR = 1.91, P < 0.05), after adjusting for multiple tests, the different drinking venues were not significantly associated with alcohol’s harms from others (all q > 0.05). Our second research question examined whether neighborhood area factors (alcohol availability and social cohesion) were related to alcohol’s harms from others. Neither bar nor off-premise outlet densities were related to harms caused by drinking others, but greater social cohesion was associated with significantly lower odds of reporting any harm (OR = 0.68, P < 0.001) and harms from drinking strangers (OR = 0.01, P < 0.001) (Table 2). These associations remained significant following FDR adjustment (all q < 0.05).

Table 2.

Full model predicting alcohol-related harms perpetrated by various others; population under 65 (N = 5425)

	Any harm	Stranger	Friend	Family	Partner (n = 2919)
	aOR (CI)	aOR (CI)	aOR (CI)	aOR (CI)	aOR (CI)
Frequent drinking in venue (≥1×/month)
Public	1.46 (0.91, 2.32)	1.10 (0.59, 2.06)	1.91 (1.03, 3.56)	1.06 (0.51, 2.21)	1.14 (0.26, 4.91)
Bars	1.39 (1.06, 1.83)	1.49 (1.01, 2.20)	1.38 (0.91, 2.09)	1.02 (0.63, 1.68)	1.30 (0.61, 2.78)
Parties	1.44 (1.10, 1.89)	1.34 (0.92, 1.95)	1.64 (1.08, 2.50)	1.14 (0.70, 1.85)	1.08 (0.42, 2.78)
Neighborhood characteristics
Bar density	0.98 (0.95, 1.02)	0.99 (0.96, 1.03)	1.00 (0.95, 1.04)	1.01 (0.92, 1.11)	0.87 (0.70, 1.09)
Off-premise outlet density	1.01 (0.97, 1.05)	1.04 (0.99, 1.10)	1.00 (0.94, 1.06)	0.96 (0.86, 1.06)	1.04 (0.95, 1.13)
Social cohesion	0.68 (0.59, 0.80)***	0.01 (0.48, 0.76)***	0.81 (0.61, 1.06)	0.78 (0.59, 1.03)	0.75 (0.49, 1.15)
High disadvantage	0.81 (0.56, 1.17)	0.80 (0.47, 1.36)	0.93 (0.54, 1.61)	0.77 (0.47, 1.27)	0.12 (0.05, 0.30)***
Metropolitan area	1.46 (1.06, 2.02)	1.24 (0.77, 2.02)	2.44 (1.27, 4.70)	1.15 (0.68, 1.95)	1.97 (0.81, 4.83)
Male gender	0.82 (0.67, 1.00)	1.25 (0.94, 1.66)	1.05 (0.75, 1.47)	0.57 (0.39, 0.84)	0.36 (0.18, 0.74)
Age	0.98 (0.97, 0.99)***	0.98 (0.97, 0.99)	0.98 (0.96, 0.99)***	0.99 (0.97, 1.00)	1.01 (0.98, 1.03)
Unmarried	1.37 (1.10, 1.71)	1.28 (0.93, 1.76)	1.39 (0.96, 2.02)	1.70 (1.18, 2.44)	-
No college degree	1.11 (0.87, 1.40)	0.95 (0.68, 1.32)	1.13 (0.75, 1.68)	1.40 (0.91, 2.16)	1.73 (0.95, 2.66)
Race/ethnicity (ref: White)
Black	0.89 (0.66, 1.19)	0.90 (0.60, 1.35)	0.94 (0.58, 1.53)	0.71 (0.44, 1.14)	1.01 (0.43, 2.36)
Latino	0.83 (0.62, 1.11)	0.91 (0.60, 1.35)	0.65 (0.40, 1.05)	1.12 (0.69, 1.80)	0.59 (0.25, 1.36)
Other/missing	1.20 (0.82, 1.76)	1.30 (0.82, 2.06)	0.79 (0.38, 1.64)	1.10 (0.57, 2.14)	3.62 (1.40, 9.37)
Income (ref: $100,001+)
Up to $20 k	1.37 (0.94, 1.99)	1.03 (0.60, 1.35)	2.00 (1.12, 3.59)	1.18 (0.64, 2.19)	2.14 (0.80, 5.74)
$20,001-60 k	1.23 (0.88, 1.72)	0.91 (0.60, 1.38)	1.59 (0.91, 2.80)	0.91 (0.51, 1.65)	1.32 (0.57–3.06)
$60,001-100 k	1.02 (0.71, 1.44)	1.30 (0.82, 2.06)	1.21 (0.66, 2.22)	1.18 (0.63, 2.22)	0.90 (0.35–2.30)
Income missing	0.83 (0.54, 1.29)	0.92 (0.51, 1.64)	1.76 (0.86, 3.62)	0.30 (0.13, 0.71)	0.48 (0.14–1.58)
Weekly drinking volume	1.02 (1.01, 1.02)***	1.01 (1.00, 1.03)	1.01 (1.00, 1.03)	1.02 (1.00, 1.03)	1.02 (0.99–1.05)

Note: ^***P < q, or P < 0.001.

