Alcoholic beverage types consumed by population subgroups in the United States: Implications for alcohol policy to address health disparities

Abstract

Introduction:

We aimed to identify alcoholic beverage types more likely to be consumed by demographic subgroups with greater alcohol-related health risk than others, mainly individuals with low socioeconomic status, racial/ethnic minority status and high drinking levels.

Methods:

Fractional logit modelling was performed using a nationally-representative sample of US adult drinkers (analytic N=37,657) from the National Epidemiologic Survey on Alcohol and Related Conditions Waves 2 (2004–2005) and 3 (2012–2013). The outcomes were the proportions of pure alcohol consumed as beer, wine, liquor and coolers (defined as wine-/malt-/liquor-based coolers, hard lemonade, hard cider and any prepackaged cocktails of alcohol and mixer).

Results:

Adults with lower education and low or medium income were more likely to drink beer, liquor and coolers, while those with a 4-year college/advanced degree and those with high income preferred wine. Excepting Asian adults, racial/ethnic minority adults were more likely to drink beer (Hispanics) and liquor (Blacks), compared with White adults. High- or very-high-level drinkers were more likely to consume liquor and beer and less likely to consume wine (and coolers), compared with low-level drinkers. High-level and very-high-level drinkers, who were less than 10% of all drinkers, consumed over half of the total volume of beer, liquor and coolers consumed by all adults.

Discussion and Conclusions:

Individuals with low socioeconomic status, racial/ethnic minority status or high drinking level prefer liquor and beer. As alcohol taxes, sales and marketing practices all are beverage-specific, targeted approaches to reduce consumption of these beverages, particularly among individuals with these profiles, are warranted.

Introduction

Health behaviours are influenced by the social, cultural and economic circumstances that frame and constrain them [1]. Alcohol consumption, including the preference of a specific alcoholic beverage type, is among such behaviours. As the circumstances that shape beverage choices likely vary across demographic subgroups, each group may have a beverage type they prefer (that is, which they tend to consume the most).

There is limited information about the beverage types preferred by demographic subgroups. US studies have documented aspects of beverage preference over time, with a study from the 1980’s reporting that individuals who preferred beer were more likely to drink and drive and to drink to higher levels of intoxication compared to those who preferred wine and spirits [2]. A study from 1990 found that spirits preference was associated with middle- and older-aged men having lower education, heavier drinking and health problems; wine drinkers tended to be younger non-smoking women with higher education and few health problems; and beer drinkers were younger men with intermediate levels of other characteristics [3]. A study in the 1990s reported that Hispanic drinkers preferred beer, African American drinkers preferred spirits and both groups were less likely to drink wine than White drinkers [4]. Studies in the 2000s documented heavy spirits and malt liquor drinking by African American men [5], Hispanic drinkers’ preference for beer and its association with binge drinking for this group [6], and beer accounting for two-thirds of all alcohol consumed by binge drinkers [7]. As beverage preferences in population subgroups could change over time, there is a need for a more recent and detailed study of US drinkers and characteristics associated with beverage preferences.

A better understanding of beverage preference by demographic subgroups may inform public health and alcohol policy interventions to reduce alcohol-related health risks, particularly alcohol price and tax policies. Alcohol control policies not only try to reduce overall level of alcohol consumption, but specifically aim to reduce alcohol-attributable harm in the subgroups that may bear the greatest alcohol-related harms [8]. Heavy drinkers, for one, have greater health risks for most health conditions (including ischemic stroke and diabetes reported in some studies to benefit from light/moderate drinking) [9,10]. Alcohol excise taxes are beverage specific, as are alcohol sales and marketing practices [7]. Tax rates and prices can shape preferences of alcohol users and hence be used to meaningfully increase prices for (i.e., lower affordability of) alcoholic beverages preferred by high-risk groups, including heavy drinkers. US tax rates, though varying by state, are typically highest on spirits compared to wine and beer and are generally low compared to other high-income countries [11]. Despite higher tax rates on spirits, low-cost spirits products in large containers are the cheapest form of alcohol, followed by large boxes of wine [11]. Low-cost beer is also available in cases of 24 and large containers, and alcohol for off-premise consumption is very affordable even to low-income consumers [11]. As low-cost alcohol products are available across beverage types, beverage choices of heavy drinkers in the US may be driven more by their preferences than by price. Knowledge about beverage preferences may not only help to reduce consumption levels in high-risk groups, but also prevent potential substitutions (e.g., by specifically devising taxation structures based on beverage and group-specific elasticities and cross-elasticities, and targeting the beverage(s) preferred by high-risk groups in setting a floor price for them).

