Sociation Today

An Exploration of the Effects of
Work/Family Roles, Gender-Role Attitudes and Demographics on
Normative Drinking Rates

by

Susan Bullers

University of North Carolina Wilmington

Background

The negative impacts of alcohol abuse on society are well documented. While the majority of studies in the alcohol literature focus on problem drinking, research concerning broad societal trends in normative drinking patterns can help inform efforts to predict and target populations who are at an increased risk for drinking problems. This study explores the effects of gender roles, attitudes and demographic factors on normative drinking rates in the US to more clearly understand how social structural factors affect alcohol consumption rates.

Consistent findings in the literature show that men drink more, and more often than women (Dawson, Grant, Chou and Pickering, 1995; Fillmore et al., 1997; NIAAA 2009; Wilsnack, Wilsnack, Kristjanson, Vogeltanz-Holm, & Gmel, 2009). Although this is true for virtually all age and ethnic groups, the gap between men's and women's rates varies by age and ethnicity and has changed over time. Current trends show that whites have the highest drinking rates, followed by Hispanics, with African Americans reporting the lowest drinking rates. The gender gap is widest among Hispanics, followed by African Americans and narrowest among whites. The gender gap is also wider among older cohorts and appears to be narrowing over time. Approaches used to explain these gender differences in alcohol use rates include cultural explanations, which focus on gendered attitudes and stigma, and structural approaches, which focus on gender roles.

Cultural approaches offer that alcohol consumption itself is deeply embedded in a gendered belief system. Heavy drinking among women has been stigmatized and is often perceived to be associated with sexual promiscuity, lack of self-worth, negligence of family responsibilities, and generally inappropriate gender-role behavior (Bergmark, 2004; Fillmore et al., 1997; Huselid &Cooper, 1991; Morgan, 1987; Robbins & Martin, 1993; Room, 1996; Trotter, 1985; Waterson, 2000). On the other hand, heavy drinking has been viewed as an acceptable way for men to have fun and gain social acceptance (Ames & Rebhun, 1996; Gilbert, 1985). Among young men especially, heavy drinking can serve as a culturally recognized symbol of masculinity (Peralta, 2007; Peralta, Steele, Nofziger, & Rickles, 2010). These traditional attitudes are diminishing somewhat, especially among younger cohorts and it is assumed that increasingly liberal gender-role attitudes are responsible for increasing drinking rates among women. On the other hand, the impact of changing gender role attitudes on men's drinking has not received as much attention. In fact, drinking rates among younger men have been decreasing. Perhaps increasingly egalitarian views on drinking behavior will weaken the cultural association between drinking and masculinity, effectively diminishing the use of alcohol to bolster masculine identity.

     Although egalitarian gender role attitudes are associated with increased drinking among women, some researchers argue that social structural factors play a more fundamental part in this relationship. The structural approach to gender and drinking patterns focuses on gendered roles themselves and suggests that increasingly egalitarian work/family roles may lead to similar drinking rates between men and women.

     Early stress-related research assumed that employment roles were inherently more stressful than homemaking roles and that increased employment among women would be accompanied by increased stress and subsequent drinking (Fillmore et al., 1997). Alternative explanations pointed out that women's' family roles were more likely to be a source of stress then employment roles (Hammer & Valgum, 1989; Mirowsky & Ross, 1989; Wilsnack & Cheloha, 1987; Wilsnack & Wilsnack, 1992) and that home-making roles required constant attention and responsibilities that limited opportunities to drink and socialize among adults (Ames & Rebhun, 1996; Caetano & Clark, 1999; Cunningham, 2007; Waterson, 2000; Wilsnak & Cheloha, 1987).

     The role literature has also found that marital status affects drinking; married men and women drink less than their single, divorced, or widowed counterparts, although this effect is stronger for men than for women (Fillmore et al., 1997; Leonard & Rothbard, 1999). Again, stress-related explanations argue that non-married individuals drink to cope with the hardship and stresses of the single status. Alternatives to the stress explanations argue that heavy drinkers select out of marriage and that spouses provide social support and encourage healthy behavior; all resulting in lower drinking rates among married individuals. Parental status is also associated with lower drinking rates, although more so for women than men. This is likely due to gender differences in parental roles where women typically spend more time on housework and child-care than men do (Bianchi, Milkie, Sayer & Robinson, 2000). However, most studies in this literature do not include measures of specific family roles that may be implicated in these gender differences.

The role of college student has long been associated with very high drinking rates and a large body of research has explored various effects of psychological and social influences on drinking rates in this population (Ham & Hope 2003). From a structural perspective, the college student role offers a plethora of drinking opportunities. College life typically entails a flexible schedule, optional class attendance, few if any family responsibilities, an abundance of accommodating and convenient drinking establishments and large social networks. These structural factors likely play a role in determining drinking patterns for this group and combined with increases in women's college enrollment, may impact aggregate drinking rates among women.

