Skip to content

Latest commit

 

History

History
27 lines (15 loc) · 2.16 KB

step-02.md

File metadata and controls

27 lines (15 loc) · 2.16 KB

Project Proposal, Step 2

Your grade will depend upon your ability to clearly respond to the assignment—concepts and constraints.

Assignment: Submit a written project proposal (~3 pages) or a presentation (3–5 minutes)

  1. Motivates and outlines your causal question of interest.
  2. Explains potential sources of selection that could lead to bias.
  3. Describes the ideal experiment which one could use to answer this question
  4. Discusses a practical research design through which one could answer the original question—clarifying how this research design avoids selection bias.

Note: Your question must be causal in nature. If it is not, come up with a new question.

Title: A title that clearly describes your question—and potentially how you would answer it.

Abstract: A brief description of your project. Clearly describe the main question, how you will answer it, and why/for whom the results matter. Be concise and clear. You hook the reader here and elaborate later.

The abstract should be less than 150 words.

Question and motivation: Explain why this area of research is interesting/important in general (not just to you). Why should your reader care/keep reading? After you briefly motivate the general topic, clearly describe your specific causal question. If necessary, motivate the specific question too.

Selection: Why is this question challenging to answer empirically? In other words, what sources of selection bias concern you? If we simply run a regression of y on X, why might the estimated effect be biased?

Ideal experiment: Describe the ideal experiment that would answer your question. This ideal experiment does not need to be practical—i.e., you do not need to be able to run it in real life.

Practical research design: How might you causally answer the your question in real life? Which data would you need? What sort of research design would you apply—selection on observables (regression with many controls, matching, propensity-score methods, etc.) or selection on observables (IV, RD, etc.)? How does this proposed research design avoid selection bias? Explain clearly.