TeX 53.6%
Python 46.2%
1# STATUS.md — Stages A–F2 complete; v1.1 after editorial-board revision; IRFA package regenerated23**Date:** 2026-10-17 (revision) — previous status 2026-07-1345## Revision 2026-10-17 (v1.1)67Full scientific review by the editorial board (`docs/REVIEW-REPORT.md`) with audit trail8(`docs/CHANGELOG-REVISION.md`). Main changes: pre-event levels quantified (+40×10⁻⁴, the9"no pre-event response" claim withdrawn); market-wide response bracketed between the ΔΔ (≈0) and a10pre-level-stratified estimator (+19×10⁻⁴ ≈ 2 % of base rate); co-arrival disclosure (22 % singleton11minutes) with estimates by class; network differential decomposed by decade (placebo contrast far12from zero within decades) and its secular rise flagged as partly a measurement effect (Epps);13co-jump severity not compared across the tick regime (1992–97 exceeds 2008 under the same14standardisation); robustness section restricted to executed checks; 19 references added, 5 removed,15124/124 Crossref-verified; `scripts/17_revision_checks.py` + Tables 9–10; macros 48 → 97; all four16PDFs recompiled (0 errors, 0 undefined references); anonymous version checked (0 identifiers).17Analyses recommended but not executed (raw data required): REVIEW-REPORT §8.181920## Completed work21221. **Project scaffold** per protocol §2: package (`src/intraday_fragility/`), configs,23 scripts, tests (7 passing), Makefile, `uv.lock`, docs, LaTeX tree.242. **Dataset probe (Stage A):** `mito0o852/OHLCV-1m` pinned at revision25 `776328445b7ac6e7815ef3a483e9c8ded1eb6d56`; inventory of 411 monthly Parquet26 files (87.7 GB), 1992-01→2026-03, **no missing months**27 (`data/manifests/repo_inventory.json`).283. **Pilot download** (1992-06, 2009-06, 2025-12; sha256 in29 `data/manifests/pilot_downloads.json`) and **pilot audit**30 (`scripts/01_pilot_audit.py`, JSON reports in `outputs/diagnostics/`,31 logs in `logs/audit/`).324. **Docs:** `DATA_AUDIT.md`, `LIMITATIONS.md`, `DECISION_LOG.md`,33 `DATA_DICTIONARY.md`, `COMPUTE_LOG.md`, `EMPIRICAL_DESIGN.md`,34 `REPRODUCIBILITY.md`, `LITERATURE_MAP.md` (candidates, none verified yet).355. **Pilot figures** (coverage, intraday seasonality; PDF+PNG; source data in36 `outputs/figures/data/`) and **summary-statistics LaTeX table**37 (`outputs/tables/tabP1_pilot_summary.{tex,csv}`).386. **Compiling LaTeX skeletons:** `paper/main.pdf` and39 `appendix/internet_appendix.pdf` build cleanly with the Data section drafted40 from verified audit facts only.417. **Stage B COMPLETE** (user instruction 2026-07-13: proceed with all data):42 all 411 monthly files downloaded (82 GB on disk), each verified against its43 LFS sha256 at download time; manifest `data/manifests/raw_files.{json,parquet,csv}`.44 Post-hoc full re-verification + per-file audit (row counts, duplicates,45 timestamp bounds) running (`verify-download` + `audit-raw`).468. **Literature review complete and massive:** 110 references verified at DOI47 level via OpenAlex/Crossref (two passes, 9 verification agents);48 `paper/sections/02_literature.tex` fully written (10 subsections organized49 around the contribution); documented novelty scan 2018–2026 supports50 positioning; manuscript compiles at 20 pages, 111 bib entries, 0 unresolved51 citations. See `docs/LITERATURE_MAP.md`.52539. **Stage C COMPLETE** (full-sample raw audit, 7.69B rows): perfect XNYS calendar54 match (0 missing/0 extra sessions, 0 weekend rows); DST confirmed clean (flat55 RTH share across calendar months); 100 OHLC violations total (1.3 per 10^8,56 concentrated 1993–95); 10,128 duplicate keys; 0 zero/negative volume. Daily57 ticker-day panel built (`data/metadata/daily/`, 411 files). 442k split-candidate58 ticker-days flagged (over-inclusive by design). 81,306 tickers, 221,957 spells59 (36% of tickers multi-spell → spell = identity unit). Coverage regime change in60 2020 (+10,965 entries, OTC expansion) identified as mandatory robustness axis.61 See DATA_AUDIT.md Stage C section, outputs/diagnostics/stage_c/.626310. **Stage D COMPLETE** (clean panel + universes + splits):64 * 6,944 high-confidence split events (4,819 fwd / 2,125 rev), ±1-day exclusion65 windows (20,662 ticker-days) — `outputs/diagnostics/stage_d/`;66 * monthly universes with strictly lagged screens (top500/1000/3000 by prior-month67 dollar volume; price/density screens; non-common ticker heuristic) —68 `data/metadata/universe_monthly.parquet` (4.44M rows);69 * clean minute panel `data/processed/minute/year=/month=/` (411 files, ~62 GB zstd):70 raw row total matches manifest EXACTLY (7,686,275,524); 10,128 duplicate rows71 merged per logged rule; sessions labelled with true XNYS closes (early-close72 truncation verified on 2025-12-24); flag_ohlc=100, split-window rows=4.30M;73 dedup verified (2020-03 spot check). RTH = 96.7% of rows.747511. **Stage E COMPLETE** (descriptive evidence): 9 JF-style figures (E1–E9, PDF+PNG,76 all with source CSVs in outputs/figures/data/) + Table