Our final research question explored whether effects of social environment factors on harms from drinking others were different by gender. Gender-stratified models are presented in Table 3. For men, drinking at least monthly at parties was related to slightly greater odds of experiencing any harm from someone else’s drinking (OR = 1.51, P < 0.05) but this was not significant after adjusting for multiple tests (q > 0.05). In terms of neighborhood area characteristics, greater off-premise outlet density was related to slightly lower risk of harm from drinking friends for women (OR = 0.90, P < 0.05), but this was not significant after adjusting for multiple tests (q > 0.05). Greater social cohesion was related to lower odds of any harm from drinking others for both men (OR = 0.65, P < 0.001) and women (OR = 0.71, P < 0.001), as well as lower odds of harm from drinking strangers for both men (OR = 0.57, P < 0.001) and women (OR = 0.65, P < 0.05). After FDR adjustment to account for potential false positives, three associations with greater social cohesion remained statistically significant: lower odds of any harm from others’ drinking for both men and women, and lower odds of harm from drinking strangers for men (all q < 0.05).

Table 3.

Predicting alcohol-related harm from drinking others by gender

	Any harm	Stranger	Friend	Family	Partner
	aOR (CI)	aOR (CI)	aOR (CI)	aOR (CI)	aOR (CI)
	Men
Frequent drinking in venue (≥1×/month)
Public	1.40 (0.76, 2.58)	1.08 (0.50, 2.33)	2.21 (1.00, 4.92)	1.00 (0.40, 2.47)	1.30 (0.16, 10.41)
Bars	1.43 (0.97, 2.11)	1.58 (0.97, 2.57)	1.42 (0.84, 2.41)	1.00 (0.45, 2.22)	0.67 (0.14, 3.28)
Parties	1.51 (1.02, 2.22)	1.37 (0.86, 2.18)	1.60 (0.91, 2.80)	1.22 (0.58, 2.58)	3.42 (0.39, 30.01)
Neighborhood characteristics
Bar density	0.99 (0.95, 1.03)	1.02 (0.98, 1.07)	0.99 (0.93, 1.04)	1.04 (0.93, 1.16)	0.86 (0.59, 1.26)
Off-premise density	1.02 (0.96, 1.09)	1.02 (0.95, 1.10)	1.06 (0.64, 1.24)	0.93 (0.74, 1.17)	1.14 (1.00, 1.30)
Social cohesion	0.65 (0.52, 0.81)***	0.57 (0.43, 0.76)***	0.89 (0.64, 1.24)	0.81 (0.51, 1.28)	0.93 (0.46, 1.87)
	Women
Frequent drinking in venue
Public	1.29 (0.65, 2.57)	0.95 (0.30, 2.97)	1.13 (0.37, 3.49)	1.00 (0.25, 4.07)	0.29 (0.0, 1.50)
Bars	1.39 (0.94, 2.05)	1.36 (0.74, 2.51)	1.34 (0.68, 2.64)	1.11 (0.58, 2.13)	1.79 (0.83, 3.86)
Parties	1.28 (0.86, 1.89)	1.27 (0.70, 2.29)	1.64 (0.85, 3.19)	0.94 (0.47, 1.88)	0.74 (0.29, 1.89)
Neighborhood characteristics
Bar density	0.96 (0.91, 1.01)	0.96 (0.90, 1.01)	1.00 (0.92, 1.10)	0.95 (0.86, 1.04)	0.84 (0.62, 1.14)
Off-premise density	1.01 (0.94, 1.08)	1.07 (1.00, 1.15)	0.90 (0.82, 0.99)	0.98 (0.91, 1.07)	1.02 (0.91, 1.13)
Social cohesion	0.71 (0.57, 0.88)***	0.65 (0.45, 0.94)	0.72 (0.47, 1.12)	0.76 (0.53, 1.09)	0.72 (0.43, 1.20)

Note: Covariates included disadvantage, metropolitan area, gender, marital status, education, income, race/ethnicity, age, weekly drinking volume.

^*** P < q, or P < 0.001.

DISCUSSION

The current study investigated whether aspects of the social environment, including different drinking venues and certain neighborhood area characteristics, were related to harms from various others’ drinking. These analyses demonstrated that perceived social cohesion was the strongest correlate of alcohol-related harm from others above all other contexts, including frequent drinking in venues and local alcohol outlet densities. For the overall sample, social cohesion was related to lower odds of any harm, and this effect was more pronounced in relation to harms experienced from drinking strangers. Subgroup analyses additionally showed that social cohesion was most strongly associated with reduced harm from drinking strangers experienced by men, rather than with stranger-perpetrated harms experienced by women.

Our findings suggest social cohesion is more important for controlling harms from drinking strangers than for harms from known drinkers, thereby building on previous work on social capital and harms from others’ drinking (Weitzman and Chen 2005). Harms from strangers are prevalent (Nayak et al. 2019), but little is understood about experiences surrounding harms from these unknown drinkers. Men are more likely to report harm from strangers than from known drinkers (Laslett et al. 2011; Nayak et al. 2019), and men also report greater physical harm from drinking strangers, along with drinking friends, relative to intimate known drinkers (Nayak et al. 2019). This may be why social cohesion appears more protective against harms from strangers for men than for women.

