In Europe, alcohol is one of the major risk factors for disease burden, although, in general, alcohol consumption, as well as irregular heavy drinking, have decreased in the last years [1
], Europe is still the region with the highest consumption levels worldwide [2
]. In order to reduce alcohol consumption further, the “European action plan to reduce the harmful use of alcohol 2012–2020” [3
] lists 10 action points which aim to improve national alcohol policies in European countries, and therefore reduce risky drinking behaviors. The 10 action points involve enhancing the population’s awareness and commitment, improving health services’ responses, encouraging community and workplace actions, implementing drinking-driving policies and countermeasures, reducing the availability of alcohol, instituting marketing restrictions, increasing pricing, reducing the negative consequences of drinking and alcohol intoxication and the public health impact of illicit alcohol, and, finally, improving monitoring and surveillance efforts. In 2016, there was a relatively high overall implementation of alcohol policies, however, some of the most effective and cost-effective policies such as pricing or marketing regulations [4
] lagged behind [1
The effectiveness of key alcohol policy measures has been substantiated by a large body of reviews and meta-analyses, although they vary in specific aims (e.g., reducing alcohol intake directly or indirectly by enhancing awareness of alcohol-related harm, see [6
]). However, public opinions, such as perceived effectiveness of and attitudes toward alcohol policies, which probably interact with the implementation and effectiveness of some alcohol policy measures [8
], are sometimes neglected by researchers but play an important role for policymakers. In fact, alcohol policy endorsement is not static but changes over time [11
] and routine monitoring of public opinion regarding alcohol policies can contribute to successful implementation of alcohol policy measures. As discussed by Giesbrecht and Livingston, a shift in public opinion can support changes in alcohol policies, and conversely changes in alcohol policies can also alter attitudes toward alcohol policies [8
]. The public acceptance of policy measures depends on the respondents’ knowledge of the expected outcomes, as demonstrated in the case of minimum unit pricing (MUP) for alcohol in the UK [15
]. The acceptability of MUP have been increased significantly when the expected outcomes of the policy are clearly outlined for the respondents. In addition, beliefs regarding the effectiveness of alcohol policy measures and beliefs regarding the harm caused by drinking have been demonstrated to affect attitudes toward alcohol policy measures [16
To date, several publications have focused on sociodemographic and demographic differences of attitudes toward alcohol policies and on the relationship between drinking behaviors and alcohol policy endorsement [16
]. In general, based on these studies, women and older adults were more likely than men and younger adults to support alcohol policies [16
]. Furthermore, individuals with higher educational levels have reported higher rates of endorsement of restrictions on alcohol use as compared to individuals with lower educational levels [17
]. Heavier drinkers showed higher rates of rejection of restrictions and taxation as compared with individuals who reported light drinking or abstaining [16
]. However, previous findings are often limited to a low number of countries, meaning that country-specific factors, such as national alcohol policies, could not be taken into account.
In this study, we used data on attitudes toward alcohol policies from 19 European countries and one subnational region to examine different clusters of alcohol policy endorsement, to compare them between supranational regions, and to evaluate their relationship to individual respondent characteristics. The data were collected during the course of the Standardized European Alcohol Survey (SEAS), which was implemented as part of the EU Joint Action on Alcohol known as the “Reducing Alcohol Related Harm” (RARHA) project in 2015 and 2016. Significant variations among attitudes toward alcohol policies across the 20 European locations were demonstrated in the published Synthesis report, which identified the following three clusters of attitudes through factor analyses using the same data (Table 1
): Population-based alcohol control policies (e.g., support for controlling the number of places selling alcohol), education and individual-based alcohol policies (e.g., support for printed warnings about alcohol-related harms), and laissez-faire alcohol policies (e.g., support for individual responsibility) [19
Using the RARHA SEAS data, this study pursued two aims which were: (1) to compare clusters of alcohol policy measures between different European regions and (2) to analyse associations between categories of alcohol policy endorsement and individual characteristics (gender, age, educational achievement, and drinking status) taking into consideration national alcohol policy scores. Building on previously identified factor structure of policy endorsement [19
], i.e., population-based alcohol control, education and individual-based alcohol policies, and laissez-faire alcohol policies, we expected to identify the same three clusters in the latent class analyses (LCAs). LCA is a probabilistic model-based clustering approach which identifies individual class membership probabilities for each individual belonging to each class instead of single class membership as defined in factor analysis [20
]. We were, therefore, were able to investigate associations between (sociodemographic and demographic characteristics and clusters of alcohol policy endorsement in all respondents using class membership probabilities. At the individual-level, we investigated the following hypotheses based on past publications: (a) Women and (b) older adults are more likely to endorse classes, which are most related to the factors “population-based alcohol control policies” and “education and individual-based alcohol policies”, whereas men and respondents of younger age are more likely to favor “laissez-faire alcohol policies” and (c) Abstainer and (d) respondents with a higher educational achievement are more likely to endorse “population-based alcohol control policies”’ in contrast to drinker and respondents with lower educational levels.
