SPB Git forge
spboucher.ai ↗

spb/wp15_uqo

Public MIT
3commits 1branches 1releases
7.6 MBsize
maindefault branch
52 min agolast push
TeX 53.6% Python 46.2%

révision scientifique complète (editorial board, 2026-10-17)

v1.0 → v1.1 (October 2026). Pre-event levels quantified, market-wide response
bracketed (ΔΔ vs pre-level-stratified), co-arrivals disclosed, network
differential decomposed by decade with the Epps/measurement caveat, co-jump
severity not compared across the tick regime, robustness restricted to executed
checks, 19 references added / 5 removed (124/124 Crossref-verified),
scripts/17_revision_checks.py + Tables 9–10, 97 macros, all PDFs and the IRFA
release package regenerated. See docs/REVIEW-REPORT.md and
docs/CHANGELOG-REVISION.md.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Simon-Pierre Boucher committed 52 min ago (Oct 18, 2026) parent 80aa1b7

70 changed files +7,011 −953

modified CITATION.cff +4 −4
@@ -1,9 +1,9 @@
1 1 cff-version: 1.2.0
2 −title: "Faster, Wider, More Fragile? The Evolution of Intraday Shock Propagation in U.S. Equity Markets"
3 −message: "Research in progress; please contact the author before citing."
2 +title: "Faster, Wider, More Fragile? The Evolution of Intraday Shock Propagation in U.S. Equity Markets, 1992-2026"
3 +message: "UQO Working Paper No. 15, version 1.1 (October 2026). Please cite the working paper."
4 4 type: software
5 5 authors:
6 6 - family-names: Boucher
7 7 given-names: Simon-Pierre
8 −version: 0.1.0
9 −date-released: 2026-07-13
8 +version: 1.1.0
9 +date-released: 2026-10-17
modified README.md +43 −33
@@ -3,26 +3,27 @@
3 3
4 4 **UQO Working Paper No. 15** — *The Evolution of Intraday Shock Propagation in U.S. Equity Markets, 1992–2026*
5 5
6 −[![Paper](https://img.shields.io/badge/paper-PDF%20(36%20p.)-b31b1b?logo=latex&logoColor=white)](paper/main.pdf)
7 −[![Journal version](https://img.shields.io/badge/IRFA%20manuscript-elsarticle%20(41%20p.)-b31b1b?logo=latex&logoColor=white)](paper/main_irfa.pdf)
6 +[![Paper](https://img.shields.io/badge/paper-PDF%20(43%20p.)-b31b1b?logo=latex&logoColor=white)](paper/main.pdf)
7 +[![Journal version](https://img.shields.io/badge/IRFA%20manuscript-elsarticle%20(50%20p.)-b31b1b?logo=latex&logoColor=white)](paper/main_irfa.pdf)
8 8 [![Python](https://img.shields.io/badge/python-3.12%2B%20·%20uv-3776AB?logo=python&logoColor=white)](pyproject.toml)
9 9 [![Data](https://img.shields.io/badge/data-7.69%20bn%20one--minute%20bars%20·%2081%2C306%20tickers-4051b5)](docs/DATA_AUDIT.md)
10 −[![Reproducible](https://img.shields.io/badge/reproducible-end--to--end%20pipeline%20(17%20scripts)-2e7d32)](scripts/)
10 +[![Reproducible](https://img.shields.io/badge/reproducible-end--to--end%20pipeline%20(18%20scripts)-2e7d32)](scripts/)
11 11 [![Institution](https://img.shields.io/badge/UQO-D%C3%A9pt.%20des%20sciences%20administratives-16365c)](https://uqo.ca)
12 12 [![Author](https://img.shields.io/badge/author-Simon--Pierre%20Boucher-16365c)](mailto:simon-pierre.boucher@uqo.ca)
13 13
14 14 > **TL;DR** — Using 7.69 billion one-minute bars on the full listed *and delisted* U.S. equity
15 −> cross-section (1992–2026), we track how intraday shocks propagate. (1) Markets absorb
16 −> idiosyncratic shocks: across 1.85 million isolated ±6σ minute events, abnormal market-wide
17 −> propagation against matched within-stock placebos is economically zero at 1–120 minutes, in every
18 −> decade. (2) Transmission is concentrated and intensifying: the shocked stock's twenty closest
19 −> lagged-correlation neighbours respond strongly (+64×10⁻⁴ abnormal jump probability at five
20 −> minutes) and this network channel tripled from the 1990s (+28) to the 2020s (+90) while
21 −> market-wide spillover stayed at zero. (3) Tail synchronisation migrated to the market-wide
22 −> minute: ordinary extreme minute returns became 108× rarer than in the 1990s, the Epps curve
23 −> flattened (one-minute pairwise correlation 0.004 in 1996 vs 0.30 in 2015), and severe co-jump
24 −> minutes spike by orders of magnitude in 2008 and 2020. Modern markets absorb the idiosyncratic
25 −> and transmit the systemic.
15 +> cross-section (1992–2026), we track how intraday shocks propagate. (1) Isolated idiosyncratic
