| Referee objection | Weak response | JHR-calibrated response |
|---|---|---|
| "Pre-trends look noisy / TWFE biased with staggered timing" | Cite a methods paper and assert robustness | Re-estimate with a heterogeneity-robust estimator, show the new event study, add a pre-trend sensitivity bound |
| "Why does your estimate differ from [prior paper]?" | A paragraph of contextual speculation | Comparative estimation: prior spec on your sample, your spec on the prior window, bridge table cited |
| "Clustering seems wrong" | Footnote defending current choice | Re-cluster at the assignment level, add wild bootstrap if clusters are few, report both |
| "Effect could be sorting at the cutoff" | Verbal institutional argument only | Density test, covariate continuity, donut estimate in a new appendix exhibit |
| "Results are a LATE; policy claim is broader" | Soften one sentence | Characterize compliers, re-scope the policy paragraph, flag what does not travel |
Comment R2.3: The staggered rollout makes the TWFE estimates hard to interpret.
Response: Done. We now report Callaway-Sant'Anna group-time ATTs as the
preferred estimates (new Table 3) and move TWFE to Appendix Table A7 for
comparison. The event study (new Figure 2) shows pre-period coefficients near
zero; a pre-trend sensitivity exercise (Appendix Table A8) indicates the
headline effect survives violations up to twice the largest pre-period
estimate. Pages 14-16 are rewritten around the new estimates.
Every response of this type names the new exhibit, the new pages, and the result of the check — referees at this venue reward verifiable specificity.
【Decision】R&R / conditional accept
【Referee map】R1/R2/R3 asks → response type each
【Reconciliation】prior-vs-ours table added? [Y/N]
【Sensitivity】new tests + cross-refs? [Y/N]
【Length】still <=40pp; overflow in appendix? [Y/N]
【Next step】resubmit via msubmit.net (no fee)