Strategy Memo — Critique
Reviewer: strategist-critic Target: quality_reports/strategy-memo.md Paper type: Descriptive predictive (asset-pricing replication) Phase: Strategy (constructive severity) Date: 2026-05-11
Score
91/100 — PASS (round 1).
Verdict
PASS. Memo clears the 80 threshold by a wide margin. All 11 required sections present, descriptive framing impeccable (no causal language anywhere), 9-check robustness suite addresses real literature concerns, inference machinery (NW 12 lags, Petersen month clustering, Shanken secondary) correctly assembled.
Issues (all Minor; -1 to -2 each)
| ID | Issue | Deduction |
|---|---|---|
| M1 | Citation form drift: plain-text “Doidge-Karolyi-Stulz” inline vs doidge2017 bibkey elsewhere |
-2 |
| M2 | Sample boundary for first monthly observation (Dec-1989 vs Jan-1990) not stated | -1 |
| M3 | CRSP return field (ret vs retx) not stated |
-1 |
| M4 | “FF12 lag” without fama1992cross bibkey |
-1 |
| M5 | R1 (FF3-residual IVOL) doesn’t say daily FF3 from Ken French | -1 |
| M6 | “Month” not explicitly defined as calendar month | -1 |
| M7 | Skip-a-month robustness omission | -2 |
Strengths
- Descriptive framing impeccable (INV-8 fully satisfied); no causal verbs.
- Falsification criteria explicit and quantitative for H1, H2, H3.
- Pre-registration discipline: 7 concrete commitments with numeric specificity.
- Anti-data-snooping: confirmatory (H1-H3) separated from supportive (R1-R9).
- Inference correct (NW 12, Petersen clustering, Shanken).
- 9 robustness checks (≥ 7 required), each tied to real literature concern.
- Honest self-critique pre-acknowledges five referee challenges.
- Lagged-ME weights for self-constructed market return; validation against
crsp.dsi.vwretd.
Summary
A strong, submission-ready Strategy-phase artifact. Strengths outweigh the minor specificity gaps. The remaining deductions are polish-level and tractable; none threaten the design. No round 2 required for gate purposes.
Recommendations (advisory)
- Add explicit statement: “We use CRSP
retthroughout. A ‘month’ is the calendar month.” - State sample boundaries: “First predictive observation Feb 1990 (IVOL from Jan 1990 daily returns); last Dec 2024.”
- Standardize citations to bibkey form throughout memo body.
- Add to R1: “Daily FF3 factors from Ken French Data Library, US Research Factors, daily.”
- Optional R10 — Skip-a-month variant — to absorb bid-ask reversal concern at month boundary (≈30 lines of code).