In the United States, every state has different voter registration and voting processes. While voting is easy and accessible in most cases, the process sometimes is so complicated that it suppresses the votes of many eligible voters. Barriers include limited voter registration periods, online or offline voter registration, requirements of documents like birth certificates, or disenfranchisement. Since these obstacles impact different groups of people differently, the actual voters are a biased subgroup of the eligible voters whose opinion is possibly not representative for all Americans. Read here for an overview.
In this project you will analyze datasets containing online voter registrations for the 2020 election and try to answer the question “What if voting was accessible to everyone?”. We suggest using US Census information in addition that functions as your ground-truth, and then applying statistical and data mining techniques to identify and correct the bias. However, you are encouraged to choose your own path as we, alongside with political scientist Ioannis Ziogas, provide guidance and support.

Duration and Type:



Please bring at least basic skills in statistics, data mining, and/or machine learning. You will also need sufficient programming skills for data handling, pre-processing, model application, evaluation, and visualization in a suitable programming language such as Python.

Supervisor and Contact:

You will meet regularly with Joerg Wicker (main supervisor), Katharina Dost, and Ioannis Ziogas. If you are interested, please contact Joerg Wicker per mail.