This is how educational level, unemployment and income affected the results in the Finnish EU-election 2019

Category: Best visualization (small and large newsrooms)

Country/area: Finland

Organisation: Svenska Yle and Yle

Organisation size: Big

Publication date: 28/05/2019

Credit: Eemeli Martti, Linus Lång, Juha Rissanen, Malin Ekholm, Janne Toivonen, Juho Salminen

Project description:

A scrollytell-vizualisation that shows how demographic factors affected the election result in the EU-parliament election in May 2019. 

The user can choose between exploring the data themselves or read the biggest findings in our scrolly tell. 

This vizualisation for example shows that the success of the populist party cannot be explained by traditional left-right theories. The areas where the populist party won were not defined by income, age or unemployment but rather by voter turnout, education level and property ownership.

Impact reached:

Combining the voting data with demographic data for such small areas (most polling areas has a population of about 3000 people) gave us the possibility to make indepth-political analysis of the voting behaviour. We tried the concept after the genral election in April 2019 and tweeked the visualization for the EU election in May.

Many academics commented that the stories we did brought new light on how Finns vote and how voting still is somewhat related to social class. Most of all, our stories delivered new fact based information to the elections related public discussion. There is never too much of that around.

Techniques/technologies used:

Because this information was not available we had to gather it ourselves. First we got the geographical info about the polling districts (which we also used for visualisations in our election result service on election night). Then we asked the statistics official in Finland to combine the polling districts to statistical grid database (250 m x 250 m). 

We also had two researchers who know this field very well help us with selecting relevant data and also analyzing the findings.

We had two data teams working with the project, from Finnish speaking and Swedish speaking departments
The project consists of 4 published stories: 2 in Finnish and 2 in Swedish:

For an initial analysis of the data, we utilized R. The data consisted of more than 100 demographic variables for each polling area. The demographic data was combined with the results data at party-level, and scatter plots describing result and demography were created for each variable and party. Visually examining the scatter plots, we could quickly identify correlational patterns between demography and support for particular parties. A handful of variables were chosen for the visualization. Population densities for polling areas were calculated with QGIS (only for the European Parliament version of the viz).

What was the hardest part of this project?

– The demography data did not include all variables that might have shown interesting correlational patterns. For example, we did not have data the about mother tongue (Finnish/Swedish/other) or the immigrational status/share of foreign-born among the residents.
– Front-end animations and responsive grids on highly deadline sensitive story.
– We would have liked to publish all the material used for the stories as open data but we couldn’t because of the contract with Finnish statistics official from whom we got/bought the data.

What can others learn from this project?

Demographical data can be used not only to describe a certain area but also to write stories about society. Whebn it comes ti electionresults it was essential that we could get demographic data for the exact geographical areas that were the polling areas. When combining datasets in this matter you should have an open mind about what the findings will be.

The dataset also provided us with backgroundstories that could be used for other projects later. 

Project links: