The values we used for cultural value orientations were those obtained by Schwartz between 1988 and 2000 (please refer to Schwartz, 2006, for full details). Specifically, we used the country averages of all seven cultural value orientations which Schwartz (2008) made openly available. Where cultural value orientations were presented for different cultural groups within a single country (e.g., East and West Germany), these values were averaged. To be clear, what this means is that for the sake of this article a culture refers to the average culture measured within a nation. Naturally, within a nation there are usually many different cultural groups, but we shall consider a nation/country as a single cultural unit. In total, the cultural value orientations for 74 nations were included in the analyses.

We obtained our estimates of total alcohol consumption per capita (15+ years; liters of pure alcohol per year) for these countries from the World Health Organization, Global Health Observatory data repository, which is available online (World Health Organization [WHO], 2017). A 3-year average value is available for males and females separately for the period 2008–2010. Year specific values are available for all years for the total adult population. We also included estimates for Income (Gross National Income per capita, based on purchasing power parity; data are in current international dollars based on the 2011 ICP round) and for Education (average years of education by people ages 25 and older) to act as control variables. The decision to control for these two variables was empirically driven because they have been found to be associated with alcohol consumption (Grittner et al., 2013; Barr et al., 2016). These values are available for individual years and were extracted from the World Bank Open Data set (The World Bank Group [WBG], 2016) and Human Development Index Data set (United Nations Developmental Programme, 2014), respectively. The average country latitudes were obtained from a public data set which can be retried at https://developers.google.com/public-data/docs/canonical/countries_csv.

Prior to analysis missing data were imputed using Multiple Imputation by Chained Equations (Raghunathan et al., 2001). We then evaluated the relationship between cultural value orientations and alcohol consumption in four stages:

First we conducted partial correlations between cultural value orientations and alcohol consumption for males, females, and total population (2008–2010 period) controlling for the effect of the mean Income and Education for these years. For significance tests, we applied a Holm′s adjustment to correct for multiple comparisons. We interpret the effect size of correlation coefficients using the guidelines described by Cohen (1988).

Secondly, in line with what might be expected from a circumplex model, VIF analyses revealed significant multicollinearity between cultural value orientations. We therefore used linear ridge regression, a technique that uses a shrinkage constant (k) to make regression coefficients more stable in the case of multicollinearity (Ryan, 2008), to determine the amount of unique variance in alcohol consumption that can be accounted for by cultural value orientations. One effect of adding a shrinkage parameter to regression equations is that the estimates of effect size are conservative (Draper and Smith, 1988). We included the same variables in the model as we did in the partial correlations with the addition of the dummy variable Gender and seven Gender-by- cultural value orientation interaction terms.

To assess the sensitivity of our findings we conducted two further sets of partial correlations. Partial correlations were calculated for alcohol consumption and cultural value orientations for three time periods (1990, 2000, and 2010). The effects of Income and Education for the corresponding years were controlled for. Partial correlations for males, females and total sample (once more for the period 2008–2010) were then calculated, but for two subsamples, one containing only European countries (n = 35), the other containing the remainder non-European countries (n = 39). The effects of mean Income and Education for the corresponding years, and corresponding countries, were controlled for.

As will become apparent, our results revealed that Cultural Value Orientations were associated with alcohol consumption across the 74 nations. Our final stage of analysis was an assessment of whether latitude – and by extension a range of ecological factors that have been shown to covary with latitude – might account for the variance between values and alcohol consumption. To do this, 14 statistical mediation analyses were conducted following the proposals of Baron and Kenny (1986). For each analysis, the effect of a Cultural Value Orientation on alcohol consumption (male or female) via the mediating effect of latitude was assessed using a series of regression analyses. To conclude that a latitude is a mediator, the effect between the cultural value and latitude must be significant (path a); the effect between latitude and alcohol consumption must be significant (path b); the total effect of the cultural value on alcohol consumption must be significant (path c); and the effect of the cultural value on alcohol consumption after adjusting for the mediator (path c′) should be smaller than that for path c. We assessed any indirect effects via an inspection of bootstrapped confidence intervals.

Based on World Health Organization defined geographical regions our sample comprised 9 countries from Africa, 12 from the Americas, 4 from the Eastern Mediterranean Region, 35 from Europe, 5 from the South East Asia Region and 9 from the Western Pacific Region. The average Income (GNI) for the period 2008–2010 was $1.08 trillion. For the same years, the mean years of schooling (Education) for countries in our sample was 9.41 years (Range: 2.23 – 12.90). In total, males were shown to consume more liters of pure alcohol per year (M = 12.97, Range: 0.40 – 23.90) than women (M = 4.32, Range: 0.00 – 7.80). In terms of cultural value orientations, Egalitarianism was the most highly valued across all countries (M = 4.68) while Hierarchy was the least (M = 2.35).

