Drinking Together: Implications of Drinking Partners for Negative Marital Quality

Abstract

Objective:

Concordant and discordant alcohol use has significant implications for negative marital quality among married couples. Yet, we know little about the frequency with which individuals drink with their spouse versus with other social partners, or the implications of drinking with one’s spouse and with others for negative marital quality.

Method:

Participants were from the third wave of the Social Relations and Health Study, in which 312 (59% women; ages 22–87) married/cohabitating respondents completed a baseline telephone survey followed by monthly web surveys regarding their alcohol use, drinking partners, and negative marital quality.

Results:

Multilevel models revealed that men and women drink more often with their spouse than with others. Drinking any amount of alcohol with one’s spouse for a greater proportion of months was associated with lower negative marital quality by gender such that more frequent drinking with others (family and nonfamily) was associated with lower negative marital quality (e.g., less criticism and fewer demands) among men and women. The effects of drinking with others varied among women and higher negative marital quality among men in the next month.

Conclusions:

These findings reveal that drinking with one’s spouse may partly explain why concordant drinking spouses report lower negative marital quality. Furthermore, drinking with others may be beneficial for women’s marriages but harmful for men’s. The findings are consistent with the concept of the drinking partnership, which suggests that drinking with one’s spouse is often linked to positive outcomes, including lower negative marital quality.

Alcohol consumption IS on the rise among adults in the United States, with increases in the number of days of light (1–2 drinks per day) and moderate (3–4 drinks a day; Kerr et al., 2014) drinking. Alcohol use is often a social phenomenon in which social network members tend to drink similar amounts of alcohol (e.g., Leonard & Homish, 2008; Leonard et al., 2000). One of the most important social contexts for examining alcohol is the marital tie, because most adults enter into married or cohabiting partnerships (U.S. Census Bureau, 2013). It is well known that spouses influence each other’s health behaviors, including drinking (Polenick et al., 2018; Roberts & Leonard, 1998).

Research shows that couples with discordant drinking patterns (e.g., a heavy drinking spouse paired with a spouse who does not drink or drinks light amounts of alcohol) report worse marital outcomes, including lower marital quality and increased divorce rates (Cranford et al., 2011; Leonard et al., 2014; Ostermann et al., 2005). Beyond the couple relationship, wives and husbands often have other social drinking partners (e.g., family members and friends), which may further affect the quality of their marriage (Leonard & Homish, 2008). Prior work on the social context of drinking has largely focused on heavy alcohol use among young or newlywed couples. However, knowledge has been limited about how any amount of drinking with one’s spouse and with others is linked to marital quality over the adult life span. Furthermore, although several studies have examined drinking patterns within marriage and intimate partner violence (e.g., Leadley et al., 2000; Quigley & Leonard, 2000), little research has considered how alcohol use is linked to more typical negative relationship qualities, including criticism and demands (with some exceptions, see Homish & Leonard, 2005; Levitt & Cooper, 2010). Negative elements of marriage (e.g., partner criticism or demands) are often more consequential for marital longevity, health, and wellbeing than the positive dimensions (Karney & Bradbury, 1995; Kurdek, 1995; Orbuch et al., 2002; Umberson et al., 2006). Given that couples in which both spouses report any amount of drinking show less negative marital quality across 4 years (Birditt et al., 2018), it is imperative to consider how drinking with one’s spouse and with other social partners may be differentially linked to negative marital quality over time.

The present study extends the literature by examining the frequency of drinking with one’s spouse versus with others (family and nonfamily) and the implications for negative marital quality on a monthly basis among adults. We addressed two research questions: (a) Do individuals drink with their spouse more frequently than with other social partners, and does this vary by gender, and (b) what are the implications of drinking with one’s spouse versus drinking with others for negative marital quality in the same month and in the next month, and is this relationship different for men and for women?

Theoretical framework

The present study was guided by compatibility theory, the concept of the “drinking partnership,” and gender role theory. Compatibility theories of marriage suggest that members of couples who are more similar to one another in their goals and behavior have greater marital satisfaction (Levinger & Rands, 1985). Roberts and Leonard (1998) introduced the “drinking partnership” concept, which refers to the matching or concordance of each individual’s drinking within marital dyads (e.g., do both members of the couple drink similar amounts?), as well as whether couples drink together or separately. According to this perspective, whether both members of the couple consume alcohol, the amount of alcohol that partners consume, and the extent to which they drink together are all important for marital functioning. Couples in which both spouses drink and drink together, for instance, may have better marital quality because of their shared leisure and social activities, along with similar beliefs about the intimacy-enhancing effects of alcohol—resulting in more frequent positive marital interactions (Derrick et al., 2010; Homish & Leonard, 2007). There are some nuances with regard to drinking partnerships that may affect marital quality. In particular, concordant heavy drinking couples tend to be happier than discrepant drinking couples, but not necessarily happier than those who are concordant abstinent or light drinkers (Leonard et al., 2014; Torvik et al., 2013, 2015).

