identification-strategy

name	identification-strategy
description	Design or audit an empirical identification strategy for economics and applied social-science work. Use when Codex needs to translate a research question into an estimand and design, compare DiD or IV or RDD or event-study options, draft a pre-analysis plan, or review whether a manuscript's causal claims match the evidence.

Identification Strategy

Use this skill before coding when the empirical design is still fluid, or as a review pass when the paper already makes causal claims.

Workflow

1. State the claim in plain language.
2. Define the estimand, treatment, control, outcome, timing, and unit of observation.
3. Compare plausible designs and rank them by credibility and data fit.
4. For the preferred design, list:
   • identifying assumptions
   • diagnostics and falsification tests
   • clustering and inference plan
   • main threats and how to answer them
5. Stop the workflow if the design cannot support the intended claim.

Output rules

• Prefer a short strategy memo plus a checklist for implementation.
• Distinguish testable implications from untestable assumptions.
• Do not jump to code until the treatment, comparison group, and estimand are explicit.

Load references as needed

• Read references/design-checklists.md for design-specific checkpoints and memo structure.