Table 1 presents the partial correlations between cultural value orientations and estimates of alcohol consumption for males and females. Inspection of these profiles reveals a similar general pattern for males and females, although some striking differences are evident.

For males there was a significant positive correlation between Harmony and alcohol consumption, and a significant negative correlation between Embeddedness and alcohol consumption. There were weak, and not significant, positive correlations between alcohol consumption and the values of Autonomy. There were also weak negative correlations between Hierarchy, Mastery and Egalitarianism, and alcohol consumption, but none of these were significant.

For females it is evident that the overall pattern of associations was stronger than for males. There were medium-sized, significant positive correlations between Autonomy (Intellectual and Affective) and Harmony with alcohol consumption, and medium-sized, significant negative correlations between Embeddedness and Hierarchy with alcohol consumption. Egalitarianism and Mastery were not significantly correlated with alcohol consumption for females.

The linear ridge regression model supported the above observations (Table 2).

The overall model accounted for a large proportion of variance in alcohol consumption, r² = 0.65, Adjusted r² = 0.60, F(17,130) = 11.91, p < 0.001. Of the control variables, Education, and Gender were found to be significantly and positively associated with alcohol consumption. Income was not significantly associated with alcohol consumption. A number of cultural value orientations accounted for unique variance in the model. Affective and Intellectual Autonomy, and Harmony, were significantly positively associated with alcohol consumption. Embeddedness and Hierarchy were significantly negatively associated with alcohol consumption. Mastery and Egalitarianism showed no significant associated with consumption. All seven interaction terms were found to be significant.

Figure 2 shows the significant gender-by-cultural value interactions plotted on an axis representing standardized alcohol consumption. What these plots therefore show is that when all else is equal, for countries low in all seven values males consume roughly the average level of alcohol for males, and females consume roughly the average levels of alcohol for females. The interactions were apparent when observing gender differences in countries with high levels of each value. For Harmony, Mastery, and Affective and Intellectual Autonomy, in countries with high levels of these values females drank around the average level of alcohol for females (although there was a small increase in consumption), but males drank roughly 0.25 standard deviations more than the mean per year. In contrast, for Embeddedness and Hierarchy, when all else was held equal, males drank similar levels of alcohol in countries with low and high levels of these values, but females drank less alcohol in countries with high levels of these values. For Egalitarianism, compared to countries with low levels of this value, males drank more alcohol in countries with high Egalitarianism, and females drank slightly less alcohol in countries with high Egalitarianism.

Inspection of Table 3 reveals that the patterns of associations between cultural value orientations and alcohol consumption at three time points were similar, although over time the strength of associations was weakening. Furthermore, these patterns are consistent with the overall pattern of associations shown in Table 1.

Table 4 presents Pearson correlations for males and females (2008–2010 values) for two subsets; 35 European nations and 39 non-European nations. The general pattern of associations between alcohol consumption and cultural values was similar between the subsamples, and to the results presented in Table 1. This was particularly the case for Harmony (medium positive), Embeddedness (medium negative), Mastery (negligible) and Affective Autonomy (weak positive). For Hierarchy, European countries showed a weaker negative correlation (r = -0.07) compared to non-European countries (r = -0.23). For Intellectual Autonomy, European countries showed a stronger, but still weak, positive correlation (r = 0.16) than non-European countries (r = 0.08), and for Egalitarianism European countries had a medium positive correlation (r = 0.30), while non-European countries showed little evidence of any association (r = 0.02).

Some clear differences were observed when inspecting consumption for males and females separately. Firstly, while males in the European subsample had similar correlations to those shown in Table 1, males in the non-European subsample showed very little association between alcohol consumption and the cultural values. The exception was for Egalitarianism, which showed a moderate negative correlation (r = -0.42) which differed from the European subsample, where males showed a positive correlation (r = 0.28). For females the correlations were remarkably similar for European and non-European subsamples, although an exception was Egalitarianism, which showed little association with alcohol consumption in non-European countries (r = 0.04), but a medium positive correlation in European countries (r = 0.37).

As a primary step to understanding the relationship between latitude and alcohol consumption, Pearson’s correlations were calculated between latitude and male and female alcohol consumption. Significant positive correlations were found for male, r = 0.35, p = 0.002, and female alcohol consumption, r = 0.37, p = 0.001. Thus there appears to be evidence of a north–south consumption gradient.