The current study is informed partly by the alcohol harm paradox that refers to the disproportionate burden of alcohol-attributable morbidity and mortality borne by adults with low socioeconomic status (SES), which neither average alcohol consumption nor heavy drinking patterns can explain [12–14]. Supporting this thesis, a systematic review found that individuals with low SES had 3.8–5.2 times higher risk for alcohol-attributable mortality and 1.6–3.6 times higher risk for all-cause mortality than high-SES individuals, with the differences insufficiently accounted for by consumption level [15]. Similarly, racial/ethnic minority adults, particularly Black and Hispanics, are also at higher risk of alcohol-related conditions (e.g., hypertension, heart disease) and mortality [16–18]. Blacks, for example, had 21% higher heart disease mortality rate than Whites in 2015 [17], although White adults were more likely to drink and engage in at-risk alcohol use than Blacks and Hispanics [19].

Potential mechanisms suggested for the alcohol harm paradox involve other behavioural risk factors such as obesity and smoking [12], limited access to health care and more harmful drinking patterns [12,15], which are likely to be more prevalent among low-SES individuals and may interact with alcohol use [15]. Beverage types associated with greater health risks—more prominently, liquor, and to a lesser degree, beer—may constitute another mechanism for the paradox [12] to the extent they are preferred by low-SES and racial/ethnic minority individuals. A study conducted in Wales reported that unit increases of spirits consumed were positively associated with increasing risk of wholly attributable alcohol-related hospital admissions, higher than for other drink types [20]. Liquor was also consistently associated with greater risks for conditions partially attributable to alcohol such as diabetes and cardiovascular disease, and mortality [21,22], with greater adverse effects than beer [22]. Beer was associated with greater risk for cardiovascular disease and all-cause mortality, compared with wine [22]. Adverse health effects of liquor are attributed to more concentrated levels of ethanol and associated consumption patterns such as rapid ethanol intake and intoxication [23], as opposed to less harmful patterns associated with wine like consuming with meals [24].

In the current study, we aim to identify alcoholic beverages preferred by the groups at greater alcohol-related health risk, mainly individuals with low SES, racial/ethnic minority group members and heavy drinkers. Given the differential health risks noted above, we hypothesized that people with low SES, racial/ethnic minorities and heavy drinkers preferred liquor and beer to wine.

Methods

Data

A nationally representative sample of US adults aged 18+years was drawn from the cross-sectional National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) Waves 2 (2004–2005) and 3 (2012–2013). NESARC used multistage probability sampling and computer-assisted personal interviewing, with the response rates being 87.6% at Wave 2 and 84.0% for Wave 3 [25,26]. Current drinkers (analytic N=37,657) were included in this study as NESARC provides beverage-specific information only for them. This secondary data analysis study was approved by the Public Health Institute Institutional Review Board.

Measures

The outcomes were the proportions of pure alcohol consumed as wine, beer, liquor and coolers (including wine-, malt- and liquor-based coolers, hard lemonade, hard cider and any prepackaged cocktails with the alcohol and mixer), each of which is the ratio of the individual’s beverage-specific volume to their total consumption volume of pure alcohol. The beverage-specific daily volume was first calculated using the reported frequency, usual amount consumed per day, and frequency of drinking of 5+ and 6+ (women/men) drinks per drinking day, following the steps that were specified in the NESARC’s data reference manual [27]. For those who consumed 5 or fewer drinks, the average daily volume was calculated using two components (the usual quantity times its frequency and the largest quantity times its frequency); for those who consumed more, an intermediate component (the geometric mean of quantities between 5 and the largest number of drinks) was added. The amount of ethanol (pure alcohol) in each drink was then calculated by multiplying the daily volume, the size of each typical drink by beverage type and the ethanol conversion factor (a NESARC variable based on the subtype of each beverage usually consumed). Our interpretations of drink sizes mainly followed the NESRAC coding except that we coded a shot as 1.5 ounces, not one ounce, based on our prior work which found shots in bars were mostly 1.5+ ounces and very few shots at home were as small as one ounce [28]. Subsequently, the proportion of each beverage for each individual was calculated by dividing the beverage-specific consumption volume by the total volume.

The average daily drinking level was operationalized as a 4-category ordinal variable reflecting the World Health Organization drinking levels: low-level (≤40 grams/<20 grams per day), medium-level (40–60 grams/20–40 grams), high-level (60–100/40–60 grams) and very-high-level (>100 grams/>60 grams) for men/women [29]. SES was operationalised using educational attainment (< high school graduation, high school diploma/equivalency, some college education and 4-year college/advanced degree) and family income ((low (<152.5%), medium (>152.5% and <354.0%) and high (>354.0%), each reflecting a tertile based upon the ratio of family income to the corresponding survey year’s US Federal Poverty Level) [30]. Self-reported race/ethnicity was assessed using the NESARC categories of White, Black, Hispanic, Asian, American Indian/Alaska Native (AI/AN) and Native Hawaiian/Pacific Islander (NHPI). Models also adjusted age (categorical: 18–24, 25–34, 35–44, 45–54, 55–64, 65+ years) and sex (male and female).

Analysis

Analyses were conducted using STATA version 17 [31]. Using the svy function in Stata, analyses were weighted to account for the complex survey design and post-stratification adjustments. We first computed the unweighted sample size and the weighted percentage (%) by demographic characteristics and drinking levels, and weighted mean and standard errors of the proportion of each beverage, also performing bivariate one-way tests to examine differences in weighted average beverage proportions across subgroups. We then computed the percentages of each beverage-specific volume consumed by drinkers of different drinking levels to understand how much of the total volume of each beverage was consumed by high- and very-high-level drinkers.

Given the compositional structure of the data (that the proportion of each beverage, ranging from 0 to 1, adds up to 1), fractional logit modelling was implemented [24,32] to estimate the associations of each proportion of beverage with drinking level, race/ethnicity, education and family income. The model used for each beverage proportion is shown below, where p is the proportion of each beverage in total consumption and $log\left[\frac{p}{1-p}\right]$ is modeled as a linear function of the predictors [31]. As the proportion of each beverage was interpretable as a probability of drinking alcohol as each specific beverage (i.e., coolers, beer, wine, liquor), we interpreted $\frac{p}{1-p}$ as the odds of drinking each specific beverage and obtained odds ratios $\left(e^{\beta }\right)$ by exponentiating the estimated coefficients from the fractional logit models.

$$E\left(log\left.\left[\frac{p}{1-p}\right]\right|\mathit{x}\right)=\mathit{x}\mathit{\beta }$$

Results

Demographic characteristics, current drinking level and proportions of beverage type consumption

Sample characteristics are shown in Table 1. The majority (83.02%) of drinkers consumed alcohol at low-levels, and less than 10% of them were high-level and very high-level drinkers. Figure 1 visually depicts the mean proportions of beverage-specific consumption by demographic characteristics and drinking levels. The mean proportion of wine was the highest among individuals with the highest level of education and the lowest among individuals with the lowest education; the opposite was the case for beer. Mean proportions of coolers and liquor were the highest among adults with some college education, closely followed by those with a high school diploma. As for family income, mean proportions of coolers, beer, and liquor were the highest among adults with the lowest income, and the lowest among those with the highest income. The opposite was the case for the proportion of wine.

Table 1.