Gender role theory suggests that men and women have distinct roles, and that women’s identities are more linked to nurturance and relationships (Bernard, 1972). These culturally influenced gender stereotypes are also associated with alcohol consumption. Women are less likely to consume alcohol, drink less, and are less likely to have alcohol-related problems than men (Chan et al., 2007; Drum et al., 2009). Drinking among women is often culturally disapproved, because drinking is associated with masculinity, aggression, power, increased sexual behavior, and risk taking (Wilsnack et al., 2005). Accordingly, patterns of drinking, and the implications of drinking for marital outcomes, may vary between men and women.

Drinking with one’s spouse versus with others

Few studies have examined the frequency of drinking with spouses versus with other social partners among married individuals. Research on newlyweds has shown that couples vary widely in their amount of drinking and in whether spouses drink together or separately (Roberts & Leonard, 1998). Prior work has found that alcohol use may remain more stable for people who have friends or relatives with similar drinking patterns (Moos et al., 2010; Orford et al., 2002), suggesting that married adults may drink with multiple social partners. Indeed, the concept of “drinking buddies” (i.e., people who meet on a regular basis to engage in activities centered on drinking) has been examined among married partners. Heavy drinkers tend to have more drinking buddies in their networks (Leonard et al., 2000). A greater number of drinking buddies has been found to predict greater heavy alcohol use and alcohol problems among couples over the first 4 years of marriage (Leonard & Homish, 2008). Nevertheless, the majority of studies have focused on heavy drinking. As a result, there is little understanding of how often individuals consume any alcohol with their spouse versus with others (e.g., children, siblings, friends, neighbors).

There may also be gender differences in preferred drinking partners. Women report more frequent drinking with their spouse than do men (Erol & Karpyak, 2015; Graham & Braun, 1999; Levitt & Cooper, 2010). In contrast, relative to women, men drink more often in public and with their peers (Erol & Karpyak, 2015), and husbands tend to report having more drinking buddies than do wives (Leonard & Homish, 2008). It is not clear, however, whether there are gender differences in the frequency with which married individuals drink with their spouse versus with other social partners.

Effects of drinking with one’s spouse versus with others on marital quality

Research on newlyweds reveals that drinking with one’s spouse is associated with better marital quality than drinking apart or discordant drinking. For example, Roberts and Leonard (1998) identified clusters of couples in their first year of marriage (ages 18–29) and found that couples who consumed alcohol together reported better marital functioning and intimacy. These findings remained after the amount of alcohol consumed was controlled for—suggesting that drinking together (even high quantities) was distinctly associated with greater marital satisfaction. In a later longitudinal study, Homish and Leonard (2005) compared the marital quality of three groups of newlywed couples who drank similar amounts of alcohol but engaged in different drinking contexts. Couples who drank together more often than they drank apart reported greater marital satisfaction than couples who drank apart more than they drank together. Wives who drank with their husbands also reported less of a decline in marital satisfaction over time than wives in nondrinking couples, whose declines were similar to those of wives who more often drank apart from their husbands. Similarly, for wives but not husbands, frequent (but not heavy) alcohol use was linked to greater intimacy and increased marital satisfaction (Roberts & Linney, 2000). Of note, these studies focused on young adult or young newlywed samples and did not examine the frequency of drinking with other social partners. Consequently, the implications of drinking together versus drinking with others for marital quality among individuals beyond young adulthood are not known.

Studies of drinking buddies and relationship satisfaction among younger couples show that husbands and wives who were less satisfied with marriage reported more drinking buddies over time (Homish & Leonard, 2008)—implying that drinking with friends or with other nonfamily drinking partners may be linked to poor marital quality. Another study assessed daily use of alcohol and relationship processes among college student romantic couples over 3 weeks (Levitt & Cooper, 2010). People who drank any amount of alcohol with romantic partners reported feeling closer to their partner and perceived fewer negative partner behaviors on the next day, compared with drinking apart from one’s partner or not drinking. Women who drank any amount of alcohol apart from their partner also reported greater negative partner relationship events on the next day, whereas this link was not found for men. Thus, among romantic/married couples, drinking with other social partners predicts lower quality partner relationships, especially among women, but this research has not considered adults of all ages.

