Research Plan

• Read CLAUDE.md for project context
• If initial_plan.md or ideas.md exist, read them
• If data documentation exists, read it (codebook, README, data dict)

2. Draft all required sections

Produce a complete research_plan.md with ALL of these sections:

§1. Research Question 1–2 sentences. Should be answerable by the data.

§2. Identification Strategy

• Source of variation (what is the treatment? who? when?)
• Treatment definition with exact cohort timing
• Control group definition
• Key identifying assumption (state formally, e.g., "parallel trends: absent treatment, treated and control states would have followed parallel trends in Y")

§3. Primary Specification

• Display equation with all terms defined (use LaTeX notation if appropriate)
• Fixed effects (level and justification)
• Clustering (level and justification)
• Estimator (name and cite if non-standard)

§4. Expected Effects and Mechanisms

• Theoretical prediction: sign, rough magnitude, and why
• Mechanism: economic channel through which treatment affects outcome
• Key heterogeneity dimensions (at least 2)

§5. Planned Robustness Checks (≥8) Label R1–R8+. Must include:

• R_: Event study / pre-trend test (always required)
• R_: Placebo test (wrong group, wrong time, or wrong outcome)
• R_: Alternative estimator or specification
• R_: Permutation or randomization inference
• R_: Sample restriction robustness

§6. Data Sources For each dataset:

• Name, URL or local path
• Authentication needed (API key, credentials)
• Time period, geographic coverage, unit of observation
• Any known data limitations

§7. Power Assessment

• Pre-treatment periods
• Treated clusters/units
• Control clusters/units
• Minimum detectable effect (MDE) estimate
• Known constraints (short pre-period, few clusters, etc.)

§8. Known Limitations and Planned Responses 3 anticipated referee concerns + how to address each.

3. Save output

Save to: quality_reports/plans/YYYY-MM-DD_[paper-name]-research-plan.md

Add frontmatter:

---