Sample characteristics (unweighted n; weighted %)

	n (%)
Sex
Female	18,470 (44.6)
Male	19,187 (55.5)
Age, years
18–24	4,455 (13.1)
25–34	8,706 (21.7)
35–44	8,501 (21.4)
45–54	7,433(19.8)
55–64	4,979 (14.0)
65+	3,583 (10.1)
Education
< High school diploma	4,074 (9.3)
High school diploma	9,179(23.9)
Some college	13,258 (34.6)
4-year college degree+	11,146 (32.2)
Family income
Low income	10,430 (23.9)
Medium income	12,140 (31.3)
High income	15,087 (44.8)
Race/ethnicity
Whites	22,548 (72.0)
Blacks	6,288 (9.4)
Hispanics	6,962 (12.8)
Asians	1,147 (3.6)
American Indian/Alaska Native	567 (1.8)
Native Hawaiian/Pacific Islander	145 (0.4)
Drinking level
Low risk	31,051 (83.0)
Medium risk	2,937 (7.6)
High risk	1,691 (4.4)
Very high risk	1,978 (5.0)

Figure 1.

Proportion of pure alcohol consumed (%). *Bars represent the mean proportions of beverage-specific consumption.

By race/ethnicity, mean proportion of wine was the highest among Asian adults, followed by White and NHPI adults. Mean proportion of beer was the highest among Hispanic adults, followed by AI/AN and NHPI adults. Mean proportions of coolers and liquor were higher among Black and AI/AN adults than other groups, although these differences were not significant.

As for drinking levels, the mean proportion of wine was the highest among the medium-level drinkers, followed by low-level drinkers, and that of coolers was also the highest among the low-level drinkers. Mean proportions of beer and liquor were the highest among the very-high-level drinkers, followed by high-level drinkers.

Proportions of beverage volume consumed by adults with different drinking levels

Table 2 shows the percentages of each beverage-specific volume consumed by drinkers with different overall drinking levels. Notably, over 40% of the total volume of coolers, beer and liquor consumed by all adult drinkers were consumed by very high-level drinkers, and over half (50.1% of coolers, 56.1% of beer and 57.0% of liquor) by high-level or very-high-level drinkers. Men accounted for a disproportionate share of each of these beverages; for example, of the 40.7% of beer consumed by very-high-level drinkers, 30.9% were consumed by men and 9.7% by women. Only 18.3% of the total volume of wine was consumed by very high-level drinkers, with almost 70% of it being consumed by low- or medium-level drinkers. Women’s share of wine consumption volume was consistently higher than men’s across drinking levels, with the sex difference most pronounced among medium-level drinkers.

Table 2.

Percentages of total alcohol volume consumed by drinkers with different drinking levels

	Low level (%)	Medium level (%)	High level (%)	Very high level (%)
Cooler
All	37.5	12.5	9.5	40.6
Male	13.8	4.0	4.7	22.1
Female	23.6	8.5	4.8	18.5
Beer
All	30.1	13.8	15.4	40.7
Male	24.6	9.7	13.1	30.9
Female	5.6	4.1	2.4	9.7
Wine
All	46.0	22.3	13.4	18.3
Male	21.7	6.3	5.9	7.4
Female	24.3	16.0	7.5	10.9
Liquor
All	29.1	13.8	13.8	43.2
Male	19.3	7.1	9.8	28.6
Female	9.8	6.8	4.0	14.6
All beverages
All	29.7	14.7	14.9	40.7
Male	21.5	8.8	11.6	28.6
Female	8.2	5.9	3.4	12.1

Low-level: up to 40 grams/20 grams per day for men/women). Medium-level: >40–60 grams/>20–40 grams for men/women. High-level: >60–100/>40–60 grams for men/ women. Very-high-level: >100 grams/>60 grams for men/women.

