Parsing the European Parliament elections -… | DataJournalism.com

Parsing the European Parliament elections

Conversations with Data: #28

Do you want to receive Conversations with Data? Subscribe

Conversations W Data header

Come election day, voters are all asked one crucial question: Who are you going to vote for? But behind every vote, there’s plenty more. What does this candidate stand for? Which party do I align with most? And, regrettably in some elections, will my vote even count?

Last weekend, up to 400 million Europeans took to the polls in the 2019 European Parliament elections to answer these questions for themselves. And, as we found out, there were many data journalists working to inform their vote in unique ways. As the results roll in, we took a look at 5 ways that data journalists have parsed the European Parliament election.

1. What does the election look like for me?

This one is key for any voter. In the UK, BBC Online News helped Bristons look beyond a (possibly) impending Brexit, putting together a region-by-region rundown on the country’s candidates. Robert England gave us a behind the scenes look:

“To gather the data we scraped publicly available lists of MEPs and candidates to find out which areas had candidates seeking re-election and which had candidates running for the first time -- potentially facing a very short political career in the EU. We did this by cross referencing the two groups, candidates and sitting MEPs, in Excel to show not only which areas should expect the most new faces, but also which regions had the widest/smallest choice of political parties to vote for.”

“The issue of MEP pay and benefits, a subject extensively covered nationally, was another area we wanted to personalise. Using the UK Office for National Statistics earnings estimates, we broke the question of pay down to a regional level by comparing the wage of MEPs against average wage estimates,” he explained.

106996961 meppay2 nc

BBC.

2. ...So, what about Brexit?

Speaking of the UK, results are now showing that the Brexit Party has dominated the country’s vote. Who would’ve thought? Well, it turns out Ashley Kirk of The Telegraph did. They published a suite of analysis pieces, with stories including a comparative analysis of all pro-Remain parties versus all pro-Brexit parties in the UK, a revelation that Theresa May's Conservative Party could slump to their lowest vote share since they were formed in 1834, and a look into how establishment parties across the EU are forecast to lose seats.

“When it comes to presenting these, we've paid particular attention to how we can help our readers understand each EU-wide bloc in the Parliament and where they sit on the political spectrum. On a UK level, we have been analysing how the two establishment parties are struggling to hold their ground, with the advance of the Brexit Party,” he wrote.

The Telegraph

The Telegraph’s polling analysis, indicating a high chance of Brexit Party votes.

3. What can we do to prevent filter bubbles?

From Brexit to filter bubbles (or is it the other way around?), the folks over at Talking Europe came up with an ingenious way to break through echo chambers and encourage dialogues between voters of different political persuasions. The platform matches people in Europe, each with different political opinions, for one-on-one chats about current issues.

Malte Steuber talked us through their artificial intelligence (AI) techniques, which leverage a matching algorithm based on answers to five yes/no political statements. “Users then get matched to somebody living in another European country and, in the best-case, who provided five different answers,” he said.

But that’s not the only AI they use: “To make it possible for everybody to chat in his/her native language we use DeepL technology to translate the chats in real-time. Users can write and read in their chosen language and select whether they want to see the translation.”

4. How popular are the populists?

According to Bloomberg’s Andre Tartar, “the big question going into the EU elections was whether populists would win enough votes to make a difference”. And there was no shortage of data teams looking into it.

For Andre’s part, they charted the combined populist seat-share rising steadily to 21.7% in 2014 and forecast to reach 29% this time around -- but their analysis wasn’t all good news for populists.

“The flip side of the story is the historic disorganisation of populists and their difficulty holding on to members. To show this we anchored the piece around a scrolling graphic of the political chamber (called the hemicycle) produced using D3. We first experimented with different representations of this, including a 3D model, and zoom on scroll, and how we could use it to walk readers through the story,” he said.

Over at SPIEGEL ONLINE, Marcel Pauly also conducted an analysis of populism in Europe (in English here), showing how populist parties have performed in national parliamentary elections over the past two decades.

Pop screenshot

SPIEGEL ONLINE.

Here’s a little more about how they did it: “I got the election results from ParlGov, a nice database with data on parliaments and governments for ‘all EU and most OECD democracies’. For the parties’ classifications I used The PopuList, an academic ‘overview of populist, far right, far left and Eurosceptic parties in Europe’. It was initiated by The Guardian, whose data team kindly provided me with a lookup table to merge the two datasets on The PopuList’s party names and ParlGov’s party IDs.” You can also recreate or reuse Marcel’s preprocessing and data analysis, available on GitHub here.

But, of course, the past doesn’t equal the present. And polling data is in constant flux in the lead-up to an election. To capture gains by populists and other parties, Sweden’s Newsworthy put out a monthly, automated analysis of European polls.

It wasn’t easy though. As Jens Finnäs told us, “just categorising the most likely parliament group of each national party is a very challenging task that needs continuous updates”. Their report was only “made possible by the polling aggregation from Politico (formerly pollofpolls.eu)” and their “huge data cleaning and refining effort”, on top of which Newsworthy laid their analyses.

5. Who will win? (c’mon, we had to!)

When Camille Borrett and Moritz Laurer, Co-founders of European Elections Stats, first asked this question a year ago, they were faced with the sad reality that there was little interest or money for pan-European polls.

Their solution: “we wrote several scripts in R which (1) scrape national polls from open data sources for all 28 member states; (2) calculate four seat projections for the European Parliament based on these polls and different political scenarios; (3) upload all the raw data into an Open Data Hub, where anyone can explore and download our data; (4) visualise the data to allow for easy comparison of political scenarios.”

EP2019 Seat Projections R Shiny

See European Elections Stats for the interactive R-Shiny application.

Now that the elections are over, they will continue to aggregate national polls and use them to calculate an aggregate index of the pan-European political mood. Watch this space for updates.

Our next conversation

While our eyes were focussed on the 400 million voters in Europe, Eva Constantaras reminded us of the important work being done to inform elections in other parts of the world. Earlier this month, for example, journalists in India worked hard to inform the nation’s 900 million (!) eligible voters. Using data, reporters revealed how many promises were kept since the last election, visualised all 8049 candidates, and more. Inspired by their work, we’re keen showcase the work of other data journalists from all over the world. To help, we’ll have Eva with us for an ‘ask me anything’ in our next edition.

Eva is an investigative trainer, with teams in Pakistan, Afghanistan, and Kenya, and author of Data Journalism By, About and For Marginalised Communities in the Data Journalism Handbook 2. Submit your questions here.

As always, don’t forget to let us know what you’d like us to feature in our future editions.

Until next time,

Madolyn from the EJC Data team

subscribe figure