Present study

This study examined participants who ranged in age from young adulthood to old age who reported whether and how often they drank with social partners (i.e., spouse, vs. others) and their negative marital quality on a monthly basis. We aimed to address the following two research questions: (a) Do adults report drinking with their spouse more often than with other social partners, and (b) is drinking with one’s spouse and/or drinking with others associated with negative marital quality? Regarding the first research question, we hypothesized that adults would report drinking with their spouse more frequently than drinking with others. We further hypothesized that the frequency of drinking with one’s spouse would be greater among women, whereas men would drink more often with others. Regarding the second question, we predicted that drinking with one’s spouse would be associated with lower negative marital quality, and that drinking with others would be associated with greater negative marital quality. We also predicted that these links would be stronger among women than among men.

Method

Participants

The participants were from Wave 3 of the longitudinal Social Relations and Health Study (SRS; Antonucci, principal investigator). All waves of data were collected by Survey Research Operations at the Institute for Social Research, University of Michigan. The Wave 1 SRS sample was drawn from the tri-county Detroit metropolitan area (ages 8–93; N = 1,703; response rate = 72%) and was completed in 1992. Wave 2, completed in 2005 using both telephone and faceto-face interviewing, included 1,076 of the original participants with 320 identified as deceased (n = 1,076, response rate = 78%). Wave 3, completed in 2015 using telephone interviews, included 720 of the original participants (383 of the original respondents were deceased, response rate = 73%). Participants received $25 for completing the phone interview. We conducted multinomial logistic regressions to determine what factors may have influenced attrition in the panel sample from Wave 1 to Wave 3. Attrition could be due to death or dropout for any other reason. Those who were deceased were older (odds ratio [OR] = 1.11, p < .001), more likely to be male (female coded as 1, male coded as 0, OR = 0.53, p < .001), and more likely to be less educated (OR = 0.84, p < .001) compared with those who completed the Wave 3 interview. Those who dropped out were more likely to be male (OR = 0.70, p < .01) compared with those who completed the Wave 3 interview. Because of the wide age range, which included children, and the fact that many participants were not married or living with a partner in Wave 1, we did not include negative marital relationship quality as a predictor in the attrition analysis. The attrition analysis shows that the Wave 3 sample was not a randomly representative group of individuals who participated in Wave 1.

Participants completed a social network diagram in Wave 3, in which they listed the people with whom they felt close, in order of closeness, using three concentric circles (Antonucci, 1986; Kahn & Antonucci, 1980). People whom respondents “could not imagine life without” were placed in the first circle, and people who were “less close” were placed in the second and third circles. After completing the interview, Wave 3 participants were invited to provide contact information for the first three people listed in their social networks. These network members (referred to hereafter as the core network sample) were then invited to participate in the study.

After completing their interviews, both the panel and the core network sample were asked to complete monthly web surveys. Participants received $5 for each monthly survey, with an additional $5 incentive if they completed 6 web surveys, an additional $10 if they completed 9 web surveys, and an additional $15 if they completed all 12 web surveys. The web survey was offered for 15 months from October 1, 2014, to December 10, 2015. The respondents who completed their baseline interviews between July 2014 and December 2014 were eligible for 12 monthly surveys.

A total of 727 respondents were married or cohabiting in Wave 3 (463 panel respondents and 264 core network members). A total of 312 participants completed at least two web surveys and had complete data. Eight participants were removed because of missing data. Those who were married (92%) had known each other for an average of 31 years (SD = 15.4), and those who were cohabiting (8%) had known each other for an average of 12 years (SD = 9.8).

The data analyzed in the present study are from Wave 3 interviews and the web surveys. Table 1 presents participant characteristics and scores on key variables for the sample of 312 participants. In Wave 3, individuals ranged in age from 22 to 87, 59% were women, 81% were White, and 12% were Black. A total of 18% were young adults (22–39), 39% were middle-aged adults (40–59), and 43% were older adults (60–87). A selection analysis using logistic regression that compared married/cohabiting respondents who were not included in the analysis with those who were included revealed that respondents in the sample were older (OR = 1.02, p < .01), more educated (OR = 1.17, p < .001), more likely to be female (OR = 1.42, p < .05), and more likely to be core network members (core coded as 0, panel coded as 1, OR = 0.47, p < .001). Negative marital quality in Wave 3 was not associated with inclusion in the analysis for this study (OR = 0.93, p > .05).

Table 1.

