Week 4 exercise: A Simulated Leniency Design
PP5001 · Ungraded practice
Use the downloadable Quarto notebook to investigate a hypothetical training programme. The notebook supplies Python code that generates applicants randomly assigned to caseworkers with different approval thresholds. You will construct a leave-one-out instrument, examine the first stage and reduced form, and compare OLS and IV estimates with the known causal effect.
The notebook includes the questions, explanations of the variables, and space for your code and interpretation. The default code generates a fresh sample each time. Instructions explain how to choose a random seed if you want to keep your sample unchanged.
Work through the notebook and render your completed analysis to HTML or PDF. This exercise is ungraded.
Critique Claude
Give Claude the following prompt:
Explain how differences in disability examiners’ approval rates can be used to estimate the effect of disability insurance on employment. What can policymakers learn from this research? Write approximately 400 words for a Master of Public Policy student.
Include Claude’s response in your notebook. Then answer:
Assess Claude’s explanation using the assigned paper and the IV assumptions discussed in class. Identify two claims that are well supported and two statements that require correction or qualification. Explain your reasoning and rewrite the latter statements.