It is important to further consider how perceptions of social processes vary by gender, as this may extend to social cohesion and harms from various drinking perpetrators. In consideration of social processes, an informative next step would be to investigate the types of harms that one incurs from strangers in various locales compared to known others, as well as the specific locations of the harms (for example, do harms occur in the local neighborhood or elsewhere?). Data on more specific harm types could target the types of drinking harms most prevalent within and around specific venues and neighborhood types (e.g., urban vs. rural). Because drinking patterns are determined by individual characteristics, it also would be worthwhile to further examine intersections of drinking venue preferences with gender.

Rather than representing the objective environment, social cohesion represents social processes within neighborhoods that may result in fewer harms from drinking. When individuals are familiar with and are willing to help their neighbors, harms from others’ drinking may be less likely to occur and/or effects of such harms may be less severe. Neighborhood organizations can foster social cohesion through local events and programs that encourage residents to meet and establish trust, as well as to build a desire to ‘look out for one another’ to mitigate the potential risks of alcohol-related harms in the community. Future research should consider modifying effects of social processes on environmental characteristics related to harms from drinking strangers, as well as explore other neighborhood processes (such as bonds between residents and social control) that may protect against alcohol-related harms from others in the community.

We expected to find social environmental influences on harms experienced from drinking others, but after adjusting for multiple tests, none of the observed associations with frequent drinking in bars, parties and public settings and densities of alcohol outlets remained statistically significant. Prior research has shown higher risk of assaults by other drinkers for men who drink heavily at parties in other people’s homes (Kaplan et al. 2017), and drinking in unfamiliar social settings is linked with higher levels of consumption (Kuntsche and Labhart, 2013) which, in turn, may lead to more alcohol-related problems (Vik et al., 2000) and harms to others (Wechsler et al., 2001). Because off-premise alcohol outlets, especially when concentrated, may be associated with higher rates of violence (Gruenewald et al. 2006), we also expected greater alcohol availability in the neighborhood area would be associated with increased risk of harm. However, results did not support our hypothesis. Alcohol outlets often are located in places of disadvantage (Romley et al., 2007), but there also is a lower prevalence of drinking in disadvantaged areas (Karriker-Jaffe et al., 2012). In our full sample analyses, greater neighborhood disadvantage was associated with reduced odds of harms from drinking partners. These countervailing effects should be examined in future studies to disentangle drinking venue and neighborhood contextual effects on alcohol’s harms to others.

There are some limitations to this study. First, these data are cross-sectional, and therefore causality may not be inferred from these results. Although the data used in the present study are useful for establishing relationships, they are not a substitute for prospective, longitudinal data. In addition to replicating these analyses with other datasets to validate the relationships between social cohesion and harms observed here, it would be worthwhile to collect repeated measures to investigate nuances of the effects of neighborhood characteristics on secondhand effects of alcohol over time. It also would be worthwhile to use qualitative inquiry and methods such as ecological momentary assessment to elicit detailed instances of harm from other drinkers that may not be adequately captured through a retrospective survey instrument. Second, the geographic catchment area for neighborhood characteristics is at the ZIP code level. Generally, ZIP codes are larger than the area an individual typically perceives to be their neighborhood, but given the scale of a nationally representative study, ZIP codes do provide sufficient variation for comparison of neighborhood area characteristics between respondents (Osypuk and Galea 2007). Finally, with respect to harms from drinkers, the NAS and NAHTOS do not assess whether reported harms occurred within the context of the respondents’ residential neighborhood. The harms also were aggregated, given the low frequency of some negative secondhand effects of drinking.

We add to the extant literature by considering neighborhood and social contexts of alcohol-related harms incurred in different settings. Although the current surveys did not go into detail on harms from others’ substance use aside from alcohol, new data suggest harms from others’ cannabis use may be common (Kerr et al. In press) and thus further investigation on the contexts of harms from co-use of alcohol and other drugs is warranted. We examined different perpetrators of harms from drinking and the ways in which harms differ across multiple contexts. Considering the complexity of these environmental factors will be an important step toward tackling the social consequences of drinking (World Health Organization 2010). Finally, given gender differences in alcohol consumption and alcohol-related harms, this work can inform policy options to more effectively target alcohol’s harm to others.

DATA AVAILABLITY STATEMENT

Datasets and codebooks from the US National Alcohol Survey Series can be requested at http://arg.org/nas-datasets. For more information about the US National Alcohol Harms to Others Survey, please contact Thomas K. Greenfield, PhD, at tgreenfield@arg.org.

FUNDING

This work was supported by the National Institutes of Health's National Institute on Alcohol Abuse and Alcoholism (grants R01AA0022791, T.K. Greenfield and K.J. Karriker-Jaffe, M-PIs, P50AA005595, W.K. Kerr, PI, and T32AA007240, S. Zemore, PI).

CONFLICT OF INTEREST STATEMENT

None declared.