Characteristics of married or cohabiting women and men in the SRS web study

	Women (n = 183)			Men (n = 129)
Variable	M	SD	Range	M	SD	Range
Age, in years	55.42	14.03	22–87	57.21	15.86	22–87
Education, in years	14.51	2.32	4–17	14.94	2.25	5–17
Monthly web surveys
Number of web surveys	7.44	3.68	2–12	7.07	3.62	2–12
Average drinks per week	2.53**	4.14	0–30	5.72	6.90	0–30
Frequency of drinking with spouse	2.77	1.11	1–5	2.91	1.20	1–5
Frequency of drinking with others	1.82	0.83	1–4.5	1.85	0.77	1–4
Monthly negative marital quality	2.27	1.06	1–5	2.26	1.03	1–5

Notes: Frequency of drinking values were 5 = daily; 4 = 3 or more times a week; 3 = 1–2 times a week; 2 = 2–3 times over the last month; 1 = once over the last month. Negative marital quality ranged from 1 = disagree to 5 = agree. SRS = Social Relations and Health Study.

Measures

Alcohol use. Participants’ average number of drinks per week in the past month was assessed as follows: Participants were first asked whether they drank any alcoholic beverages such as beer, wine, or distilled spirits in the past month (yes or no). Participants who reported drinking were then asked, on average, in the last month, (a) how many days per week they drank alcohol and (b) about how many drinks they had consumed on the days they drank. This measure is used in other national data sets, including the Health and Retirement Study. Although the participants were not reminded in the web surveys of how a drink was defined, this information was provided in the Wave 3 baseline interview. Specifically, interviewers stated, “When I use the word ‘drink’ in the next questions, I mean either a glass of wine, a can or bottle of beer, or a shot or jigger of liquor either alone or in a mixed drink.” Average drinks per week were capped at 30 to adjust for skewness.

Participants then reported the social partners with whom they drank. Each month, participants were asked to (a) list the names and relationships of up to five people with whom they drank alcohol in the last month and (b) report how often they drank with each person using the following choices: 1 = daily, 2 = 3 or more times a week, 3 = 1 to 2 times a week, 4 = 2 to 3 times over the last month, 5 = once over the last month. Items were reverse coded so that higher scores reflected more frequent drinking. We then created scores for monthly frequency of drinking with (a) one’s spouse or cohabiting partner (referred to hereafter as one’s spouse) and (b) other relationships (e.g., family, such as parents, siblings, adult children; and nonfamily, such as friends, coworkers). Because participants could report drinking with multiple other people, we averaged their monthly drinking frequency with all people other than one’s spouse.

Negative marital quality.

Using a scale from 1 (disagree) to 5 (agree), each month participants reported the extent to which their spouse/partner: (a) gets on their nerves; (b) makes too many demands; (c) lets them down when counted upon; and (d) criticizes them. Mean scores were calculated (mean α = .81; α range: .74–.88).

Gender.

Gender was identified as 1 (female) or 0 (male).

Covariates.

We controlled for baseline age and years of education, because studies show that the frequency of drinking is lower as people age, and individuals who are more educated tend to drink more frequently (Assari & Lankarani, 2016). Age was included as a continuous variable. Participants reported the highest grade or year of education they had completed. We also controlled for whether the respondent was a panel member or a core network member (1 = panel, 0 = core network member), the month of survey, and average number of drinks per week.

Analysis strategy

Descriptives on key variables were calculated in preliminary analyses. Because the data were from panel respondents and their identified close social network members over time, we first estimated multilevel models without predictors to estimate the nonindependence with intraclass correlations (Kenny et al., 1998). These models included three levels in which Level 1 was the month, Level 2 was the participant, and Level 3 was the social network (e.g., panel and core network members). Multilevel models were estimated in STATA 15 (StataCorp LP, College Station, TX).

Next, to examine whether frequency of drinking varied by social partner (i.e., spouse vs. others), we estimated a multilevel model predicting monthly frequency of drinking with type of social partner as the predictor (1 = spouse, 0 = others), controlling for participant gender, baseline age, education in years, whether the respondent was a panel member or a core network member, and month of survey. Models were estimated in two steps, in which we first examined the main effects, followed by an examination of the interaction term (Social Partner × Gender) to determine whether the social partner–frequency of drinking link varied between women and men. These models included three levels, in which Level 1 was the month, Level 2 was the participant, and Level 3 was the social network (e.g., panel and core network members). We estimated models with four levels, in which the lowest level was the social partner but there was nonsignificant variance. In the interest of parsimony, we present results here from the three-level model.

To consider how drinking with each social partner is associated with negative marital quality, we estimated a series of multilevel models. These models included three levels, in which social network was Level 3, participant was Level 2, and month was Level 1. First, we assessed whether drinking any amount with one’s spouse or with others, in the same and previous month, was associated with within-person variations in negative quality. These models included dichotomous predictors, in which each month was coded as drank with one’s spouse (1 = yes, 0 = no) and drank with others (1 = yes, 0 = no). This model also included a between-person variable (percentage of months one drank with one’s spouse or with others), which allowed for a between-person assessment of whether individuals who drink more frequently with their spouse or with others report higher or lower negative quality relationships. Nondrinkers were included in this analysis. Drinking with one’s spouse and drinking with others were estimated in separate models.