2. Materials and Methods
2.1. Data Collection
Survey data were used from the RARHA SEAS in 2015 and 2016, where 33,237 adults from 19 European countries and one region (Autonomous Community of Catalonia) participated (see Table 2
for sample characteristics by location).
In all locations, randomized sampling procedures were applied to select a representative general population sample aged 18 to 64 years. The sample size was about 1500 people in most locations, with few exceptions. The mode of administration differed between locations, but for most locations either computer-assisted telephone interviews or computer-assisted personal interviews were applied.
To assess individual consumption levels for the past 12 months, respondents provided data on their usual intake of three basic beverage types (beer, wine, and spirits using the beverage-specific quantity frequency approach). In addition, questions on frequency of risky single-occasion drinking (RSOD) were asked assuming a RSOD threshold of 40 grams of pure alcohol for women and 60 grams for men. Both measures were combined to estimate annual consumption of pure alcohol. To avoid overestimation of individual consumption, capping procedures were applied at 0.5 L of pure alcohol for daily alcohol intake of any alcoholic beverage, and at 182.5 L of pure alcohol for annual consumption.
In our analyses, we considered the following covariates: Gender (women and men), three age groups (≤34 years, 35 to 49 years, and ≥50 years), socioeconomic status as defined by educational achievement (primary and lower secondary education, secondary education, and high education), and drinking status grouped into (past-year) abstainers, low, and high-risk drinking (intake of pure alcohol per drink day, women ≥ 20 g and men ≥ 40 g). Individuals with missing data on gender, educational achievement, or quantity of drinking were excluded (n = 457, 1.4% of the sample).
2.2. Attitudes Toward Alcohol Policies
Attitudes toward eleven different alcohol policy measures were assessed (see Table 1
for details). Respondents identified how strongly they agreed or disagreed with each statement (strongly agree, somewhat agree, somewhat disagree, and strongly disagree). For statistical analyses, the variables were dichotomized and coded as “endorse”, if a respondent indicated “somewhat agree” or “strongly agree”, or as “not endorse”, if a person indicated “somewhat disagree” or “strongly disagree”. For the current contribution, we considered three clusters of attitudes toward alcohol policies based on the factor analysis reported in the Synthesis report [19
]: Population-based alcohol control policies, education and individual-based alcohol policies, and laissez-faire alcohol policies. For a total number of 32,641 respondents (i.e., 0.42% missing values on the policy endorsement variables), data was available and analyzed.
2.3. Country-Level Variables
Country-specific alcohol policy scores were taken from the AMPHORA project [21
]. The summary score represents the degree of implementation of six alcohol policy measures in 2010, namely, (1) control of production, retail sale, and distribution of alcoholic beverages; (2) age limits and personal control; (3) control of drunk driving; (4) control of advertising, marketing, and sponsorship of alcoholic beverages; (5) public policy; and (6) alcohol taxation and price, which were taken into account using different weights (for details, see [21
]). The resulting alcohol policy score indicates three levels of strictness of alcohol policy (liberal, medium, and strict) whereby the strictest policies were identified in the Nordic countries (e.g., Finland) and the most liberal in Central (e.g., Germany) and South Europe (e.g., Italy). For the Autonomous Community of Catalonia, we used the same score as for Spain.
The assignment of countries to European regions was based on the AMPHORA classification as well. Iceland, Finland, Norway, and Sweden were grouped as Nordic countries. Denmark, which is usually classified into this group, was included in the West and Central European region because of the more liberal alcohol policies found there as compared with the other Nordic countries. Austria, France, and the UK were also associated with the West and Central European region. Greece, Italy, Portugal, Spain, and the Autonomous Community of Catalonia were included in the Southern European region and the remaining countries formed the Eastern European region (Bulgaria, Croatia, Estonia, Hungary, Lithuania, Poland, and Romania).