Men have traditionally dominated roles that accommodate drinking opportunities (employment and college) while women's traditional roles (family and child-care) limit opportunities to drink. Increases in women's' participation in employment and college roles may lead to increases in their drinking rates through the increased opportunities provided by these roles. However, the effects of changing gender roles on men's drinking have been largely ignored. In fact, aggregate trends suggests that the gender gap convergence is due to decreases in men's drinking rather than increases in women's drinking (McPherson, Casswell, and Pledger, 2004; NIAAA, 1994; NIAAA, 2009; Wallace et al., 2003; Wilsnack, Wilsnack, Kristjanson, Vogeltanz-Holm, & Gmel, 2009). The role literature tends to frame the influence of changing gender roles as process in which women's employment will result in assimilation to male drinking patterns. However, increases in men's participation in family roles may also be associated with men's assimilation to women's drinking patterns. Increasingly gender-segregated work and family spheres will likely result in complex social influences in which norms will be determined by the composition of these new social networks.

Although the role and attitude factors discussed above are likely associated with men's and women's drinking patterns, broad trends in the US population may also contribute to aggregate drinking rates. One factor that is strongly associated with decreased drinking rates is age and the US population is ageing. This trend could reduce aggregate drinking rates (US Bureau of the Census, 2011b). Similarly, drinking rates are lower in the US south than other regions and the US south is growing rapidly compared to other regions of the country (US Bureau of the Census, 2011c). Another factor that may be suppressing aggregate drinking rates is ethnicity. Ethnic differences in US drinking rates have a long complex history but consistent trends show that whites have the highest rates, followed by Hispanics and then African Americans who report the lowest rates among these three groups (Bales, 1946; Christmon, 1995; Gilbert, 1991; Gordon, 1985; Herd, 1985; Rebach, 1992; Collins & McNair, 2004). Relative increases in minority population sectors with low drinking rates could contribute to the overall suppression of aggregate drinking rates (US Bureau of the Census, 2011a).

Another factor shown to affect drinking rates is "objector" religious affiliation. Objector religions include those religions or denominations that explicitly prohibit alcohol use. These religious traditions are associated with very low drinking rates and membership varies substantially by ethnicity. African Americans are most likely to belong to religious denominations that prohibit alcohol use, followed by whites. Hispanics in the US are more likely to be Catholic; a denomination not typically associated with drinking prohibition. Minority ethnicity and objector religious identification are both associated with low drinking rates and some of the effects of ethnicity on drinking and found in the literature may be due to these religious differences. Similarly, population trends regarding growth among ethnic groups or objector religious affiliations may play a role in aggregate drinking rate trends.

Given the same drinking opportunities and expectations, some people become problem drinkers and others do not. The personal, economic, and social costs of alcohol abuse in the US are huge and a vast body of research has explored the complex relationships among the social, psychological, and physiological factors associated with problem drinking. Although the process and causal factors associated with problem drinking undoubtedly differ from the correlates of light or moderate drinking, it is likely that structural factors will be associated with problem drinking to some extent, simply as a matter of initial exposure and opportunity. The extent of differences between structural causes of drinking status and the structural causes of drinking amount is unknown.

In summary, changes in gender role attitudes and work/family roles have been associated with increases in women's drinking. However, the effect of these factors on men's drinking need to be examined as well. In addition, any effects of age, ethnicity, region, and religion on drinking are likely to effect aggregate drinking rates though the population growth or decline within these groups. This study will analyze a large representative sample to examine the effects of attitudes, roles, and demographics on drinking status and drinking quantity among men and women.

Methods

Data

This analysis uses data from Wave II of the US National Survey of Families and Households (NSFH) to explore the effects of demographic, work/family, gender-role attitude, religious, and geographic context variables on drinking status and drinking quantity. The NSFH is a nationally representative longitudinal panel (three Waves) survey conducted between 1987 and 2002 (a full description of data collection methods can be found in Sweet & Bumpass, 1996). Wave II data were collected in 1993 and consist of 10,007 respondents. For these analyses, data were weighted by a normalized total person weight for Wave II, which includes a total person weight from Wave I, a correction for tracing bias, a correction for non-response bias, and a post-stratification correction. Although Wave II of the NSFH is somewhat dated it includes a unique collection of role, attitude, demographic and drinking variables not available elsewhere. These data are ideal for exploring the simultaneous and independent effects of these variables in a nationally representative sample. The demographic effects regarding ethnicity, region, urban residence, and age, in particular, may help inform current trends regarding aggregate drinking rates. Although the distributions of these characteristics have changed over time, the associations among the variables should be relatively stable and applicable to current behavior. More recent population rates on key variables will be offered where appropriate.

Quantity and frequency drinking variables revealed that nearly half of the respondents did not drink at all in the past year and less than 4% typically drank four or more drinks, resulting in highly skewed distributions for these drinking variables. A dichotomous "Drinking Status" variable was constructed and coded "1" for any drinking in the past year or "0" for no drinking in the past year. A "Drinking Level" variable was also constructed, for drinkers only, which included aspects of quantity and frequency of alcohol consumed. This variable was a product of the number of drinking days per month and the number of drinks typically drunk per drinking occasion. The resulting value was divided by 30 to give mean drinks per day. This was then dummy coded into "light" drinkers (< .5 drinks per day) and "moderate/heavy" (.5 + drinks per day). Although there are many measurements and definitions of light drinking, this one is consistent with the CDC's definition of light drinking. There were 59 cases with missing values on the drinking variables; these were dropped from the sample.

