rfs-identification
Identification Strategy (rfs-identification)
When to trigger
- The empirical core is OLS + controls with an open endogeneity threat
- A DID uses two-way fixed effects (TWFE) without addressing staggered-adoption bias
- An IV has a weak first stage or a contestable exclusion restriction
- A cross-sectional asset-pricing claim rests on a factor that may be data-mined
- Reviewers will ask "what is your source of variation?" and you lack a crisp answer
The RFS identification bar
RFS applies the same high causal-inference standard as JF and JFE: a claim of causality requires a credible source of exogenous variation, not a richer control set. RFS is more receptive than the others to genuinely new questions and to structural / theoretical identification — but novelty never substitutes for a clean design. Pick the strongest feasible strategy below.
RFS-specific lever — Stage 1 design review. Because RFS pioneered Registered Reports (pre-results review; Karolyi, "Kick-Starting the Review Process," RFS 27(2), 2014), an identification strategy can be refereed before the results exist. If you pursue this route, the design must be airtight on paper: pre-specified sample, treatment definition, estimator, diagnostics, and the exact tables to be produced — because that protocol becomes the binding commitment that earns in-principle acceptance. Even for a standard submission, draft the design as if it had to survive Stage 1 review with no results to fall back on.