I (sample construction by77 decade) + Table II (summary stats by decade, top-1000). Aggregates in78 outputs/diagnostics/stage_e/ (411 monthly minute-level aggregate files on the79 lagged top-1000 universe + Epps curves for 1996/2005/2015/2024).80 Headline descriptive facts (verified from saved outputs):81 * Extreme one-minute returns (|r|>1%) fell ~100× in frequency from the 1990s to82 the 2010s on the top-1000 universe, with sharp spikes in 2008 and 2020 —83 the paper's central tension in one figure (figE5).84 * The Epps effect has essentially disappeared: mean pairwise 1-minute correlation85 was ≈0.00 in 1996 (vs 0.16 at 30-min) but ≈0.21–0.30 in 2015/2024 with a nearly86 flat curve — cross-stock information incorporation is now sub-minute (figE8).87 * Intraday |return| and range shifted down decade by decade; volume concentrates88 increasingly at the close in recent decades (figE3).89 * CAVEAT for interpretation: pre-2001 minute returns embed 1/8–1/16 tick bounce;90 the 1990s "extreme" frequency partly reflects mechanical bid-ask bounce.91 Robustness planned: midpoint proxy, coarser sampling, range-based measures.929312. **Stage F baseline COMPLETE**: 1,854,384 real shock events (|z|≥6, isolated) +94 1,766,058 matched placebos, 410 months, lagged top-500 universe95 (outputs/estimates/events/, 410 files). Baseline result (Table IV, figs F1–F3,96 registry stage_f_baseline.json): **average abnormal propagation ≈ 0** —97 ΔΔbreadth between −1 and −3 ×10⁻⁴ across h∈{1..120} with day-block bootstrap98 SEs ≈ 1×10⁻⁵…1.5×10⁻⁴; no decade positive; weak insignificant positive gradient99 for |z|>12. Honest interpretation: isolated idiosyncratic shocks are absorbed.100 Falsification good news: pre-shock outcomes ≈ post placebo levels (no leakage).101 Next (protocol-specified): cumulative breadth through h, network-weighted102 receivers, and the common-shock (simultaneous-extreme) sample as separate object.10310413. **Stage F2 COMPLETE — the paper's headline mechanism found**: neighbor−non-neighbor105 ΔΔ breadth = +64.0 (SE 1.9) ×10⁻⁴ at h=5 (t≈34), decaying to +49.6 at h=30;106 **tripled across eras** (+28 in 1992–99 / +30 in 2000–09 → +81 in 2010s / +90 in107 2020s). Market-wide cumulative remains null. Co-jump measure corrected to108 severe-synchronization (>10% of names): crisis spikes 2008/2020; severity 9.3%109 (2008) / 6.4% (2020) vs 3.5% calm. Placebo-network falsification passes (−2.7±2.8).11014. **Robustness battery (Table VIII)**: null + network channel stable across111 post-decimalization, mid-day, ex-announcement windows, calm/crisis, |z|≥8/12,112 ex-non-common, month-cluster bootstrap. Network diff declines with magnitude113 (+64→+34) — documented.11415. **IRFA submission package** (outputs/release/): manuscript_irfa.pdf (elsarticle,115 38 pp, 0 unresolved refs, elsarticle-harv bib), highlights.pdf, cover_letter.pdf,116 online_appendix.pdf, submission_checklist.md. All numbers macro-generated117 (48 macros). Provenance framed as Finnhub.io-sourced aggregation (user directive).118119## Key findings (verified)120121- Schema uniform: `timestamp` (ns, UTC, bar-start), OHLC (USD, **unadjusted**),122 `volume` (shares, integer-valued), `ticker`. No nulls.123- Timestamps are true UTC; DST handled correctly (09:30 ET session start maps to124 13:30 UTC in June, 14:30 UTC in December).125- Zero OHLC violations in 58.8M pilot bars; 14 benign duplicate pairs (OTC only).126- **No zero-volume bars: missing minute = no trade.** Trade-conditional panel.127- Liquid names complete (390 bars/day); sparse tail is huge in 2025 (23k tickers,128 median 5.5 bars/day — OTC expansion).129- Delisted securities present (only 22–26% of 1992 tickers persist).130- Early-close days handled by data; sessions must be truncated by exchange calendar.131- Unscreened minute-return SD is outlier-dominated (splits/bad prints) ⇒132 corporate-action screens are prerequisites for variance-based measures.133134## Quality control135136- `uv run pytest`: 7/7 pass.137- Both LaTeX documents compile (TeX Live 2026).138- Figures validated for CVD safety (palette validator) + grayscale line styles.139140## Unresolved issues / decisions requiring attention1411421. **Data license unspecified** on the HF card (Finnhub-sourced). Research use with143 citation documented; public redistribution of derived *data* deferred until144 clarified. (DATA_AUDIT §1)1452. Venue scope of bars (consolidated vs single venue) unknown — Stage C probe planned.1463. Security-type classification requires an external reference or suffix heuristics.1474. Literature map has candidates only; verification pass (DOI-level) still to do.148149## Next commands150151```bash152# when download finishes:153PYTHONPATH=src uv run python -m intraday_fragility.cli verify-download154PYTHONPATH=src uv run python -m intraday_fragility.cli audit-raw # fills manifest row counts155# then Stage C full raw audit scripts (04–07)156```157