Age was measured in years and gender was coded as male (1) or female (0). Marital status was coded as currently married (1) and not currently married (0), to capture the social roles of marriage. Education was coded as any college (1) or no college (0). This measure was chosen to reflect a college culture experience as well as a broad proxy for socioeconomic status. Similarly, the role or social experience of paid employment was measured with "work" coded as currently employed (1) and not currently employed (0). Presence of children in the household was re-coded from a variable measuring number of children under age 18 in the household. Over half the respondents had no children under 18, again resulting in a non-normal distribution, so this variable was coded dichotomously; any children under 18 currently living in the home (1) or none (0). The hours of household labor scale was constructed by summing the scores from 9 items asking about hours spent per week on preparing meals, washing dishes, cleaning house, outdoor tasks, shopping, washing/ironing, paying bills, auto maintenance, and driving.

Gender role attitudes were measured with a scale constructed from responses to five items concerning gendered family responsibilities; "It is much better for everyone if the man earns the main living and the woman takes care of the home and family," "It is all right for children under three years old to be cared for all day in a day care center," "Preschool children are likely to suffer if their mother is employed," "It is all right for mothers to work full-time when their youngest child is under age 5," and "Both the husband and wife should contribute to family income." Responses for these items ranged from strongly agree (1) to strongly disagree (5) and response codes were selectively reverse coded so that higher scores reflected more traditional attitudes. Item scores were summed and divided by 5; final scores ranged from 1 to 5.

The objector religion variable was constructed by re-coding an item asking for respondents' religious preference. There were 66 response categories for this item. Items were coded as objector religions (1) if doctrine prohibited (not just discouraged) alcohol use. All Evangelical, Muslim, and known fundamentalist responses were coded as objector religions as well as religions for which a formal statement on alcohol use could be located. Religions for which alcohol use beliefs could not be ascertained or where alcohol use was merely discouraged, were coded as non-objector (0). Metropolitan status was recoded to represent metropolitan (1) or non-metropolitan and adjacent (0). Region of Country was re-coded as south (1) or all others (0).

Ethnicity included three categories; African American, white, and any Hispanic. The African American and Hispanic categories were each re-coded into dichotomous variables for use in the regression analyses with "white" used as the reference category. All "other" ethnic groups and those with missing values on ethnicity were dropped from the analysis (n=133). The final sample size was 9,872.

Analysis

The main analyses were conducted using SPSS version 19. First, percentages (or means) for all drinking, role, attitude, and demographic variables by gender and ethnic group were calculated to describe the sample and the distribution of these characteristics (Table 1).

Table 1

Descriptives of Drinking Status, Drinking Level, Role, Demographic and Attitude by Variables by Ethnicity and Gender

		Men					Women		N
(% in race)	White	Blk	Hisp	Total	White	Blk	Hisp.	Total
Drinker	59.8	47.5	55.1	58.2	43.5	26.7	26.1	40.3	9,807
College	53.9	34.0	30.6	50.1	44.1	35.7	29.6	42.1	9,850
Work	75.6	70.6	76.1	75.1	57.0	57.6	56.0	57.0	9,872
Married	73.0	49.7	62.6	69.9	63.1	32.5	52.5	58.7	9,870
Kids at home	34.7	34.5	52.2	36.1	36.2	49.6	55.6	39.2	9,872
Mean hrs/wk housework	12.5	18.4	12.4	13.1	30.6	34.4	39.9	31.7	9,220
Mean age	47.6	44.6	42.1	46.8	49.5	45.6	43.5	48.6	9,869
South	31.6	55.4	32.2	34.1	32.4	54.9	31.9	35.0	9,843
Metro	79.2	87.9	97.6	81.5	78.2	88.4	97.6	80.8	9,841
Objector Rel.	24.4	69.0	12.7	27.9	27.2	73.0	11.9	31.3	9,802
Mean Trad. Attitude	3.14	2.84	3.24	3.12	3.02	2.74	3.04	2.99	9,388
>Lightdrink^a	29.3	21.2	21.6	27.7	18.3	13.8	15.5	17.6	4,786
Column %	39.0	4.8	3.8	47.6	42.5	6.1	3.9	52.4	(100)
N=	3852	472	372	4696	4195	599	383	5176	9872

^aAmong drinkers only

NSFH, Wave II, 1993, weighted by normalized Wave II total person weight (Wave I total weight, tracing correction, household correction, post-stratification correction). Gender attitude scale; higher values are more traditional

To examine the effects of the role, attitude and demographic variables on Drinking Status (drinker/non-drinker) and Drinking Amount (light/other) two sets of logistic regression models were run; one each for men and women. Logistic regression models are suited for analyses with dichotomous dependent variables and results are expressed as logged odds ratios that represent the increase or decrease in odds of obtaining the dependent variable value (1), given the values of the independent variables. Total model effects are expressed as a Chi square statistic that tests the increase in model fit from the addition of the independent variables. The first set of models regressed Drinking Status on ethnicity, college, employed, metro area, age, married, kids, household labor hours, objector religion, Southern region, and gender-role attitudes for men and women separately (Table 2). Next, Drinking Amount was regressed on those same role, attitude, and demographic variables for men and for women (Table 3).