Associations of beverage proportions with SES, race/ethnicity and drinking level

Table 3 shows the results from fractional logistic regression modelling to examine the associations between beverage proportions and SES, race/ethnicity and drinking level, controlling for age and sex. Overall, there was a gradient between the two SES variables, education and family income, and the likelihood of drinking beer and wine. Compared with adults with a 4-year college/advanced degree, other adults all had a higher likelihood of drinking beer and a lower likelihood of drinking wine, with those who did not graduate from high school had the highest likelihood of drinking beer and the lowest likelihood of drinking wine. Similarly, with high income as reference category, the likelihood of beer drinking was the highest (and the likelihood of wine drinking the lowest) for those with low family income, followed by those with medium income. The patterns of associations of education with coolers and liquor were not linear, with adults with a high school diploma and those with some college education having the highest likelihood of drinking coolers and liquor, respectively. Adults with low and medium income were more likely to drink coolers and liquor, compared with those with high income.

Table 3.

Associations of beverage proportions with sex, age, socioeconomic status, race/ethnicity and drinking level for US adults (N=37,657)

	Cooler		Beer		Wine		Liquor
	aOR	95% CI	aOR	95% CI	aOR	95% CI	aOR	95% CI
Sexa
Male	0.27***	0.25–0.30	4.14***	3.95–4.35	0.27***	0.26–0.29	0.80***	0.77–0.84
Age, yearsb
25–34	0.77***	0.69–0.86	1.43***	1.33–1.54	1.33***	1.19–1.49	0.69***	0.64–0.75
35–44	0.59***	0.52–0.66	1.42***	1.31–1.53	1.81***	1.6–2.04	0.61***	0.57–0.66
45–54	0.42***	0.37–0.48	1.28***	1.17–1.39	2.46***	2.18–2.77	0.58***	0.53–0.63
55–64	0.32***	0.2 −0.37	0.89*	0.81–0.98	3.53***	3.10–4.00	0.66***	0.61–0.73
65+	0.15***	0.12–0.19	0.58***	0.52–0.64	5.22***	4.54–6.00	0.83***	0.75–0.91
Educationc
Did not graduate high school	1.47***	1.22–1.78	2.64***	2.39–2.90	0.22***	0.19–0.26	0.91	0.82–1.01
High school diploma	1.78***	1.56– 2.04	1.83***	1.71–1.96	0.34***	0.31–0.36	1.13**	1.05–1.21
Some college education	1.56***	1.40–1.74	1.35***	1.28–1.43	0.52***	0.49–0.56	1.24***	1.17–1.32
Family incomed
Low income	1.25***	1.12–1.41	1.26***	1.18–1.34	0.60***	0.56–0.65	1.10**	1.03–1.17
Medium income	1.24***	1.12–1.37	1.17***	1.11–1.24	0.70***	0.66–0.74	1.09**	1.03–1.16
Race/ethnicitye
Black	1.36***	1.22–1.51	0.62***	0.58–0.67	1.00	0.91–1.09	1.50***	1.40–1.60
Hispanic	0.82*	0.70–0.96	1.14**	1.06–1.23	0.98	0.89–1.08	0.89**	0.82–0.96
Asian	0.66**	0.49–0.88	0.87	0.75–1.02	1.46***	1.26–1.69	0.99	0.85–1.16
American Indian/Alaska Native	0.98	0.74–1.31	1.04	0.86–1.24	0.76*	0.60–0.96	1.17	0.98–1.41
Native Hawaiian/Pacific Islander	0.51	0.20–1.29	0.87	0.64–1.19	1.44	0.97–2.14	1.04	0.74–1.47
Drinking levelf
Medium level	0.30***	0.20–0.36	1.21***	1.11–1.31	1.08	0.98–1.18	0.99	0.91–1.08
High level	0.33***	0.25–0.42	1.12*	1.01–1.24	0.86**	0.77–0.96	1.23***	1.11–1.35
Very high level	0.36***	0.29–0.46	1.13*	1.02–1.24	0.54***	0.46–0.63	1.54***	1.38–1.72

aOR, adjusted odds ratio; CI, confidence interval.

^***p <0.001

^**p <0.01

^*p <0.05.

^aFemale as reference group.

^bAges 18–24 as reference group.

^c4-year college/advanced degree as reference group.

^dHigh family income as reference group.

^eWhite as reference group.

^fLow level drinking as reference group.

