"Namentliche Abstimmungen" in the Bundestag

How do individual members of the federal German parliament (Bundestag) vote in "Namentliche Abstimmungen" (roll call votes)? How does the individual align with the different political parties? And how may the members vote on upcoming bills? The bundestag cli provides tools to assist to answer those questions by providing tools download and transform the required data.

Data sources

The German parliament makes roll call votes available as XLSX / XLS files (and PDFs ¯\_(ツ)_/¯ ) here: https://www.bundestag.de/parlament/plenum/abstimmung/liste.

The NGO abgeordnetenwatch provides an open API for a variety of related data. They also provide a great way of inspecting the voting behavior of members of parliament and their (non-)responses to question asked by the public.

Insights from the collected data

"Fraktionszwang"

Do all the members of a party always follow the party line? Clearly not. But that "discipline" is similar across parties. The significant deviation are the factionless, as measured here using Shannon entropy. The curious mind could even estimate the energy it takes to enforce the disciplines.

Embedded members of parliament

As a side effect of trying to predict the vote of individual members of parliament, we can obtain embeddings for each member. Doing so for the 2017-2021 legislative period, we find that they cluster into governing coalition (CDU/CSU & SPD) and the opposition:

If you want to see more check out this site or this notebook.

How to install

pip install bundestag

or

uv add bundestag

to get access to the cli to download and transform bundestag roll call votes data from bundestag.de or abgeordnetenwatch.de.

By to keep things light the machine learning dependencies are made optional. If you want to get those and related functionality as well run

pip install bundestag[ml]

or

uv add bundestag[ml]

For development

git clone https://github.com/eschmidt42/bundestag
cd bundestag
make install-dev-env

How to use

The bundestag cli

A tool to assist with the data processing.

For an overview over commands run

bundestag --help

To get preprocessed data simply run

bundestag download huggingface

To download data from abgeordnetenwatch, for a specific legislature id

bundestag download abgeordnetenwatch 132

To transform the downloaded data run

bundestag transform abgeordnetenwatch 132

To find out the legislature id for the current Bundestag, visit abgeordnetenwatch.de and click on the "Open Data" button at the bottom of the page.

To download data from bundestag.de

bundestag download bundestag-sheets --do-create-xlsx-uris-json

Note that this is using selenium and is therefore starting a browser. Currently this is not using a headless browser so it is easy to see when the scraping should be broken.

To transform the downloaded data run

bundestag transform bundestag-sheet --sheet-source=json_file

Note: If you run the extraction for the legislature with id 132, i.e.

bundestag transform abgeordnetenwatch 132

the data is damaged for some reason. To fix it run

uv run python scripts/fix_empty_fraction.py

download commands will store artefacts in ./data/raw and transform commands will transform that data and store artefacts in ./data/preprocessed.

The get_xlsx_uris cli

Pre-processing cli for bundestag cli.

uv run get_xlsx_uris run --help

Module for collecting and storing XLSX URIs from Bundestag data sources. Also done with

bundestag download bundestag_sheet --do-create-xlsx-uris-json

Notebooks

• fraktionszwang
• embedded members of parliament