Table 2

Logistic Regression of Drinking Status on Explanatory Variables, by Gender

Men Women

	B	SE	exp(B)	B	SE	exp(B)
Afric. Amer.	-.069	.063	.933	-.210	.060	.810***
Hispanic	-.239	.133	.787	-1.043	.140	.352***

College	.363	.072	1.438***	.342	.069	1.408***
Employed	.435	.096	1.545***	.177	.076	1.194*
Married	.019	.089	1.019	.103	.072	1.108
Kids at home	-.190	.086	.827*	-.365	.084	.694***
HH Labor hrs	-.007	.003	.993	-.005	.002	.995**

Age	-.017	.003	.983***	-.034	.003	.967***
South	-.194	.075	.824**	-.246	.072	.782**
Metro Area	.175	.090	1.191	.380	.087	1.462***
Religion	-1.007	.082	.365***	-.938	.081	.391***
Trad. Attitude	-.316	.051	.729***	-.282	.048	.754***
(Constant)	2.108	.247		2.287	.247
Model Chi Square=			548.97***			811.30***
Nagelkerke R²=	.168			.218
N=	3372			5378

*<.05, **<.01, ***<.001

NSFH, Wave II, 1993, weighted by normalized Wave II total person weight (Wave I total weight, tracing correction, household correction, post-stratification correction).

Drinking status; drink in past year (1), no drinking in past year (0)

Table 3

Logistic Regression of Drinking Level on Explanatory Variables, by Gender

Men Women

	B	SE	exp(B)	B	SE	exp(B)
Afric. Amer.	-.285	.087	.752***	-.223	.121	.800
Hispanic	-.204	.163	.816	-.056	.249	.946

College	-.271	.087	.762***	.056	.110	1.058
Employed	.037	.130	1.038	-.176	.127	.838
Married	-.452	.106	.636***	-.175	.112	.840
Kids at home	.090	.101	1.094	-.536	.127	.585***
HH Labor hrs	.001	.004	1.001	.004	.003	1.004

Age	.001	.004	1.001	-.004	.004	.996
South	.258	.093	1.294***	.077	.119	1.080
Metro Area	.178	.114	1.195	.336	.157	1.399*
Religion	-.204	.119	.815	-.399	.159	.671*
Trad. Attitudes	-.237	.063	.798***	.012	.075	.988
(Constant)	.823	.307		-.680	.380
Model Chi Square=			68.67***			52.59***
Nagelkerke R²=	.034			.039
N=	2026			2193

*<.05, **<.01, ***<.001

NSFH, Wave II, 1993, weighed by normalized Wave II total person weight (Wave I total weight, tracing correction, household correction, and post-stratification correction).

Drinking level; light (0), mod/heavy (1)

For both descriptive and logistic regression analyses, the data were weighted to correct for tracing, stratification, and response biases in the original data collection at Wave I and in Wave II. These weights were normalized to retain the sample size. There is debate about the utility of weights for regression analyses using the NSFH (Johnson & Elliott, 1998; Fox, Benson, Demaris, & Van Wyk, 2002). However, the descriptive analysis in the current study explores the population distribution of several variables and relationships, so sampling weights were used in these analyses. These weights were also retained in the regression analyses because variables in the models (especially ethnicity) were likely to be affected by the over-sampling.

Design effects concerning possible biases from cluster sampling homogeneity were explored with a post-hoc analysis of the logistic regression models using the NSFH sampling unit information and Stata data analysis software. These design effects do not alter coefficient estimates but can increase (or decrease) standard errors, possibly overestimating statistical significance. Again, the utility of these design effect corrections are debated; it is argued whether the estimate benefits outweigh the compromises in power. The regression models used in this study include the main NSFH cluster-generating variables (region, metropolitan area, ethnicity) so additional cluster-design corrections were not deemed necessary.

Results

The descriptive results (Table 1) generally reflect previous research regarding drinking among US gender and ethnic groups. For both Drinking Status and Drinking Level, whites have the highest rates, followed by Hispanics and then African Americans. Men drank more than women in each case but the gender gap was widest among Hispanics, followed by African Americans, and narrowest among whites. Also reflecting previous research, Hispanics had the lowest rates of objector religious identity (Hispanics were predominantly catholic), the highest rates of traditional work/family roles and the most traditional gender role attitudes. Consistent with previous research, the descriptive results also show that older age, female gender, objector religious identification, and southern region were associated with decreased drinking, and college and employment were associated with higher drinking rates.

The logistic regression analyses that regressed Drinking Status on attitude, role, and demographic variables showed that for both men and women, college and employment were positively associated with drinking, while age, objector religion, southern residence, children at home, housework hours and traditional gender-role attitudes were associated with abstinence for both men and women (table 2). For men there were no independent effects of ethnicity on drinking status after controlling for other variables in the model. For women, both African American and Hispanic ethnicity were positively associated with abstinence. Metro residence was positively associated with drinking for women only. The regression model for women explained more variation in Drinking Status than did the men's model.

The logistic regression analysis of Drinking Level (Table 3) included only drinkers and showed that, for men, African American ethnicity, college, marriage, and traditional gender-role attitudes were associated with lower drinking levels. Among these drinking men, Southern residence was associated with higher drinking levels. For women drinkers, children in the household and objector religion were associated with lower drinking levels and metro residence was associated with higher drinking levels.