Next we assessed whether the frequency of drinking with one’s spouse or with others in the same and previous month was associated with within-person variations in negative marital quality. These models also included the between-person effect of average frequency of drinking either with one’s spouse or with others, which allowed for a between-person assessment of whether people who report drinking with their spouse and/or with others more frequently report better or worse marital quality. The covariates included age, gender, education, whether the respondent was a panel member or a core network member, average number of drinks per week, month of survey, and previous-month reports on negative marital quality. Frequency of drinking with one’s spouse or with others was estimated in separate models.

For all models, continuous predictors and covariates at baseline were centered on the grand mean, whereas continuous time-variant predictors and covariates were centered on the participant-level mean (i.e., the individual’s mean across all time points). Month of survey was centered on the baseline. We assessed whether there were significant differences between the fit of the model at each step relative to the previous step by subtracting the -2*log likelihood estimations (Singer & Willett, 2003).

Results

Description

First, we calculated descriptive statistics to examine the frequency of drinking and drinking with social partners. A total of 24% of participants reported not drinking alcohol at all during the time they participated in the web study. On average, participants reported drinking 3.8 drinks per week (SD = 5.6, range: 0–30) and in 62% of the months they participated. In 38% of the months, participants reported not drinking, and in 8% of the months participants reported drinking alone. When considering the proportion of months in which participants reported drinking with different social partners, participants reported drinking with their spouse in 43% of the months and drinking with others in 36% of the months. A total of 25% of participants reported drinking with their spouse every month.

Next, we estimated multilevel models without predictors to estimate the degree of nonindependence in reports of frequency of drinking and negative marital quality between panel respondents and core network members. The intraclass correlation for frequency of drinking was .26, which represents the average correlation between any two network members. The intraclass correlation for negative marital quality was .24. Thus, network members were moderately correlated in their reports of drinking frequency and negative marital quality.

Do people drink with spouses more frequently than with other social partners?

As presented in Table 2, multilevel models predicting frequency of drinking as a function of social partner (coded as 1 = spouse, 0 = others) revealed that participants reported drinking more frequently with their spouse than with others. Estimated means showed that individuals reported drinking with their spouse between two times a month and two times a week (M = 2.71, SE = 0.06) and drinking with others between one and three times per month (M = 1.66, SE = 0.06). There was no gender difference in the frequency of drinking with one’s spouse or with others.

Table 2.

Multilevel model examining frequency of drinking by type of social partner

	Step 1		Step 2
Parameters	B	SE	B	SE
Intercept	1.606***	0.099	1.572***	0.102
Age	0.004	0.004	0.004	0.004
Gender	-0.067	0.091	-0.005	0.102
Education in years	0.024	0.022	0.024	0.023
Month of survey	0.005	0.006	0.005	0.006
Panel vs. core network	0.140	0.088	0.141	0.089
Drinking with spouse	1.043***	0.039	1.104***	0.058
Drinking With Spouse × Gender			-0.109	0.078
Variance estimates
Between network	0.309***	0.073	0.306***	0.073
Between person	0.199***	0.053	0.202***	0.054
Within person	0.510***	0.020	0.509***	0.020
-2*log likelihood	3,799.373		3,797.405
Δ -2*log likelihood	582.869***		1.968

Notes: N = 312 married or cohabiting adults. Estimates are presented from each step of the model. Step 1 change in -2 log likelihood represents change from the model with only covariates (-2*log likelihood = 4,382.242) and Step 2 change in -2 log likelihood represents change from Step 1. Continuous predictors and covariates at baseline were centered on the grand mean, whereas continuous time-variant predictors and covariates were centered on the group mean (i.e., the individual’s mean across all time points).

^***p < .001.

Is drinking with one’s spouse and with other social partners associated with negative marital quality?

First, we assessed whether drinking with one’s spouse or with others at all, in the same and previous month, was associated with within-person variations in negative marital quality. There were no within-person links between drinking with one’s spouse or with others and negative marital quality. There was a between-person association between the proportion of months individuals drank with their spouse and negative marital quality. Individuals who drank with their spouse on more months had lower negative marital quality on average (Table 3). The proportion of months they drank with others was not associated with negative marital quality.

Table 3.

