Bachelor’s and Master’s Thesis Guidelines
- A thesis showcases your ability to do science.
- Before you start the thesis, you should have some knowledge of the broader topic (equivalent to a seminar or elective lecture).
- If not, spend one or two weeks reading around the topic before starting the actual project.
- Before you start the thesis, you should have a workflow ready for:
- Finding literature
- Programming (if the thesis has an empirical part)
- Writing
- All three can and probably should be AI-assisted.
Timeline
Before Registration
- Initial contact by mail. Send me a short email expressing your interest and, if you already have one, a topic idea. If not, I will suggest topics based on the questionnaire.
- Questionnaire. Fill out the thesis questionnaire available on my teaching website. It helps me propose a topic that fits your skills and interests.
- Broad preparation. After we have a broad topic idea, start reading around it to prepare the first meeting to discuss the specific topic title. Current topics I am interested in are listed on my teaching website.
First Meeting
- We discuss potential topics and your background.
- We agree on a direction.
- Registration happens after this meeting, once the topic is clear enough to commit to.
After Registration
- Start working. Begin with a literature review and, if applicable, data exploration.
- Share intermediate results. Send me drafts as you progress — I will give feedback. Use email for this; no special format is required.
- Meetings. We will have 1–2 additional meetings during the project. You can also use my office hours or reach out by email to schedule an online meeting.
- Ideally, you produce text snippets and code with plots/table output in this phase that you could already use in the final thesis.
- Before writing up. Check in with me before you start writing the final paper.
Final 2-3 weeks
- write up the paper and focus on finishing
- check, clean, and read several times before submitting
Communication
- Contact me by mail or in meetings.
- I typically respond to emails within a few business days.
- Check in regularly — to flag problems early or to share progress.
- Come prepared: describe what you have done, what your next steps are, and what specific questions you have.
- At minimum, meet with me once before registration and once before you start writing up.
Scope and Format
For formatting and style requirements (citations, figures, academic writing, software), see the Guidelines for Seminar Papers on my teaching webpage.
Word count:
Bachelor thesis: maximum 7,000 words (or 60,000 characters)
Master thesis: maximum 10,000 words (or 80,000 characters)
this is for main text only, excluding title page, references, tables, figures, etc.
you can submit additional supplementary material if necessary
Quality over quantity. Shorter is fine if the content is precise and complete.
Reproducibility
If your thesis contains an empirical analysis, the submission must include reproducible code in R or Python, unless we have agreed otherwise.
AI use declaration
Submission via mail should include a short text (< 1 page) on what AI tools were used and for what. If paid subscriptions were used, this should also be mentioned.
Submission Checklist
Next, to the official submission at the Exams Office, send me an email with
Using AI
AI tools are permitted and encouraged, but with boundaries:
- AI is not a citable source.
- All errors are your own.
- Also see the AWI guidelines:
Literature Work
- Use deep research tools to survey a field or identify relevant papers.
- Share PDFs in chatbots with prompts asking for explanations or connections to your topic.
- AI is a useful learning tool for basics in economics, statistics, and data analysis.
- Important: LLMs frequently fabricate literature references. Always verify sources independently before citing them.
Empirical Work and Programming
Autocomplete
Tools such as GitHub Copilot (free for students) suggest code completions as you type. Useful for reducing boilerplate.
Chatbot Dialog
Copy-paste code or error messages into a chatbot, describe your goal, and iterate. Effective for debugging and learning.
Agents
Agents can execute multi-step tasks autonomously. They can run in the command line (CLI) alongside your editor (RStudio, VS Code) or be integrated into it. Selected tools:
- Codex (OpenAI)
- Gemini CLI (Google)
- GitHub Copilot
- free tiers are available for students (subject to change)
- available as VS Code integration or Copilot Cli
- Cursor (IDE with integrated agent)
Understanding Your Code
Be sure to understand every part of your analysis. The logic should be described in the paper, but abstracted from implementation details.
- Good: “The data includes quarterly inflation rates defined as year-on-year changes.”
- Bad: “I merged the CPIq4 variable with the command joinx and divided by lag(CPIq4, lag=4).”
Recommended AI Workflow
Slow down. Work step-by-step. Prioritize learning over speed.
Share your background (coding experience, preferred language, etc.)
Share your goal — for example: “I want to conduct an empirical analysis of X using data source Y in language Z. The code should be simple and documented. Walk me through it step-by-step and explain each part. Ask if my prompts are ambiguous.”
Workflow of one step: 1. Discuss options 2. Generate a plan 3. Implement 4. Have the AI explain each step; ask questions until you understand
Writing
Use AI for structuring, copy editing, and feedback, but not for generating content.
- Avoid generating whole sentences or paragraphs without specific, deliberate prompting. AI-generated output may be adequate, but scientific writing should contribute new thinking, not repackage existing knowledge without citation.
- Make space for your own thoughts: use pen and paper, take a walk, talk to other people.
- Software tools are a great support, but can hinder creative and structured thinking.