Discussion

The effects of housework hours on Drinking Status were significant and similar for men and women. However, the difference in mean housework hours between men and women was substantial; 13.1 hrs/wk for men vs. 31.7 for women (Table 1). These data are from 1993 and current population trends show that men's housework contribution and family role participation has increased steadily (Cunningham, 2007). If this relationship between housework hours and drinking holds, aggregate decreases men's drinking could be due, in part, to increases in their housework hours. From a role perspective, it may be that housework hours themselves limit opportunities to drink or they may reflect a broader sense of responsibility for family roles that is not compatible with drinking. Although housework was associated with Drinking Status, it was not associated with Drinking Level for either men or women. This pattern suggests that the housework variable taps a sense of responsibility rather than a time constraint.

The parental role was associated with increased abstinence for both men and women but the effects were substantially stronger for women. This may reflect the persistently gendered responsibilities of family roles in which women are more likely to be the primary caretaker (Bianchi, Milkie, Sayer & Robinson, 2000). Child care is the family role most closely linked to a social structural explanation for gender differences in drinking. The time, venue, and social constraints associated with full-time child-care can be functionally, or apparently, incompatible with drinking. At any rate there is a heavy stigma attached to drinking mothers (Waterson, 2000).

These parental status findings are limited in that they did not use a measure of child care hours or specific responsibilities but rather used a measure indicating whether there were children under 18 in the household. Although data concerning hours of child-care and number of children would provide more detail, it may also be the responsibility associated with the role of primary caregiver that affects this relationship between parenthood and drinking. To further explore the effects of parenthood on drinking, future research should explore the distinction between own- and step-children in the household and should include own children living in another household. Children are more likely to live in their mother's household and there may be gender differences in felt responsibility for step-children living in one's own home. The strength and gender discrepancy in these effects, along with recent trends regarding men's family role participation, warrant further exploration of the effects of changing work-family roles on men's drinking behavior. Among drinkers, the effect of children in the home had a substantial negative effect on amount of drinking for women only. Again, this may reflect the primary caregiver role that women are more likely to occupy.

Marriage had no effect on Drinking Status for either men or women. It may be that other variables in the model accounted for factors typically associated with abstinence among married individuals, or perhaps couples self-select into similar Drinking Status pairs. Although marriage had no effect on abstinence it was associated with substantially lower drinking levels among drinking men. As discussed in previous research, marriage may offer a protective effect through social support and active discouragement of harmful behaviors. However, these protective effects may themselves entail gendered behaviors if women are expected to be responsible for their husbands' drinking behavior. Or, husbands may simply assimilate to their wives drinking patterns.

Employment was positively associated with Drinking Status for both men and women but the association was substantially stronger for men. Although the diverse social networks, drinking norms, personal income and drinking opportunities provided by employment may encourage or enable drinking, the gender differences here suggest that this relationship is weaker for women than for men. Possible intervening factors include women's additional family responsibilities, occupational culture, and gender segregation of the workplace. At any rate, this pattern does not offer strong support for a stress explanation for women's employment and drinking. Stress effect drinking is likely to be heavier drinking and in this analysis, employment was associated with lower levels of drinking among drinking women. These findings suggest an increased likelihood of drinking with women's employment but also an increased likelihood of light drinking rather than moderate or heavy drinking, among employed drinkers.

    College experience was associated with an increased likelihood of drinking for both men and women. The College experience provides the environment, social opportunities and diverse network exposure that encourages drinking and for many, it is a time of few responsibilities that would constrain drinking. However, in this analysis, the college variable reflects a past role experience (for most respondents) that would most accurately reflect exposure to the college drinking culture. Although this past college experience is associated with increased likelihood of current drinking, among men who do drink, it is associated with lower levels of drinking.   This may suggest a protective effect of past heavy drinking on current drinking, or it may simply reflect socioeconomic status. However, most research shows that drinking is positively associated with socioeconomic status. The implications of this finding are fairly substantial. These long-term effects of the college experience on drinking should be investigated further.

     This quantitative exploration of ethnicity does not delve into the large body of literature on the cultural, political, and historical aspects of ethnicity, religion and drinking (Bales, 1946; Christmon, 1995; Gilbert, 1991; Gordon, 1985; Herd, 1985; Rebach, 1992; Collins & McNair, 2004). However, several researchers have suggested that there is an over-reliance on cultural explanations for ethnic differences in drinking patterns that can otherwise be explained by social structural factors (Darrow, Russell, Cooper, Mudar, & Frone, 1992; Gilbert & Collins, 1997; Guttmann, 1999; Hunt & Barker, 2001; Nyaronga, Greenfield & McDaniel, 2009; Nielsen, 2000). In this analysis, the lack of significant effects of ethnicity on Drinking Status among men, and drinking level among women, suggests that other variables in that model account for all of the ethnic differences in those models. However, ethnicity remains a strong predictor of women's abstinence, and Drinking Level among African American men. The effects of Hispanic ethnicity on women's drinking were among the strongest in that model. These results suggest a complex interaction between gender and ethnicity that is not explained by simple ethnicity effects.