Multilevel model examining negative marital quality as a function of drinking at all with spouse or with others

Variable	Drinking with spouse				Drinking with others
	Step 1		Step 2		Step 1		Step 2
	B	SE	B	SE	B	SE	B	SE
Intercept	2.388	0.121 ***	2.360	0.139 ***	2.241	0.119 ***	2.284	0.134 ***
Age	-0.006	0.004	-0.006	0.004	-0.005	0.004	-0.006	0.004
Gender	0.027	0.101	0.073	0.149	0.037	0.102	-0.035	0.147
Education in years	-0.039	0.024	-0.038	0.024	-0.050	0.024 *	-0.051	0.024 *
Month of survey	0.000	0.004	0.000	0.004	0.001	0.004	0.001	0.004
Panel vs. core network	-0.032	0.100	-0.034	0.100	-0.018	0.101	-0.015	0.101
Previous month negative marital quality	0.070	0.025 **	0.070	0.025 **	0.071	0.025 **	0.070	0.025 **
Drinks per week	0.001	0.006	0.002	0.006	0.001	0.006	0.001	0.006
Proportion of months drank	-0.003	0.002 *	-0.003	0.002	-0.001	0.002	-0.003	0.002
Drank in the same month	-0.029	0.048	0.010	0.069	-0.004	0.043	0.095	0.066
Drank in the previous month	0.033	0.045	-0.010	0.065	0.044	0.041	0.032	0.062
Proportion Months Drank × Gender			-0.001	0.003			0.003	0.003
Drank in the Same Month × Gender			-0.073	0.093			-0.171	0.086
Drank in Previous Month × Gender			0.083	0.089			0.017	0.082
Variance estimates
Between network	0.279	0.088	0.282	0.089	0.276	0.091	0.274	0.091
Between person	0.543	0.084	0.539	0.084	0.562	0.087	0.564	0.088
Within person	0.294	0.010	0.294	0.010	0.294	0.010	0.293	0.010
-2 log likelihood	3,913.951		3,912.229		3,919.880		3,915.405
Δ -2 log likelihood	7.17		1.723		1.24		4.475

Notes: N = 312 married or cohabiting adults. Estimates are presented from each step of the model. Step 1 change in -2LL (log likelihood) represents change from the model with only covariates (-2*LL = 3,921.117 and Step 2 change in -2LL represents change from Step 1). The model change from the covariate model to Step 1 for drinking with spouse is only marginal (p < .10) because of the nonsignificant main effects for drank with spouse in same and previous month. A Step 1 model with only percentage of months drank with spouse reveals a significant improvement with 1 df (change = 6.198, p < .05). Continuous predictors and covariates at baseline were centered on the grand mean, whereas continuous time-variant predictors and covariates were centered on the group mean (i.e., the individual’s mean across all time points).

^*p < .05;

^**p < .01;

^***p < .001.

Second, we assessed whether the frequency of drinking with one’s spouse and the frequency of drinking with others was associated with variations in negative marital quality on a monthly basis. As presented in Table 4, this model shows that the frequency of drinking with one’s spouse was not associated with negative marital quality. There were no between-person associations between average drinking frequency with one’s spouse or with others and negative marital quality. However, there was significant interaction between frequency of drinking with others in the previous month and gender when predicting negative marital quality. The analysis of the simple effects revealed that women who reported drinking with others more frequently had lower negative marital quality (B = -0.11, SE = 0.06, p < .10) in the next month, whereas men who drank more frequently with others had increased negative marital quality in the next month (B = 0.15, SE = 0.09, p < .10) (Figure 1). Although the simple slopes are not significantly different from zero, the interaction indicates a significant gender difference in the slopes.

Table 4.

Multilevel model examining negative marital quality as a function of frequency of drinking with spouse or with others

Variable	Drinking with spouse				Drinking with others
	Step 1		Step 2		Step 1		Step 2
	B	SE	B	SE	B	SE	B	SE
Intercept	2.090	0.284 ***	2.184	0.389 ***	2.337	0.303 ***	2.394	0.453 ***
Age	-0.006	0.006	-0.006	0.006	-0.006	0.006	-0.006	0.006
Gender	0.031	0.158	-0.128	0.489	0.029	0.169	-0.037	0.563
Education in years	-0.004	0.040	-0.005	0.040	-0.050	0.042	-0.048	0.043
Month of survey	-0.007	0.009	-0.007	0.009	-0.017	0.009	-0.019	0.009 *
Panel vs. core network	-0.071	0.161	-0.071	0.161	-0.217	0.163	-0.222	0.164
Previous month negative marital quality	-0.004	0.048	-0.002	0.049	0.067	0.059	0.077	0.058
Drinks per week	0.017	0.010	0.018	0.010	-0.017	0.009	-0.019	0.009 *
Mean frequency	0.016	0.088	-0.014	0.122	0.055	0.141	0.031	0.227
Same-month frequency	-0.038	0.045	-0.032	0.064	0.072	0.051	0.059	0.075
Previous month frequency	-0.064	0.045	-0.082	0.061	-0.015	0.052	0.150	0.086
Mean Frequency × Gender			0.054	0.158			0.036	0.290
Same-Month Frequency × Gender			-0.013	0.089			0.000	0.100
Previous Month Frequency × Gender			0.040	0.090			-0.256	0.107 *
Variance estimates
Between network	0.308	0.164	0.305	0.164	0.163	0.150	0.157	0.150
Between person	0.573	0.156	0.574	0.156	0.627	0.166	0.639	0.168
Within person	0.260	0.018	0.260	0.018	0.206	0.018	0.201	0.017
-2 log likelihood	1,139.123		1,138.756		786.073		780.361
Δ -2 log likelihood	2.497		0.367		2.450		5.712

