From jcf-skills
Helps choose and defend a causal identification strategy (DID, IV, RDD, event study, matching) for JCF corporate-finance empirical papers with endogenous firm-level data.
How this skill is triggered — by the user, by Claude, or both
Slash command
/jcf-skills:jcf-identification-strategyThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- Picking a credible design for a corporate-finance question with endogenous choices
Corporate-finance variables (leverage, governance, payout, M&A) are choices, so OLS-with-controls invites endogeneity, omitted-variable, and reverse-causality critiques. JCF is empirical corporate finance: a clean identification strategy is what separates a publishable paper from a desk reject. Match the design to the source of variation.
xtabond2-style GMM with instrument-count discipline.Not every "exogenous" source of variation survives a JCF referee. Grade the shock before building on it:
Variation source | Credibility at JCF | Known objection to pre-empt
Staggered state law adoption | High if modern DID | Lobbying/timing endogeneity; heterogeneity bias
Federal regulation with size threshold | High | Bunching at the cutoff; anticipation effects
Index inclusion/exclusion (RDD) | High near cutoff | Local estimate only; index rules changed over time
Shareholder vote near 50% (RDD) | High | Close votes not random across firm types — test it
Import tariff / trade shocks | Moderate | Industry-level treatment; exposure-measure disputes
Natural disasters / plant-level shocks | Moderate | Location selection; general-equilibrium spillovers
CEO deaths / health shocks | Moderate | Small N; succession-planning selection
Instrument built from lagged choices | Low | Exclusion fails by construction — expect rejection
Hypothetical, numbers illustrative: a paper claims staggered anti-takeover statutes raise leverage. TWFE gives 0.024 (t = 3.1). The JCF hardening sequence: (1) a Goodman-Bacon decomposition shows 31% of identifying weight comes from late-versus-early treated comparisons — a red flag; (2) Callaway–Sant'Anna on clean controls gives 0.015 (t = 2.2) — smaller but alive; (3) event-study leads are flat for five pre-years (joint p = 0.41); (4) one state adopting after a lobbying scandal is dropped — the estimate moves to 0.014. The paper then reports the modern estimator as the headline, TWFE as a legacy comparison, and the decomposition in the appendix. That ordering — not the TWFE number — is what survives review here.
Every JCF design with treated firms needs one explicit paragraph: who became treated, why, and what that implies. Cover (a) the institutional reason treatment landed where it did, (b) a pre-treatment covariate comparison or trends table, (c) the direction of bias if a selection story survives, and (d) why the estimate is then a lower or upper bound. Omitting this paragraph is among the most common reasons an otherwise clean JCF design draws a second-round identification objection.
【Design】<DID/IV/RDD/event/matching> 【Variation】<source>
【Top threat】<x> → handled by <y>
【Diagnostics】pre-trend/first-stage/density/balance: [Y/N each]
npx claudepluginhub brycewang-stanford/awesome-journal-skills --plugin jcf-skillsStress-tests causal identification strategies (natural experiments, IV, DID, RDD) for empirical corporate finance papers targeting The Journal of Finance.
Stress-tests causal identification designs for JFE manuscripts: natural experiments, IV, staggered DID, RDD, and endogeneity/selection treatment.
Builds credible identification and research design for JFQA empirical finance papers: portfolio sorts, Fama-MacBeth, panel FE, staggered DID, IV, RDD, event studies. Also supports theoretical submissions.