    Objector religion, Southern region, and traditional gender role attitudes comprise a group of variables generally associated with restrictive vs. liberal attitudes regarding both drinking and gender roles. All of these variables were significantly associated with increased abstinence for both men and women. Objector religion was one of the strongest predictors of abstinence in the models. It also appears that these variables represent independent effects which vary substantially by ethnicity. The descriptive results show that African Americans have the highest rates of southern residence and objector religious membership, yet the least traditional gender role attitudes. Hispanics have the most traditional gender role attitudes yet the lowest rates of objector religious affiliation. The ethnic variability and independent effects of these variables speak to their importance in epidemiological research on drinking. It may be that some of the effects of ethnicity on drinking found in the literature are due to differences in these demographic and attitude variables.

The attitude-related variables did not have as strong of an effect on Drinking Level as they did on Drinking Status. Objector religion had a weak but significant negative association with drinking level among women, and traditional gender role attitudes reduced men's drinking substantially. Southern residence, although strongly associated with increased abstinence for both men and women, was associated with increased drinking among men who did drink. A common explanation for this type of effect is that the lack of a model of for light normative drinking in the formative years will lead to heavy and problem drinking among those who do drink. However, this effect was not observed in any of the other high-abstinence groups such as ethnic minorities or objector religious affiliates. African American men who drank were more likely than white men to be light drinkers. Hispanic women, the group least likely to drink, and African American women, did not differ from white women in the amount they drank, if they did drink.

Finally, many of the factors found to affect Drinking Status and Drinking Level in these analyses may account, at least in part, for gendered trends in aggregate drinking rates. Current US demographic trends show increased population shares among the groups with lower drinking rates in this analysis; the percent of the US population in the 45-64 age range increased 31.5% between 2000 and 2010, whereas the 18-44 age group increased only 0.6% (US Bureau of the Census, 2011b), the Hispanic population grew 43% between 2000 and 2010, compared to 10% for the total US population (US Bureau of the Census 2011a), the US south is experiencing disproportionately rapid population growth (US Bureau of the Census, 2011c) and there is relative stability in objector religious denominations as opposed to declines in other denominations (Kosmin & Keysar, 2009). Although these factors are rarely addressed in studies concerned with individual drinking problem risk factors, they could have a large impact on aggregate drinking rates through population growth trends in the high-abstinence groups.

Conclusions

The effects of objector religious affiliation on drinking status were substantial for both men and women and the rate of participation in objector denominations was relatively high (averaging over 25%) and varied widely by ethnicity. These findings confirm the need to consider this factor in drinking research analyses especially those looking at ethnic differences. Gender role attitudes also offered additional, independent effects on drinking status for both men and women. Although these variables were clearly associated with drinking status, their effects on drinking quantity among drinkers were less clear and varied by gender; traditional gender role attitudes had a strong negative effect on drinking quantity among drinking men.

Employment, any college, housework and children were all associated with drinking status for men and women, although rates of participation in these roles varied along traditional gender role lines. The research regarding women's drinking is often framed as an assimilation affect in which women are assimilating to men's drinking patterns as they move into traditionally male employment patterns. Findings here suggest that the influence of changing roles on drinking patterns may be reciprocal. The family role constraints that have traditionally limited women's drinking may limit men's drinking as they increase their participation in family roles (although the distinction between family role participation and primary caretaker may remain an important distinction). Increasingly gender-segregated work and family spheres are likely to produce new drinking norms influenced by the gender composition of these changing networks.

Drinking rates regarding gender, age, ethnicity, and region found in the literature were confirmed here. Coupled with differential population growth patterns among these groups, current decreases in aggregate drinking rates may be due to demographic trends that correspond to group drinking rates. In order to avoid an ecological fallacy these relationships need to be explored with current representative longitudinal data that include adequate drinking measures. The effects of ethnicity appeared to be accounted for by other variables in two of the models; men's drinking status and women's drinking quantity, but there also appear to be complex relationships among ethnicity, gender, objector religion and gender-role attitudes that suggest an intersectionality not modeled here. One critique of ethnicity research which this study was unable to surmount was the use of broad "ethnic" categories to define vastly heterogeneous groups. The "Hispanic" category in the NSFH data contains individuals of diverse national heritage and cultures. Retaining these distinctions would leave small ethnic group sample sizes that would be difficult to interpret in this context. A qualitative exploration might yield helpful insights regarding ethnicity, gender, and drinking.

Finally, one result here was inconsistent with the literature on drinking patterns; among drinkers, men who had attended college were more likely to be light drinkers than those who had not. The reasons for this are unclear but if this effect is replicated, it may be interpreted as a protection effect of past heavy drinking. This effect does need further exploration.

References

Ames, G., & Rebhun, L. (1996). "Women, Alcohol and Work: Interactions of Gender, Ethnicity, and Occupational Culture." Social Science and Medicine 43(11), 1649-1663.

Bales, R.F. (1946). "Cultural Differences in Rates of Alcoholism." Quarterly Journal of Studies on Alcohol 6:480-499.

Bergmark, K.H. (2004). "Gender Roles, Family, and Drinking: Women at the Crossroad of Drinking Cultures." Journal of Family History 29 (3), 293-307.