The complete output from the 14 statistical mediation analyses are presented in Supplementary Tables S1–S7. Mastery and Egalitarianism did not present all the significant effects, as defined by Baron and Kenny (1986), for the effect of these values on alcohol consumption to be mediated by latitude. In all cases, the Sobel test did not indicate a significant statistical mediation effect. However, for Harmony (c′_male = 0.095, [CI 0.022, 0.217]; c′_female = 0.081, CI [0.011, 0.200]), Hierarchy in males (c′_male = -0.094, [CI -0.216, -0.011]), and Affective Autonomy (c′_male = 0.077, [CI 0.006, 0.213]; c′_female = 0.066, CI [0.001, 0.192]), and inspection of the 95% confidence intervals suggested that latitude was able to account for a partial but meaningful amount of the effect of these values on alcohol consumption. Embeddedness (c′_male = -0.072, [CI -0.199, 0.009]; c′_female = -0.049, CI [-0.167, 0.011]), and Intellectual Autonomy (c′_male = 0.098, [CI -0.019, 0.246]; c′_female = 0.047, CI [-0.055, 0.178]), and Hierarchy for females only (c′_female = -0.069, CI [-0.189, 0.001]) did not show evidence of meaningful partial mediation effects.

The finding that Harmony was positively associated with alcohol consumption needs further explanation because the neighboring cultural value orientation, Embeddedness, was negatively correlated with it. Based on the motivational synergies proposed by Schwartz, we implicitly expected Harmony to be unrelated or slightly negatively related to alcohol consumption because the underlying motivational synergies are related to Embeddedness, which was, as predicted, negatively correlated with alcohol consumption. We believe that this can be explained with the definition of Harmony that “emphasizes fitting into the world as it is, trying to understand and appreciate rather than to change” (Schwartz, 2006, p. 141). In line with this, social motives related to Harmony such as having a good time with friends have been found to be predictors of alcohol consumption (Kuntsche et al., 2005). Furthermore, research that has looked into the relationship between cultural value orientations and peacefulness related variables, such as violent inequality, has also found that the sign of the correlation coefficients was reversed for Embeddedness and Harmony (Basabe and Valencia, 2007). This suggests that the underlying motivation for at least those two adjacent cultural value orientations can also be conflicting in relation to specific external variables.

Our results indicated some interesting gender effects. Firstly, the partial correlations between cultural values and alcohol consumption were stronger for women than for men, suggesting that cultural values are a more reliable predictor of alcohol consumption in women than for men. Secondly, the gender-by-cultural value interactions were all found to be significant. These indicated that alcohol consumption in women appeared to differ only slightly, all else being equal, between countries with high and low values of each value. Nonetheless, despite the small changes in consumption, it was evident that female consumption was higher in countries that highly valued Autonomy, and lower in countries that highly valued Embeddedness. Men tended to consume more alcohol in countries with higher levels of each value, the exceptions being Hierarchy and Embeddedness, for which the differences were small. These findings are in line with what is expected from opposing cultural values and supports the relevance of the Autonomy-Embeddedness dimension for determining cultural alcohol consumption (the Harmony-Mastery and Egalitarianism-Hierarchy dimensions showed similar patterns of results for each of the values). They also suggest that Embeddedness and its adjacent value Hierarchy, which are more highly associated with less gender equality (Schwartz, 2006) and negative attitudes about women drinking alcohol (Pretorious et al., 2009), are relevant cultural values to consider in accounting for gender differences in alcohol consumption.

A broad aim of international organizations concerned with public health is to collect and use data to inform national governments on how best to tackle health issues. In relation to alcohol, the WHO in particular has made a continued effort to collect information and to use this to help nations reduce the harmful use of alcohol and its consequences (World Health Organization [WHO], 2014a). Although the WHO uses considerable amounts of information to help with this process, we argue that cultural information such as a country’s value profile may well serve as a useful tool for helping to inform policy. At a basic level, our results suggest that the WHO should prioritize tackling alcohol consumption in countries that are high in Autonomy and Harmony, and low in Embeddedness and Hierarchy. At the very least, our results indicate that future research should be directed at further understanding the causal mechanisms between cultural values and alcohol consumption, and their interaction with gender, the results of which could be used to tailor intervention programs that are culturally sensitive.

Importantly, alcohol consumption is just one of four modifiable behavioral risk factors of NCDs, which are responsible for 70% of global deaths (World Health Organization [WHO], 2014b). The other risk factors are smoking, physical inactivity, and diet. Having shed some light on the cultural underpinnings of alcohol consumption, an obvious next step is to create similar profiles and models to understand the cultural underpinnings of these other risk factors. When combined, this information may prove to be highly impactful on how governments opt to implement health policies aimed at reducing the prevalence of NCDs.