Notes: N = 312 married or cohabiting adults. Step 1 change in -2*LL (log likelihood) represents change from model with only covariates (drinking with spouse: -2*LL = 1,141.620; drinking with others: -2*LL = 788.523). Step 2 change in -2LL represents change from Step 1. The model change from Step 1 to Step 2 for drinking frequency with others is only marginal (p < .10) because of the nonsignificant interaction between same-month frequency and gender. A Step 2 model with only lagged Drinking × Gender reveals a significant improvement with 1 df (change = 5.697, p < .05). Continuous predictors and covariates at baseline were centered on the grand mean, whereas continuous time-variant predictors and covariates were centered on the group mean (i.e., the individual’s mean across all time points).

^*p < .05;

^***p < .001.

Figure 1.

Negative marital quality as a function of drinking with others and gender

Thus, consistent with our prediction, drinking with one’s spouse was associated with better marital quality, but drinking status across months mattered more than the frequency of drinking. On the other hand, when examining frequency of drinking, drinking with others (and not drinking with one’s spouse) emerged as a significant predictor of negative marital quality in the opposite direction for men and for women. These findings are only partially consistent with our hypotheses.

Post hoc analyses

We tested a series of models to assess the stability of the findings. First, we included reports of drinking with one’s spouse and drinking with others in the same models. The findings from the main analysis remained for the models examining both drinking status and drinking frequency with one’s spouse and with other social partners, confirming their stability.

We next assessed whether there were interactions with age across all of the models. There was a significant interaction between age and relationship type predicting frequency of drinking. Findings revealed that there was a larger difference between frequency of drinking with others and with one’s spouse for older respondents than for younger respondents. The estimated means of drinking frequency by relationship type for older adults were 1.64 with others and 2.83 with one’s spouse, whereas the estimated means for younger adults were 1.69 with others and 2.58 with one’s spouse. There were no interactions with age when predicting marital quality.

We also evaluated the interaction between drinking partners to assess whether the frequency of drinking with one’s spouse interacted with the frequency of drinking with others such that particular patterns of drinking are better or worse for marital quality. There were no significant interactions between drinking frequency with one’s spouse and drinking frequency with others when predicting negative marital quality.

Last, we examined whether the findings remained after controlling for the length of the marital relationship. The same pattern of findings emerged.

Discussion

Using a monthly research design, this study revealed that individuals ranging in age from young to older adulthood drink with their spouse more often than with other social partners. Furthermore, drinking with one’s spouse is associated with less negative marital quality, whereas the effects on marriage of drinking with others vary by gender. This study extends the previous literature by examining with whom and how often individuals drink (i.e., with their spouse as well as other social partners). The findings provide a window into the nuances of drinking partners and their implications for marital quality across adulthood.

Frequency of drinking with one’s spouse versus with others

Both men and women drank more frequently with their spouse than with others. This finding is consistent with the concept of the drinking partnership (Roberts & Leonard, 1998). Spouses tend to view drinking together as an intimacy booster as well as a shared leisure activity. Inconsistent with previous literature and our hypothesis, there were no gender differences in the frequency of drinking with one’s spouse (Graham & Braun, 1999). This may be because the present study examined any drinking rather than heavy drinking, which is more common among men. All of these findings uniquely contribute to the literature by showing that, among individuals ranging in age from young adulthood to old age, spouses appear to be the preferred drinking partner for both men and women.