Bianchi, S.M., Milkie, M., Sayer L.C., & Robinson J.P. (2000). "Is Anyone Doing the Housework?; Trends in the Gender Division of Household Labor." Social Forces 79(1), 191-228.

Caetano, R., & Clark C.L. (1999). "Trends in Alcohol Consumption Patterns among Whites, Blacks and Hispanics: 1984 and 1995." Drug and Alcohol Dependence 54(1), 45-56.

Caetano, R. & Clark C.L. (1998). "Trends in Alcohol Consumption Patterns among Whites, Blacks, and Hispanics: 1984-1995." Journal on Studies of Alcohol 659-668.

Christmon, K. (1995). "Historical Overview of Alcohol in the African American Community." Journal of Black Studies 25(3), 318-330.

Collins, R.L., & McNair, L.D. (2004). "Minority Women and Alcohol Use." Retrieved from the Legacy Group on August 30, 2004.

Coltrane, S. (2000). "Research on Household Labor: Modeling and Measuring the Social Embeddedness of Routine Household Tasks." Journal of Marriage and the Family 62(4), 1208-1233.

Cunningham, M., (2007). "Influences of Women's Employment on the Gendered Division of Household Labor over the Life Course: Evidence from a 31-Year Panel Study." Journal of Family Issues 28(3), 422-444.

Dawson, D.A., Grant, B.F., Chou, S.P., & Pickering, R.P., (1995). "Subgroup Variation in US Drinking Patterns: Results of the 1992 National Longitudinal Alcohol Epidemiologic Study." Journal of Substance Abuse 7, 331-344.

Darrow, S.L., Russel, M., Cooper, M.L., Mudar, P. & Frone, M.R. (1992). "Sociodemographic Correlates of Alcohol Consumption among African-American and White Women." Women and Health 18(4), 35-51.

Fillmore, K.M., Golding, J. M., Leino, E.V., Motoyoshi, M., Shoemaker, C., Terry, H., Ager, C.R. & Ferrer, H.P. (1997). "Patterns and Trends in Women's and Men's Drinking." In Wilsnack, R.W. & Wilsnack S.C. (Eds.) Gender and Alcohol: Individual and Social Perspectives. (pp. 21-47) New Brunswick, NJ: Rutgers Center of Alcohol Studies.

Fox, G. L., Benson, M.L., Demaris, A.A., & Van Wyk, J., (2002). "Economic Distress and Intimate Violence: Testing Family Stress and Resource Theories." Journal of Marriage and Family 64, 793-807.

Gilbert, M. J. (1985). "Mexican Americans in California: Intracultural Variation in Attitudes and Behavior Related to Alcohol." In Bennett, L.A. and Ames, M.A. (Eds.) The American Experience with Alcohol: Contrasting Cultural Perspectives. (pp. 255-277) New York, NY: Plenum Press.

Gilbert, M.J. (1991). "Acculturation and Changes in Drinking Patterns among Mexican-American Women." Alcohol Health & Research World 15(3), 234-241.

Gilbert, M.J. & Collins R.L. (1997). "Ethnic Variation in Women's and Men's Drinking." In Wilsnack, R.W. and Wilsnack S.C. (Eds.) Gender and Alcohol: Individual and Social Perspectives. (pp. 357-377) New Brunswick, NJ: Rutgers Center of Alcohol Studies.

Gordon, A.J. (1985). "Alcohol and Hispanics in the Northeast A Study of Cultural Variability and Adaptation." In Bennett, L.A. and Ames, M.A. (Eds.) The American Experience with Alcohol: Contrasting Cultural Perspectives. (pp. 297-313) New York, NY: Plenum Press.

Gutmann, M.C. (1999). "Ethnicity, Alcohol, and Acculturation." Social Science and Medicine 48, 173-184.

Ham, L.S., & Hope, D.A. (2003). "College Students and Problematic Drinking: A Review of the Literature." Clinical Psychology Review 23, 719-759.

Hammer, T. & Vaglum, P. (1989) "The Increase in Alcohol Consumption among Women: a Phenomenon Related to Accessibility or Stress?; A General Population Study." British Journal of Addiction 84: 67-775

Herd, D. (1985). "Ambiguity in Black Drinking Norms: An Ethnohistorical Interpretation." In Bennett, L.A. and Ames, M.A. (Eds.) The American Experience with Alcohol: Contrasting Cultural Perspectives. (pp. 149-170) New York, NY: Plenum Press.

Hunt, G. & Barker, J.C. (2001). "Socio-cultural Anthropology and Alcohol Research: Towards a Unified Theory." Social Science & Medicine 53, 165-188.

Huselid, R.F. & Cooper, M.L. (1991). "Gender Roles as Mediators of Sex Differences in Adolescent Alcohol Use and Abuse." Paper presented at the Annual Meeting of the American Psychological Association, San Francisco, CA.

Johnson, D. & Elliot, L. (1998). "Sampling Design Effects: Do They Affect the Analyses of Data from the National Survey of Families and Households?" Journal of Marriage and Family 60, 993-1001.

