Simon Munzert

data science and public policy. politics. experiments. surveys. and all that.

Thesis supervision

I'm always happy to hear from students who want to write a data-driven thesis on politics, public opinion, or the digital public sphere. I supervise master's theses across Hertie School programmes (MDS, MPP, MIA, EMPA). This page tells you which topics I'm keen to supervise and what working with me looks like. It should help you decide whether we're a good fit—and, I hope, get you excited about a project.

In a nutshell: I supervise empirical, quantitative work connected to my research. You're welcome to bring your own idea; you can also pick up one of the scoped topics below, several of which come with data ready to go. The goal is a thesis good enough to become a publishable research paper.

Prefer to watch or listen?

OK — heads up: what follows below is a lot of text trying to summarize what my research is about and how it could connect to your thesis ideas (or just get you inspired). So if you're more of a video person, I (well, NotebookLM) made you a short summary. And if you're into podcasts, I (well, NotebookLM) made you one of those too. Both are based on my last 15 publications. I watched and listened to both, and I can confirm that this portrait of my research is about 70% correct and 30% bullshit — but hopefully accessible and fun.

Video summary of Simon Munzert's research and thesis opportunities Watch the video summary
Podcast about Simon Munzert's research and thesis opportunities Listen to the podcast

What I supervise

I'm a political scientist by training and a data scientist in practice. My work spans public opinion and political behavior, digital communication, hate speech, and election forecasting. Two questions drive most of my current agenda:

  • How do online media change political communication, preference formation, and participation in democratic systems?
  • How can new digital data sources—social media data, data on online behavior, user-generated metadata—be used to study political and social phenomena?

I prefer to supervise empirical work that uses quantitative methods. I'm glad to advise on: survey design, experiments (online, survey, and field), measurement, causal inference and impact evaluation, regression analysis, multilevel modeling, panel data, web and social media data collection, and quantitative meta-analysis. If you want to work with web-based data, I can help you set up automated data collection—though the actual collection is your job.

Topics I'd be happy to supervise

You're strongly encouraged to propose your own topic, ideally with a connection to my areas of expertise. That said, finding a focused, feasible research question is one of the hardest parts of the thesis process—many proposals are too broad or ambitious for a master's thesis. To help, here are scoped directions I'd enjoy supervising. Several come with existing data.

Content moderation

Preferences for online content moderation

How does online media use shape preferences for hate speech moderation and regulation? What's the penalty of poor education on being moderated online? Where do people draw the line on what shouldn't be allowed on social media? Data available: a survey of ~19,000 respondents across eleven countries on hate speech moderation, with political preferences, attitudes toward free speech, interactive moderation decisions, demographics, and open-text data for NLP.

Evidence & policy

How policymakers and citizens make sense of scientific evidence

What do policymakers and the public know about data science and AI, and how does it shape their policy preferences? Is there a "myside bias" in accepting scientific evidence? Data available: a survey of government workers and the general population assessing competence in making sense of scientific evidence—with room to collect more observational or experimental data.

NLP & policy advice

Tracing scientific advice in parliamentary discourse (partner: TAB)

In collaboration with the Office of Technology Assessment at the German Bundestag (TAB), trace the semantic footprint of scientific policy advice in political debate. Which NLP approaches—semantic similarity, embedding-based retrieval, corpus comparison—best identify traces of TAB reports in Bundestag discourse? Data available: 1,000+ TAB reports via the KITopen API and Bundestagsdrucksachen via the DIP API. Ideal for students interested in applied NLP in a policy context.

Forecasting

What makes people good (or bad) forecasters?

What's the effect of election forecasts on expectations and voting behavior? How reliable are forecasts of political and economic outcomes made by media pundits? Data available: a pre-election panel survey of 40,000+ respondents from the 2025 German federal election, covering expectations, political preferences, knowledge, and views on polling and forecasting.

Measurement

Measuring the prominence and influence of political elites

How can we measure the visibility and importance of political elites over time and across countries? Data available: the Comparative Legislators Database (67,000+ legislators worldwide, with rich individual-level data) and Wikipedia-based daily attention statistics.

Collect your own data

Collaborative survey project

Co-design a survey to collect your own data on public preferences, attitudes, knowledge, or behavior—survey experiments (framing, exposure, conjoint) very welcome. You get to collect original data; in return, expect real coordination effort and a crash course in the do's and don'ts of survey design and analysis.

EMPA students: I'm also happy to supervise theses with a technical / data-science angle, or linked to any of the substantive topics above.

Working with me: what to expect

  • The thesis should read like a publishable research paper: it (a) motivates the question, (b) gives a concise literature overview, (c) precisely outlines the research design, and (d) presents a rigorous empirical analysis that answers the question.
  • I give detailed feedback on the structure of your work and on your ideas as early as possible—but I don't revise text drafts in advance.
  • I accept both single-authored and co-written theses. Tell me whether you'd prefer to work alone or collaborate (you don't need a partner in mind when you propose).
  • I strongly encourage you to extend your statistical and programming skills along the way. For tough R problems, my door is open.
  • Beyond scheduled sessions, I'm available during office hours by appointment. If you'd like to meet, email me a few bullet points with the agenda first.

A sample of previously supervised theses

  • Online behavior & prediction: Privacy and Democracy in a Cambridge Analytica World: Predicting Party Choice from Browsing History; Predicting Voting Behavior from Search Engine Queries.
  • Hate speech & NLP: Watching the Watchers: A Comparative Audit of Cloud-Based Commercial Content Moderation Services; Fine-Tuning Language Models for Context-Aware Dog Whistle Detection; How Do Haters Hate? Hate and Offensive Speech against Greta Thunberg on Twitter.
  • Online media: YouTube News and Politics Media Diets: A User-based Analysis of Filter Bubbles and the Recommendation Algorithm.
  • Public opinion & experiments: The Distant Impact of Nuclear Disasters: Consequences of Fukushima on German Political Preferences; Do Natural Disasters Affect Electoral Outcomes? A Meta-Analysis.
  • Digital data & AI governance: Public Money, Public Code: Evaluating Public Investment in Open-Source Digital Infrastructure; Attitudes Towards Surveillance Technologies in Australian Public Spaces.

Recommended general readings

  • Imai, K. (2018). Quantitative Social Science: An Introduction. Princeton University Press.
  • Kellstedt, P., & Whitten, G. (2018). The Fundamentals of Political Science Research. Cambridge University Press.
  • Gelman, A., Hill, J., & Vehtari, A. (2020). Regression and Other Stories. Cambridge University Press.
  • Salganik, M. J. (2019). Bit by Bit: Social Research in the Digital Age. Princeton University Press.
  • Cunningham, S. (2021). Causal Inference: The Mixtape. Yale University Press.
  • Grimmer, J., Roberts, M. E., & Stewart, B. M. (2022). Text as Data: A New Framework for Machine Learning and the Social Sciences. Princeton University Press.
  • Wickham, H., Çetinkaya-Rundel, M., & Grolemund, G. (2023). R for Data Science (2nd ed.). O'Reilly.