Low-level: up to 40 grams/20 grams per day for men/women). Medium-level: >40–60 grams/>20–40 grams for men/women. High-level: >60–100/>40–60 grams for men/ women. Very-high-level: >100 grams/>60 grams for men/women.

With White adults as the reference group, Black adults were more likely to drink liquor and coolers, and less likely to drink beer, and Hispanic adults were more likely to drink beer and less likely to drink coolers and liquor. Asian adults were more likely to drink wine and less likely to drink coolers, and AI/AN adults were less likely to drink wine than White adults.

With low-level drinkers as the reference group, medium-, high- and very-high-level drinkers all were less likely to drink coolers; high- and very-high-level drinkers were less likely to drink wine. Also compared to the low-level drinkers, medium-, high- and very-high-risk level drinkers were more likely to drink beer, although the effect size was larger for medium-level drinkers than for high- and very-high-level drinkers. The likelihood of drinking liquor, on the other hand, was the highest for very-high-level drinkers, followed by high-level drinkers. There was no significant association between medium-level drinking and wine or liquor drinking.

Addressing a reviewer comment, we conducted an ad hoc analysis to better understand beverage preference of heavy drinkers within each disadvantaged subgroup, that is, another set of fractional logit regressions focusing on beer and liquor preference among Blacks, Hispanics and individuals with low income and with only high school education/diploma (Table 4). With low-level drinking as reference category, very-high-level drinking was consistently associated with liquor preference for each subgroup (excepting those with only high school education/diploma). Also notably, this pattern was observed for Hispanics, who in aggregate were found to prefer beer. Beer preference was not associated with drinking level among Hispanic adults, but was significant for all other drinking levels compared to low-level drinking in other demographic groups.

Table 4.

Associations of beverage proportions with drinking level for demographic subgroups

	Blacks (n=6,211)		Hispanics (n=5,759)		Low income (n=10.430)		No college education (n=13,253)
	aOR	95% CI	aOR	95% CI	aOR	95% CI	aOR	95% CI
Beer
Medium levela	1.43***	(1.18–1.74)	1.04	(0.82–1.33)	1.66***	(1.45–1.91)	1.52***	(1.32–1.74)
High levela	1.33*	(1.02–1.74)	1.15	(0.82–1.61)	1.43***	(1.18–1.74)	1.38***	(1.17–1.63)
Very high levela	1.45***	(1.18–1.80)	0.81	(0.59–1.11)	1.29**	(1.08–1.53)	1.38***	(1.17–1.61)
Liquor
Medium levela	1.05	(0.87–1.28)	1.07	(0.82–1.40)	0.85*	(0.74–0.99)	0.85*	(0.73–0.99)
High levela	1.10	(0.86–1.40)	1.23	(0.92–1.64)	1.03	(0.86–1.24)	1.04	(0.87–1.24)
Very high levela	1.27*	(1.04–1.57)	1.63**	(1.21–2.21)	1.36***	(1.17–1.60)	1.13	(0.96–1.32)

Adjusting for sex, age, education and family income.

aOR, adjusted odds ratio; CI, confidence interval.

^***p <0.001

^**p <0.01

^*p <0.05.

^aLow level drinking` as reference group.

Discussion

In this study we aimed to identify alcoholic beverage types preferred by US subgroups at greater risks for alcohol-related harms, namely, those with lower SES, racial/ethnic minority status and high drinking levels. We found that these groups preferred beer and liquor and, less consistently, coolers. That is, adults with lower education and low/medium income were more likely to drink beer, liquor and coolers, while those with a 4-year college/advanced degree and those with high income preferred wine. Excepting Asian adults who preferred wine, racial/ethnic minority adults were more likely to drink beer (Hispanics) and liquor (Blacks), compared with Whites. In a similar vein, high- or very-high-level drinkers were more likely to consume beer and liquor and less likely to consume wine (and coolers), compared with low-level drinkers. Very-high-level drinkers within disadvantaged subgroups (i.e., Blacks, Hispanics and low-income) preferred liquor.