Drinking with one’s spouse versus with others and marital quality

Drinking with one’s spouse across a greater proportion of months was associated with lower negative marital quality. Thus, consistent with the concept of the drinking partnership, it appears that drinking with one’s spouse is associated with less-negative marriages. This finding is in line with research on young married couples, which has shown that drinking together is associated with greater marital adjustment and intimacy (Roberts & Leonard, 1998). Drinking with one’s spouse may increase intimacy, because it includes time together to exchange thoughts, ideas, and daily experiences that improve bonding. Interestingly, the association between the proportion of months of drinking with one’s spouse and negative marital quality did not vary by gender. This is inconsistent with the previous research demonstrating that women’s perceptions of marital quality are more affected by drinking with one’s spouse than men’s (Homish & Leonard, 2007; Roberts & Linney, 2000). However, the findings in the present study apply to drinking any amount of alcohol rather than heavy drinking. It therefore seems that both men and women may equally benefit from drinking with their spouse across the life span.

In contrast, when examining the frequency of drinking with spouses and with others, only frequency of drinking with others was associated with negative marital quality. Of note, the effects varied by gender such that drinking with others tended to be beneficial for women’s marriages and harmful for men’s marriages. This finding is inconsistent with previous literature indicating that drinking with others is associated with poorer partner relationship quality for both women and men (Homish & Leonard, 2008; Levitt & Cooper, 2010).

The present finding may be the result of this study’s definition of drinking as drinking any amount rather than heavy drinking. When engaging in light to moderate social drinking, for example, women may benefit from time away from their spouse to socialize with friends and family. Women are also more likely than men to drink at home and to include their partner when drinking (Erol & Karpyak, 2015). As a consequence, drinking with others may enhance the couple’s social environment, in turn contributing to less-negative evaluations of marriage among wives. For men, social drinking more often involves going out without their partner (Erol & Karpyak, 2015). Hence, when husbands drink with others, this may be a source of marital tension and resentment. It is possible that men’s drinking with others leads to heavier alcohol use that is problematic for marriage, yet the findings remained after we controlled for the amount of alcohol consumed. Taken together, the results imply that drinking with one’s spouse, in terms of drinking status overall, is associated with better marriages; but the frequency of drinking with others (and not with spouses) may be consequential for marital quality in ways that differ between wives and husbands.

It is important to note, however, that the association between negative marital quality and drinking with one’s spouse is likely bidirectional. Specifically, individuals who report better marriage quality (i.e., less-negative marriages) may be more likely to drink with their spouses, perhaps in part because spouses who feel less negative about their relationship want to spend more leisure time together. Nevertheless, the frequency of drinking with others predicted perceptions of negative marital quality in the next month among both wives and husbands. This finding implies that social drinking patterns may negatively affect subsequent marital interactions, underscoring the need for future studies to disentangle the nature and direction of potential causal pathways.

Limitations and future directions

There are several limitations to this study that should be addressed in future research. First, the sample was a relatively small and select group of individuals who completed web surveys from the baseline Wave 3 study. In particular, respondents in the sample were older, more educated, more likely to be female, and more likely to be core network members than the baseline Wave 3 study participants. The models in this study controlled for these participant characteristics, but future research should examine whether the findings are similar in larger and more diverse groups. Indeed, even though the models revealed that the findings did not vary by age group, further research should explore these associations among age groups that are more evenly distributed than the current study. It is possible, for example, that the findings varied by age group, but there was not sufficient power in the younger group to find effects.

In addition, this study focused on individual reports of one’s own drinking and not spouses’ reports of drinking. Future research should include marital dyads (i.e., both members of the couple). Still, it is important to note that spouses are accurate reporters of one another’s drinking behavior (Graham & Braun, 1999). This study also did not include measures of alcohol use disorders or problematic drinking. Alcohol use disorders are problematic for marriages, and future research should examine the frequency of drinking with one’s spouse among individuals with alcohol use disorders (Cranford et al., 2011). Furthermore, it is not clear whether individuals were drinking with multiple social partners at the same time (e.g., with one’s spouse and friends). Future studies should consider these issues in greater detail, including the use of daily diary designs that allow for more nuanced accounts of daily social interactions. It is also not clear whether the findings regarding drinking with one’s spouse and marital quality reflect other aspects of the spousal relationship quality (Akiyama et al., 2003), such as the amount or quality of time spent together.

In addition, the individuals who were cohabiting were together for a shorter time than those who were married, and so future research should examine whether the findings vary by marital status and length of relationship. However, the models revealed the same findings after we controlled for the length of the relationships. Despite these limitations, the present findings underscore the value of considering social drinking patterns and their implications for marital quality over time.

In sum, this study demonstrates that individuals drink with their spouses more often than with others, and that drinking with one’s spouse appears to be beneficial for marital quality, whereas the effects of drinking with others vary by gender. Additional research is needed to understand the particular mechanisms accounting for these effects as well as their long-term implications for health throughout the adult life span.

Acknowledgment

The authors thank Angela Turkelson for her assistance with the analysis.