Beers Across the Atlantic

Decoding Beer Preferences in North America and Europe

Data sources
Naive analysis
Beer styles
Time evolution


Why is there a seasonality trend between Europe and North America?

We will try to answer this question by first decomposing the time series for each region into seasonality and trend.

Seasonality; NA vs EU

From the graph above, we can conclude the following points:

Question: What are the suspects which are the reason for seasonality?

Oktoberfest Suspect 1

We suspect that the difference in the seasonality pattern of American and European ratings might be due to significant beer events that take place in Europe, such as Oktoberfest, which is, in fact, the world’s largest beer festival and takes place in late September and the first weekend of October.

Fresh and light during Summertime Suspect 2

Have you ever relaxed at a beach in summer, with your sun glasses on, sipping gently on your 12% Russian Imperial stout? Yes… Me neither. We suspect that people tend to drink lighter beers overall during hot summer days and rather prefer stronger kinds of beer during the colder parts of the year. This does not seem too unlikely, given that the seasonality pattern in both continents show higher ABV during winter periods on average than during summer!

To be sure, we will take a closer look at the trend and the seasonality (including residuals):


When keeping the residuals, the second peak in Europe looks more like a plateau, where North America seems to drop more quickly.

St. Patrick’s Day Suspect 3

Have you ever heard of St. Patrick's Day?

- It is an Irish traditional holiday, taking place in March. This might be the reason for the second peak appearing in spring!
St. Patrick

Let’s address both of the suspects

Since Guinness, and Stouts in general, are rather on the strong side of beer spectrum, an increase in consumption could easily spark a peak in the ABV pattern.

In a simple approach to investigate this, we decided to decompose the seasonality of each beer style into Fourier modes by running a simple discrete Fourier transform over it. Naturally converting from frequency space to periodicity space allows to easily filter beer styles that show a significant mode corresponding to a 12-month period.

Comparing the amplitudes and accounting for the popularity of the beer style, in addition to extracting the phase shift of the peaks, allows us to gain deeper insight on what the seasonality is composed of.

After some filtering for months that contribute to seasonality, a normalised Fourier transform would look like this:

Normalised FT of filtered beer styles

We can summarise all the information, including the phase shift, which equals the peak ABV of the oscillation here in a fancy graph:

Peak Seasonality Plot

Let us recall our initial trends, highlighting the months of May and October:

Seasonality highlighted in May and October