Our findings that low-SES and racial/ethnic minority adults were more likely to prefer beer and liquor might partly explain the alcohol harm paradox. That is, their preference of these beverages associated with greater health risks may contribute to the disproportionate burden of alcohol-related conditions and mortality they bear. In this respect, differential preferences for beverage types may indeed point to a mechanism for the alcohol harm paradox. The cross-sectional design of this exploratory study, however, precludes us from making causal inferences. Future research should further investigate this mechanism in longitudinal study designs.

Given the greater health risks associated with liquor than all other beverages, it is concerning that it is a preferred beverage of Black adults, a group with high mortality from alcohol-related conditions such as heart disease [17] and cancer [18] than all other racial groups in the US. Liquor was also preferred by very-high-level drinkers within most disadvantaged subgroups, including Hispanic adults who in aggregate preferred beer. The combination of liquor preference and very-high-level drinking within disadvantaged groups that have generally lower levels of alcohol consumption offers another, pointed explanation for the alcohol harm paradox.

Importantly, we found that high-level and very-high-level drinkers, who constituted less than 10% of all drinkers, were likely to drink over half of total volume of beer, coolers, and liquor consumed by all adult drinkers. A large body of literature has demonstrated the effectiveness of alcohol price and tax increases in reducing alcohol consumption [33] and alcohol-related morbidity and mortality [34]. Heavy drinkers are more likely to purchase cheaper alcohol than other drinkers [35] and to show the steepest decline in consumption levels when prices increase given their high baseline consumption [36]. Pricing policies resulted in greater consumption changes among low-income alcohol users [36]. Heavy drinkers could switch to lower-quality alcohol products or lower-cost brands in response to higher prices mostly within a given category and not across beverage types [37], especially in the US where low-cost alcohol products are available across beverages [11]. Therefore, increasing prices or taxes on the beverages preferred by heavy drinkers—e.g., by introducing minimum unit pricing to set a floor price for them [36]—might be effective in reducing their consumption. Such a strategy is likely to have similar effects on low-SES and racial/ethnic minority adults as they favour the same beverages. Concerns have been raised that alcohol taxation and pricing policies are regressive (hence adversely affecting low-SES individuals), but research has found any regressive effects are small and concentrated among heavy drinkers [38]. This is likely to be particularly the case in the US where alcohol is affordable, for example, with individuals spending only 0.29%, 1.37% and 0.35% of per capita disposable income to purchase one drink daily of the cheapest brand of spirits, Budweiser beer and low-priced wine, respectively [11].

Prior studies have reported differential effects of alcohol taxes on demographic subgroups. For example, a study reported stronger effects of beer and spirits taxes on consumption volume for Black and Hispanic women than for White women [39]. Our results offer some clues to this. That is, the differences may be attributed, at least partly, to preference of liquor and beer by Black and Hispanic adults, which may have resulted in stronger effects of taxes on these beverages found for these populations.

We acknowledge several limitations to the current study. First, we used self-reported drinking measures, which typically underreport heavy drinking. As noted above, we followed the steps that were specified in the NESARC’s data reference manual [27] in calculating drinking volume to address this. Second, we used survey data to estimate the share of aggregate beverage-specific volume consumed by high-level and very-high-level drinkers. Survey data tend to underestimate ‘real consumption volume’ (that includes recorded and unrecorded volume) at the aggregate level in part because of different alcohol coverage rates across beverages (lower for spirits than for beer and wine) [40]. Third, we had less statistical power for Asian, AI/AN and NHPI adults because of the smaller number of these groups in our sample. This may have hampered our ability to detect significant associations between beverage types and membership in these groups.

Conclusions

To our knowledge, this is the first study to identify alcoholic beverage types preferred by disadvantaged US subgroups that are more likely to experience greater alcohol-related health harms, i.e., those with low SES, racial/ethnic minority status, or high drinking levels. It has been argued that alcohol policies can reduce alcohol-related health risks and thus should be a key component of comprehensive strategies to address health disparities [41]. Our findings meaningfully contribute to the evidence base that can inform the development and implementation of beverage-specific alcohol price and tax policies to help address health disparities.