Alcohol Policies - ADA Project

Why is this important?

Harmful use of alcohol causes more than 5% of the global disease burden. The death of more than 3 million people a year is caused by the substance. Simultaneously it is source of a 1344 billion dollar industry, and an important cultural factor. This makes it a highly important, and highly complex, matter for regulators which have to take into account many aspects when creating policies. Alcohol control policies exist since many years, from taxes to total bans. The World Health Organization (WHO) has gathered significant amount of data on both alcohol consumption and related policies in different countries. This data can be used to evaluate the effectiveness of the policies and later policymakers can use it to make informed decisions.

Research Questions

What kind of legislation is the most common in which type of country?
Which policies correlate the most with low alcohol consumption?
Does policies or socio-economic factor explain better different levels of alcohol consumption?

Analysis of alcohol policies

From the data from the World Health Organization allows one can identify three major means frequently used by states in their politics of alcohol consumption limitations and reduction:
1. charge money for alcohol production, retail and consumption. This is an interesting feature since the effectiveness of tax policy interventions for reducing excessive alcohol consumption is subjected to a lot of studies. Results tend to show a strong positive impact, which according to this study is proportional to the magnitude of the taxation.
2. restrict access to alcohol in public places (e.g. in public transport, workplaces, places of worship, or during sporting events, etc)
3. raising awareness of public health problems caused by harmful use of alcohol and ensuring support for effective alcohol policies. This is the soft approach since it aims to convince or persuade individuals to limit their own alcohol consumption. In practice, the government enforces laws that constraints alcohol companies and alcohol services in their relation with the clients (A known example are the health warning labels on alcohol advertisements) We can observe that in the world the most common alcohol control policy is to charge the consumer an extra fee. This makes sense as it is the easiest to enforce also for lower-income countries, who do not have the budget to have government programmes to reduce alcohol consumption We look a bit closer inside each category

Policies by regions

Observation: We observe that in the world the most common way to control alcohol consumption is through charging taxes at different scales which range from production to consumption. In fact, it is almost as important as the sum of the frequency of usage of the two other features (0.81 for Charges and 0.89 for Awareness and Restrictions). Policies linked to raising awareness of the consumer is the least used of all the means.
Explanation: Charges is one of the easiest policies to enforce since countries need not use extensive human resources to support the enacted law. Hence, it makes sense especially for lower-income countries which do not have the budget to launch governmental programmes to reduce alcohol consumption. Taxation policies are however highly subjected to critics because they are a source of income for the government. The line between security driven politics and money driven politics is often blurry, in particular in countries subjected to lobbying and corruption. Awareness probably is less popular due to the high cost of enforcement and only developed governments can afford it. The effectiveness of this method is hard to assess since psychological leverages have an important latency. It involves processes which influence the mentalities on the long term. While enacting a taxation policy or a new restriction of alcohol usage is straightforward, awareness procedures demand an actual campaign costly in time and money.
Special attention to: A possible explanation of this plot is the need of human resources. Indeed, when it comes to decision making, private and public actors are all constraints by a budget. Even in governments, where the security and health of the population should be a major concern, money is still a cross cutting issue and a high lever. Knowing that in any project a major expenditure is human resources. We can therefore reasonably carry that the more civil servants will be needed to set up the plan, the less this plan is likely to be chosen. This theory seem to adapt quite well to the data we have here.

Policies by religion

We can observe that the popularity of different kinds of policies is similar for certain regions. The hypothesis made above about awareness being a policy put forward by developed countries does not stand as Western Europe has little difference with the other region in this category

We can observe a few differences between countries with different religions

Islamic and officially atheist (China) coutries have high number of regulations. The former due to the ban of alcohol imposed by the religion, the second due to the autocratic organization of the state
Christian groups, except Roman Catholic and Orthodox, seem to have fewer restriction policies compared to other countries.
Orthodox and non-specified Christian countries have the lowest usage of taxes on alcoholic beverages. This can be due to the high cultural importance of alcohol in many post-soviet nations which are also orthodox

Influence factors

Several things influence alcohol consumption. We can initially think that the religion or the region in which a country is highly influences the alcohol consumption. We can see how these two factors compare to alcohol consumption in the following images.

We try to find a more accurate way to separate countries in categories and understand what influences alcohol consumption. We have several alcohol control policies and socio-economic policies that probably influence alcohol policies. We have two ways to assess the influence: train a regression model and see the weights of each feature and see the statistical correlation of each feature with alcohol consumption. ''

Even though the weights are less likely to accentuate on features that strongly correlate between themselves, they can also be less reasonable, especially if the model did not have much data to learn from, as it is in our case. For instance, the restrictions on alcohol use in parks and streets should certainly have a negative effect on alcohol consumption, but our model would put a larger weight on some other similar features (i.e. on use in leisure events) and use the given feature only for somewhat unreasonable adjustments due to correlations between said features. For that reason, we are going to use correlations with alcohol consumption directly, instead of the generated weights, even though they give some redundant information due to inter-correlations.

