This code is the result of a Civic Tech hackathon in Göteborg in September 2018 🤓. It was written by Emil Hemdal, Albin and Pierre Mesure.

What does it do? 🖨

This is a script that scraps data from regeringen.se. For now, it only looks for remisser, lists them and builds a dataset containing information about these remissprocesses.

You can read more on how to run the script and which data it saves.

The result is saved in an SQL database.

Why? 📕📗📘

Regeringskansliet, the Swedish government's chancellery, has very little open data. Theoretically, all of their documents are available publicly on regeringen.se, but a weak search engine (which doesn't search for keywords through the content of the page or documents) and the lack of structured data (most of the documents are available as PDF with various formattings) make it virtually impossible for journalists, researchers and activists from civil society to delve in this information and scrutinize the government's actions.

Apart from this valuable work by counterpowers, this lack of open structured data also prevents civil servants from working efficiently and other public agencies to use the information released.

The Swedish parliament, Riksdagen, includes some of the government's documents on their website (bills, SOUs) but they get this as PDF from Regeringskansliet and have to parse the content to display it, making it harder for them to inform citizens.

The end result is impossible to use in a good way. Formatting is lost, page numbers and tables get inserted in the text and no structure is preserved. See example here.

Goal 🎯

Our goal with this small script is to show how ridiculously hard it is to find simple information such as:

• which organisations does the government ask?
• which organisations answer?
• how many times have they asked each organisation and how many times do they answer?

Statistical information would already help counterpowers to say who the government is picking to help them to shape a bill. Big interest groups? NGOs? Local associations? These are important democratic issues.

And that would only require a structured list of remisser and remisssvar.

If Regeringskansliet was also publishing remissvar's content in a structured format, techniques such as semantic analysis could be applied to analyze the proximity of a remissvar and a bill to determine which organisations influenced most. This is already done to study the impact of lobbying in the EU and the US parliaments.

First results 📊

EDIT: The script has now been improved and you can find better updated results in this spreadsheet. The hackathon was just one day and we realised while doing it that the quality of the "data" was poorer than expected:

• regeringen.se only shows the last 894 remisser, the oldest one being from the 02/04/2015. We harvested the name and link of 23353 attached files.
• we used the file names as the name of the organisation (for lack of something better) and although that worked most of the time, a lot of names include a typo (Stockholm universitet instead of Stockholms universitet), a number (35. Stockholms universitet) or something else which requires manual cleaning of the data.
• we had no choice but to group local branches of national organisations (LO Dalarna becoming LO) and suborganisations (Sverigesingenjörer, Sveriges Läkarförbund, etc. becoming SACO, which explains the high number for them as they usually send their answers separately for each remiss)

However, we managed to get a shortlist of organisations sorted by which answers the most to remisser, and we filtered it manually to show only interest groups:

🏆

You can run the script yourself to get the raw data or find it in this Google spreadsheet. Don't hesitate to use it to make your own rankings and visualisations.

What does that show us? 🧐

There's really no conclusion we can jump to by seeing that Svenskt näringsliv answers to more remisser than LO. But as basic as this information can be, the fact that you have to know how to code and spend a Sunday in front of a screen to access it shows us a lack of transparency.

In just one day, we didn't manage to extract, clean and restructure enough data to make conclusions on which organisations the government listens to most. But we hope it will motivate others to dive into more of this data and create more robust models to answer that question and more!

How can I contribute? 🙌

Did we manage to get your attention with this small demo? Do you want to see more insights on all the information that Regeringskansliet has but doesn't publish in a usable format?

There are 3 ways you can help:

• if you can script, fork this project and try to fetch and analyse more data from regeringen.se! Write some code to clean the data or to visualise it! 👩🏽‍💻
• if you can't but have a profession/occupation which would benefit from having this data available (journalist, politician, researcher), contact the government to ask them to release it! Explain to them what you could do with it. 👨🏻‍⚕️
• ask the government to reform its offentlighetsprincip to require that any piece of public data be available online in a structured format. Canada, the UK, Germany or France are doing it, it's time for Sweden to catch up! 🙋🏻‍♀️

And don't hesitate to contact us, we love hearing from opengov enthusiasts! ❤️