The results of our study are interesting because our variables – cultural value orientations, Education, Income and Alcohol consumption– were measured independently and are based on different sources (self-report vs. official statistics). This rules methodological bias out as a hidden moderator – a problem which many studies that rely on self-reports face (Richardson et al., 2009). Also, the data we used is highly reliable and stable, given that it relies on around 40,000 participants, whose responses were used to estimate the 74 country averages of the cultural value orientations.

A test of the validity revealed a similar overall CVO profile for three different time points over 20 years and for the European subset of countries for men and women for the time period 2008–2010: These findings suggest firstly that the identified pattern of associations between cultural value orientations and alcohol is not specific to a single time point, although it appears that the overall association between cultural values and alcohol is decreasing: likely a result of converging levels of alcohol consumption across countries (Room and Mäkelä, 2000). This is interesting since the global trend and current projections for 2025 are an increase in alcohol consumption (World Health Organization [WHO], 2014a). At a regional level, however, there are some differences, with Europe showing a trend for decreases in consumption.

The pattern of associations was replicated with a European subsample of our data. Inspecting the total levels of alcohol consumption, the pattern of correlations between the European sample and the total sample (Table 1) were, with the exception of Hierarchy, remarkably similar. Hierarchy had a weaker negative correlation in the European sample than the total sample. The non-European sample was also fairly similar to the total sample, particularly for Harmony, Embeddedness, Hierarchy, Mastery and Affective Autonomy.

In our analyses we controlled for two variables that have been shown to be associated with alcohol consumption; income and education (Grittner et al., 2013; Barr et al., 2016). Of course, other variables such as the average age of a population, the price of alcohol, national alcohol laws/policies, degree of urbanization etc. also influence alcohol consumption and were not controlled for. These variables should be considered in future research. Since the number of predictors/variables determines the sample size required for sufficiently powered statistical analyses, including additional covariates in future research will require multi-level data from a larger sample of countries. Further, although not reported, our analyses revealed that cultural value orientations were correlated with income and education. Thus it is likely that our analysis controlled out some of the cultural variance in alcohol consumption.

When examining the association between cultural value orientations and alcohol consumption at different years we used the same CVO values and so it might be argued that our results do not entirely accurately reflect the time specific associations. However, cultural value orientations are claimed to be stable over time (Schwartz et al., 2000; Hofstede, 2001; Manfredo et al., 2017; Tormos et al., 2017) and this suggests that over a relatively short period of time using the same values of cultural value orientations is legitimate.

While the associations between total alcohol consumption and values for European and non-European countries were similar for Harmony, Embeddedness, Mastery and Affective Autonomy, some notable differences were evident between the European subsample and non-European subsample. Firstly, with the exception of Egalitarianism, alcohol consumption in males from non-European countries showed little evidence of being correlated with cultural value orientations. In this sense, the pattern of associations for males displayed in Table 1 can be said to be driven by European countries. In other words, while encultured drinking levels by men in Europe is associated with cultural values, outside of Europe these values are unrelated to consumption. Why this might be, lies beyond the scope of this discussion. For women, the associations between consumption and values were, more or less, consistent between the subsamples, although of note was that Egalitarianism was more strongly associated with alcohol consumption in European countries than non-European countries. This highlights the complexity of the relationship between values and alcohol consumption, specifically demonstrating that beyond the presented two-way gender-by-value interactions, there may also be three-way region-by-gender-by-value interactions. Future research, incorporating larger data-sets should consider this complex interplay of variables.

In addition to investigating the interplay between cultural values, alcohol consumption, and variables such as religiosity, we suggest that future research should look into causal relations between cultural value orientations and alcohol consumption, using a multi-level longitudinal design. Prior pseudo-longitudinal work using cultural value orientations to predict socioeconomic development and democratization has found that cultural value orientations predict said national variable rather than vice versa (Schwartz, 2006). Also, a longitudinal study has found that values and value-expressive behaviors influence each other reciprocally, but values tended to have a stronger influence on behavior then behavior on values (Vecchione et al., 2015). We are therefore expecting that cultural value orientations influence alcohol consumption directly, but that alcohol consumption can also influence cultural value orientations through indirect routes such as conflict and increased levels of psychopathology: both values and alcohol consumption have been found to be related to or causing aggression and depression (Hoaken and Stewart, 2003; Boden and Fergusson, 2011; Hanel and Wolfradt, 2016; Souchon et al., submitted). A multi-level design would allow us to test whether the same pattern observed on a country level can also be found on an individual level consistently in each country.