2.4. Statistical Analyses
LCAs were run to identify the cluster structure of policy endorsement. LCAs are defined by the prevalence of each class and the probability that an individual in a certain class would endorse a certain item [20
]. Therefore, this is a person-centered probabilistic approach which is not exclusive to single-class membership (i.e., in contrast to exploratory factor analysis). Every class is described by all variables of interest based on different endorsement probabilities. In our investigation, responses to the eleven items assessing alcohol policy endorsement were used to identify latent classes. Because previous factor analysis suggested that all policy measures could be grouped into three distinct categories, a three-class LCA model was performed. Weights were taken into consideration to account for sampling bias. On the basis of the fitted LCA model, we obtained three variables ranging between 0 and 1, indicating the probability of each individual belonging to each class. Those probabilities were used as dependent variables in the following regression analyses.
Two sets of regression models were run. In the first set, three hierarchical regression models were calculated to compare European regions (independent variable) for the probabilities to endorse in each of the three classes (dependent variable). In the second set of hierarchical regression analyses, the same three dependent variables, gender, age, educational achievement and drinking status were included as predictors. The latter set of models were adjusted to the national alcohol policy score to control for effects of local policy strictness. As the dependent variable ranged between 0 and 1, we fitted a logistic regression with standard error estimation based fractional response regression models [22
]. In these models the intercept was allowed to vary across locations (i.e., random intercept model). Survey weights were applied in all regression analyses. All statistical analyses were performed using Stata 15.1 [23
] and R version 3.6.1 [24
Our findings suggest that attitudes toward alcohol policy vary significantly between European regions. We identified three classes of alcohol policy endorsement which did not fully overlap with the previously identified factor structure. The three classes referred to support and rejection of alcohol control policies in addition to an “acquiescence tendency” class, which we discuss below. Overall, rejection of alcohol control policies was the most prevalent class, followed by its antipode, support for alcohol control policies. We showed that the support for alcohol control policies was mostly driven by Nordic European countries, while Eastern European countries showed significantly less support. In accordance with the dichotomy of both classes, two contrary patterns of those who support or reject control policies were identified as follows: Women, middle-aged adults, and abstainer preferably supported alcohol control policies, in contrast to men, adults aged younger than 35 years, and drinkers, who were more likely to reject alcohol control policies. Notably, educational achievement was not predictive of either support or rejection of alcohol control policies.
For a comparison of the current findings with the Synthesis report [19
], methodological differences should be taken into account. In the current study, classes were characterized by the endorsement probabilities of all items of interest, instead of a selection as is usual in factor analysis. Therefore, the class structure and their properties differ from those presented in the Synthesis report. For example, belonging to the first class (i.e., support of alcohol control policies) means to report more support for items with high endorsement probabilities for this class (e.g., taxation, p
= 69.9%) and less support for items with low endorsement probabilities (e.g., alcohol is a commodity as any other, p
= 4.7%) in the survey. As a result of using a divergent approach, we were not able to replicate the cluster structure from the Synthesis report. Particularly critical is the third class where all items questioned had high endorsement probabilities. Since some items are contradictory to each other in this class (e.g., “parents, and no legal authorities, should decide at what age their child is allowed to drink alcoholic beverages“ versus “public authorities have the responsibility to protect people from being harmed by their own drinking”), we assumed that this class might reflect an acquiescence tendency among respondents. Bias due to acquiescence can occur for several reasons; the stimuli (e.g., cognitive load to the respondent), the respondent (e.g., education), or country-level indicators (e.g., collectivism) can be sources for higher acquiescence [25
]. Remarkably, previous studies reported similar patterns of individual [25
] and cross-country variations [27
] in the acquiescence tendency in Europe as we found for class 3. For example, Van Herks and colleagues described a higher acquiescence tendency in the Mediterranean region than in the Northwestern European region. However, the high endorsement probabilities to all items in this class could also arise from the respondents’ tendency to at least “somewhat agree” with the policy statements, which would not be visible after a dichotomization of the response options.
Some further limitations have to be taken into consideration. First, the representativeness of survey data is limited due to the following reasons: (a) Samples are not representative of the entire population because they exclude certain groups of people, including the homeless and the imprisoned or otherwise institutionalized persons [28
]; (b) alcohol surveys have a high nonresponse rates, often exceeding 50%, which was shown to be related to an underrepresentation of certain groups of people (i.e., individuals with low income or heavy episodic drinking [29
]); and (c) there is an undercoverage of reported alcohol consumption compared to “real consumption” estimates [30
]. Furthermore, limited generalizability of the results has to be considered due to great variations in the population size between locations, and therefore they might be not transferable to the whole European region. Finally, we investigated associations, and therefore we cannot draw conclusions about the direction of effects or why individuals endorse particular alcohol policies.
Our study presents evidence for variations in alcohol policy endorsement across European regions. In addition to alcohol policies being traditionally stricter in the Nordic region as compared to other European countries [21
], we show that restrictive alcohol measures received more support in this region than elsewhere in Europe. In all studied regions, alcohol policies are considerably more liberal and control strategies were more likely to be rejected. Several publications reported a substantial increase in supporting alcohol control policies in Sweden, Finland, and Norway, since the millennium [12
]. Although positive attitudes toward those policies increased, alcohol politics in Nordic countries were characterized by liberalization. But why did liberal alcohol policies lead to an increase in support for alcohol control measures there? A possible explanation is that individuals who experienced alcohol-related harm in their personal environment are more likely to endorse restrictive alcohol policies [33
]. Referring to the Nordic countries, consumption levels and alcohol-related harm increased during the time period, where alcohol policies became more liberal [34
]. With alcohol-related harm becoming more prevalent, more individuals are directly or indirectly affected by harm, which could lead to higher levels of awareness and, consequently, to alcohol control policy endorsement. In addition, even the knowledge about alcohol-related harm can predict changes in attitudes [17
]. Another mediator discussed in the literature is a change in the individuals’ beliefs on the effectiveness of alcohol control strategies [11
On the other side, our results indicate that in the Eastern European region there is considerably lower endorsement toward those restrictive policy measures although this is the region with the highest alcohol-related harm and alcohol-attributable mortality in Europe [1
]. It can be argued, that the mediators explained above, i.e., the knowledge on alcohol-related harm and beliefs on the effectiveness of alcohol control strategies, are less widespread in the general population. This hypothesis goes hand in hand with a high prevalence of general acquiescence found in this region in our analyses, whereas the rejection of alcohol control policies was not significantly higher than in the other European regions. Another explanation can be related to general attitudes toward market economy and beliefs in self-regulatory powers of the market or support for liberal, laissez faire economic policies. In all countries which underwent transitions to market economy, the 1990s saw rapid liberalization of alcohol policies and dismantling of a previous control mechanism despite visible growth of alcohol-related problems [37
]. In the decades of transitions toward market economy in Poland, the attitudes toward alcohol policy has changed dramatically. The general population survey carried out in 1992 indicated that only 12.8% of respondents endorsed a statement that “alcohol beverages should be treated as all other commodities and their sales should not be restricted” while 72.2% confirmed that “alcohol beverages must not be treated as all other commodities and their sales should be restricted by the State” [38
]. Twenty-five years later in the RARHA SEAS survey, 62% of Polish respondents endorsed that “alcohol is a product like any other and does not require any special restrictions” [19
]. The Polish experience, as well as experiences of other countries in transition, suggest that prevailing economic ideologies have a crucial impact on attitudes toward alcohol policy [39
]. After decades of liberal policies, many Eastern European countries have considered returning to more restrictive policies. In a recent publication, increasing support of evidence-based alcohol control policies by the members of the Lithuanian Parliament were observed in the years between 2016 and 2018 [40
]. In addition, with regard to the implementation of alcohol control policies in Lithuania between 2004 and 2019, traffic harm, injury, and mortality attributable to alcohol have been found to decrease ever since [41
]. Nevertheless, studies on attitudes toward alcohol policies are scarce in most European countries. In light of our results, it would be interesting to study the interaction of public opinion and policy implementation and examine, for instance, if the former followed or influenced the latter.
We further investigated individual-level differences in alcohol policy endorsement. Our findings are in line with previous studies [16
] and expand existing knowledge on individual characteristics associated with attitudes toward alcohol policies for a large number of European countries. Only one curious finding has to be discussed as both supporting (class 1) and rejecting alcohol control policies (class 2) were related to higher levels of education. We suggest that this is based on a higher tendency for acquiescence by individuals with a primary or lower secondary education [25
]. This was underlined in the regression analysis, since the probability of endorsing the third class (i.e., acquiescence tendency) was up to twice as high in this subgroup as compared with higher educated respondents. However, in terms of individual differences such as gender, age, and educational achievement, it should be noted that they explain only a little about individual attitudes to alcohol policies as compared to individual differences in beliefs regarding alcohol-related harm and the effectiveness of alcohol-control measures [17
]. Moreover, the framing of policies, i.e., how and who presented them to the public, is an important influencing factor to garner public acceptance and higher levels of alcohol policy endorsement [16
]. These changes in public acceptance can be accomplished by targeting those who have higher levels of false beliefs and negative opinions to enhance alcohol policy endorsement more broadly.