From the most important features chosen through the correlation we create a clustering to divide countries. Now we can visually see if this clustering matches with the alcohol consumption

Case studies

In this section we will do a more qualitative analysis of alcohol consumption in several countries and what seems to influence it. We start with some insights coming from our alcohol consumption dataset in which we analyse the coutries with the biggest change in alcohol consumption between the year 2010 and 2015.

It is quite surprising to see that 39 countries where subjected to a very high variation of their alcohol consumption. A great increase can be seen in Cambodia, Latvia and Uruguay that we analyse above. Estonia is additionally analyzed as it shows an important increase and is closely related to the increase in Latvia Many countries also experience an important decrease. In particular, while looking at the horizontal bar plot above, Ukraine is a country which got the third highest drop of alcohol consumption in 6 years: -3.5 Liters. Lead by his curiosity, the ukranian group member decided to analyse closely his country to understand the reasons behind this variation.

Latvia & Estonia

Latvia passed from 9.9 to 11.1 litres of alcohol per capita from 2010 to 2015. Estonia in this same period passed from 15.2 to 15.8. For both these neighbour countries this trend started in 2009, but ended for Estonia in 2013 for then again decrease, while in Latvia the consumption continued to rise. These changes can have many explanations. Just before 2009 we see a significant drop of GDP for both countries, of which economies alcohol consumption is contributor. It is very common for Estonians to buy alcohol in Latvia as it is cheaper, and for Finns to buy it in Estonia for the same reason. It seems that during these years there has been a game to the bottom between Estonia and Latvia on alcohol taxes, which until 2013 brought prices down significantly. In Estonia a new alcohol strategy was adopted in 2013, increasing tax rates and raising awareness. Latvia also implemented an action plan in 2012-2014, but it didn’t contain any action on alcohol taxation. We see the median age growing, but having no correlation with changes in alcohol consumption. The Education Index remains stable in Estonia, while we can observe a decrease in Latvia between the years 2008 and 2012.

Uruguay

Uruguay passed from 8 to 9 litres per capita from 2010 to 2015. Looking more into the consumption data provided by the WHO we see that this is a trend started in 2003 and accelerated in 2009. We do not find online many information available in English on Uruguay alcohol policy history, but from the GDP data it seems that the economy started growing again in 2003, at the same time as the alcohol consumption started to rise again. The Education Index dropped significantly in 2009 and got back only in 2014 to its previous level, so this could be correlated with the alcohol consumption. We see no correlation with the median age which continues to rise slowly in this period.

Cambodia

Cambodia passed from 2 to 3.5 litres per capita from 2010 to 2015, the largest increase worldwide for this period of time, although total consumption remains low compared to many countries. This country seems to have a big problem related to alcohol advertising target towards youth and for the moment is has very limited policies to regulate alcohol consumption and advertising in general. For example small and informal producers are completely excempt of any taxation framework This problem seems highly entagled with economic incentives of the advertising industry for which alcohol accounts for more than 50 percent of all advertising revenues in the country . Relating alcohol consumption in these years to GDP seems difficult to make any relevant observation as it increases between 2010 and 2012, but drops again after that. Median age rises slowly and steadily, showing no correlation with alcohol consumption and the Education Index rises from 2010.

We now want to analyse the high decrease of alcohol consumption. As one team-member is from Ukraine it offers us the possibility to analyse well it's alcohol policy. We take the occasion to analyse the countries of the rest of our fellow team members: China, France and Switzerland.

China

France

Switzerland

Ukraine

Limitations

The initial goal of the project is was to make an observational study and discover which alcohol control policies are the most effective. This was not possible due to several reasons. The first hurdle was the high intercorrelation of different variables. As can be seen in the heatmap below many of the features have fairly high correlation with alcohol consumption (eg. GDP, Education index and HDI), the strong correlation among themselves mean that there is some redundancy in the information they provide us with. Partially for this reason, the previously generated weights for GDP, for example, are quite low, despite it having one of the highest correlations with the alcohol consumption.

We can also think it in terms of the following causal diagram, which illustrates how some socioeconomic-factors are linked to the government programs policies

These highly correlated variables would need to control for many factors to draw any valid conclusions. The WHO alcohol control policy dataset is limited to the year 2016, which unfortunately does not allow to control for all these parameters, as there are too few datapoints. Another feature that is important for a government when implementing a policy is to consider the cost of deploying it. In the dataset found we have no information about the this parameter, so we cannot give information about cost-effectiveness of policies. Lastly we noticed that several policies target a specific group of people, so even if effective we cannot see a variation in the total alcohol consumption. For example if a policy targets alcoholics, even if policy successful it won’t be observable in the national alcohol consumption as it is too small of a share of the population

Conclusion

In this work we discover the importance of socio-economic factors on the alcohol consumption of a country. Median age, the Education Index and slightly less GDP might have a causal effect to increase alcohol consumption. To make further investigation and draw more conclusive results access to more data of alcohol control policies is needed. In the qualitative analysis we have discovered that there the variations of alcohol consumption within a country are affected the most by changes in the tax rate of alcoholic beverages, which themselves are often affected by socio-economic factors for political reasons.