Kosmin, B.A. & Keysar, A. (2009). "American Religious Identification Survey (ARIS) 2008."
http://b27.cc.trincoll.edu/weblogs/AmericanReligionSurvey-ARIS/
reports/ARIS_Report_2008.pdf. Hartford, Connecticut, USA: Trinity College. Retrieved 5 August 2012.

Leonard, K.E. & Rothbard, J.C. (1999). "Alcohol and the Marriage Effect." Journal of Studies on Alcohol 60, 139-146.

Mirowsky, J. and Ross, C.E. (1989). Social Causes of Psychological Distress. New York, NY: Aldine de Gruyter.

McPherson, M., Casswell, S. & Pledger, M. (2004). "Gender Convergence in Alcohol Consumption and Related Problems: Issues and Outcomes from Comparisons of New Zealand Survey Data." Addiction 99, 738-748.

Morgan, P. (1987). "Women and Alcohol: The Disinhibition Rhetoric in an Analysis of Domination." Journal of Psychoactive Drugs 19(2), 129-133.

National Institute on Alcohol Abuse and Alcoholism (1994). "Alcohol and Minorities." Alcohol Alert 23, p. 347.

National Institute on Alcohol Abuse and Alcoholism (2009). "Percent Distribution of Current Drinking status, Drinking Levels, and Heavy Drinking Days by Sex for Persons 18 Years of Age and Older: United States." HIS, 1997-2008.

Nielsen, A.L. (2000). "Examining Drinking Patterns and Problems among Hispanic Groups: Results from a National Survey." Journal of Studies on Alcohol and Drugs 61, 301-310.

Nyaronga, D., Greenfield, T.K. & McDaniel, P.A. (2009). "Drinking Context and Drinking Problems Among Black, White, and Hispanic Men and Women in the 1984, 1995, and 2005 U.S. National Alcohol Surveys." Journal of Studies on Alcohol & Drugs 70, 16-26.

Peralta, R.L. (2007). College Alcohol Use and the Embodiment of Hegemonic
"Masculinity Among European American Men." Sex Roles 56, 741-756.

Peralta, R.L, Steele, J.L., Nofziger, S, & Rickles, M. (2010). "The Impact of Gender on Binge Drinking Behavior among U.S. College Students Attending a Midwestern University: An Analysis of Two Gender Measures." Feminist Criminology 5(4), 355-379.

Rebach, H. (1992). "Alcohol and Drug use among Minority African American Minorities." Ethnic and Multicultural Drug Abuse 23-57.

Robbins, C. & Martin, S.S. (1993). "Gender, Styles of Deviance, and Drinking Problems." Journal of Health and Social Behavior 34, 302-321.

Room, R. (1996). "Gender Roles and Interactions in Drinking and Drug Use." Journal of Substance Abuse 2, 227-239.

Sweet, J.A. and Bumpass, L.L. (1996). "The National Survey of Families and Households - Waves 1 and 2: Data Description and Documentation." Center for Demography and Ecology, University of Wisconsin-Madison (http://www.ssc.wisc.edu/nsfh/home.htm)

Trotter, R.T. (1985). "Mexican-American Experience with Alcohol South Texas Examples." In Bennett, L.A. and Ames, M.A. (Eds.) The American Experience with Alcohol: Contrasting Cultural Perspectives. (pp. 279-296) New York, NY: Plenum Press.

U.S. Bureau of the Census. (2011a). "The Hispanic Population: 2010, 2010." Census Briefs,
http://www.census.gov/prod/cen2010/briefs/c2010br-04.pdf

U.S. Bureau of the Census. (2011b). "Age and Sex Composition: 2010, 2010." Census Briefs,
http://www.census.gov/prod/cen2010/briefs/c2010br-03.pdf.

U.S. Bureau of the Census. (2011c). "Population Distribution and Change: 2000 to 2010, 2010." Census Briefs,
http://www.census.gov/prod/cen2010/briefs/c2010br-01.pdf

Wallace, J.M., Bachman, J.G., O'Malley, P.M., Schulenberg, J.E., Cooper, S.M. & Johnston, L.D. (2003). "Gender and Ethnic Differences in Smoking, Drinking, and Illicit Drug use among African Americans 8th, 10th, and 12th Grade Students, 1976-2000." Addiction 98, 225-234.

Waterson, J. (2000). Women and Alcohol in Social Context. Hampshire, UK: Palgrave Publishers, St Martin's Press.

Wilsnack, R.W. & Cheloha, R. (1987) "Women's Roles and Problem Drinking Across the Lifespan." Social Problems 34(3), 231-248.

Wilsnack, R.W. & Wilsnack, S.C. (1992). "Women, Work, and Alcohol: Failures of Simple Theories." Alcoholism: Clinical and Experimental Research 16(2), 172-179.

Wilsnack, R.W., Kristjanson, A.F., Wilsnack, S.C. & Crosby, R.D. (2006). "Are U.S. Women Drinking Less (or More)? Historical and Aging Trends, 1981-2001." Journal of Studies on Alcohol 67, 341-348.

Wilsnack, R.W., Wilsnack, S.C., Kristjanson, A.F. , Vogeltanz-Holm, N.D. & Gmel, G. (2009). "Gender and Alcohol Consumption: Patterns from the Multinational GENACIS Project." Addiction 104, 1487-1500.