16 +> shocks barely spread: across 1.85 million isolated ±6σ minute events, the market-wide response
17 +> against matched within-stock placebos is bracketed between zero (difference-in-differences,
18 +> −1 to −3×10⁻⁴) and about 2 % of the base rate (pre-level-stratified, +19×10⁻⁴) at 1–120 minutes,
19 +> in every decade. (2) Transmission is concentrated: the shocked stock's twenty closest
20 +> lagged-correlation neighbours respond (+64×10⁻⁴ at five minutes, +51 in singleton minutes) and
21 +> this network channel rose from +28 (1990s) to +90 (2020s) — a rise that partly reflects sharper
22 +> identification of neighbours as the Epps effect receded. (3) Common information is impounded
23 +> within the minute (one-minute pairwise correlation 0.004 in 1996 vs 0.30 in 2015), ordinary extreme
24 +> minutes became ~10× rarer after decimalisation (~100× from the wide-tick 1990s), and severe co-jump
25 +> minutes are roughly ten times more frequent in 2008 and 2020 than in calm years. Modern markets
26 +> absorb the idiosyncratic, transmit it narrowly, and synchronise on the systemic.
26 27
27 28 ---
28 29
@@ -30,9 +31,9 @@
30 31
31 32 | Item | Value |
32 33 |---|---|
33 −| Version | 1.0 (July 2026) — UQO working-paper cover added 2026-10-17 |
34 +| Version | **1.1 (October 2026) — revised 2026-10-17** after a full editorial-board review (`docs/REVIEW-REPORT.md`, `docs/CHANGELOG-REVISION.md`); v1.0 = July 2026, tag `pre-revision-20261017` |
34 35 | SSRN | not yet submitted |
35 −| Journal | *International Review of Financial Analysis* — submission package ready in `outputs/release/` (manuscript, anonymous manuscript, highlights, cover letter, title page, online appendix), **not yet submitted** (`outputs/release/submission_checklist.md`: online appendix and data-licence note still open) |
36 +| Journal | *International Review of Financial Analysis* — submission package **regenerated for v1.1** in `outputs/release/` (manuscript 50 p., anonymous manuscript, highlights, cover letter, title page, online appendix), **not yet submitted**; before submitting: complete acknowledgments/funding, re-run the novelty scan, check the current Guide for Authors (`outputs/release/submission_checklist.md`) |
36 37 | Code & data | reproducible from `scripts/` (see Quick start); raw minute bars (88 GB) not redistributed |
37 38
38 39 ## 🎯 What this paper does
@@ -43,28 +44,31 @@ and delisted common stock from 1992 to 2026 — and three complementary measurem
43 44
44 45 1. **Shock-propagation event studies**: 1.85 million isolated ±6σ one-minute events, each matched to
45 46 a within-stock placebo minute, tracked at horizons of 1 to 120 minutes for breadth (share of other
46 − stocks jumping), gradient (response by liquidity/size tier) and network response (twenty nearest
47 − lagged-correlation neighbours);
47 + stocks with a tail exceedance |z|>2), intensity, signed response, shock-magnitude gradient and
48 + network response (twenty nearest lagged-correlation neighbours);
48 49 2. **Epps curves**: pairwise correlation as a function of sampling interval, by year;
49 50 3. **Co-jump counts**: the number of stocks with extreme minute returns in the same minute, with a
50 51 severity scale calibrated on calm, GFC and COVID periods.
51 52
52 53 All headline numbers are macro-generated from machine-readable registries
53 −(`paper/results_macros.tex`, built by `scripts/15_build_results_macros.py`); no number is hand-typed.
54 +(`paper/results_macros.tex`, 97 macros built by `scripts/15_build_results_macros.py`); in-text
55 +references to table entries quote the generated tables.
54 56
55 57 ## 🏆 Headline results
56 58
57 59 | Object | Estimate |
58 60 |---|---|
59 61 | Panel | 7,686,275,524 bars, 81,306 tickers, 221,957 listing spells |
60 −| Isolated ±6σ events (real / placebo) | 1,854,384 / 1,766,058 |
61 −| Abnormal market-wide breadth, 5 min / 30 min | −1.04 (0.90) / −3.10 (1.01) ×10⁻⁴ — zero |
62 −| Network response (20 neighbours), 5 / 15 / 30 min | +64.0 (1.8) / +55.4 (1.8) / +49.6 (1.8) ×10⁻⁴ |
63 −| Network response by decade (5 min) | 1990s +28 → 2000s +30 → 2010s +81 → 2020s +90 |
64 −| Extreme minute returns (>1%), 1990s → 2010s | 518 → 4.8 per million (÷108) |
62 +| Isolated ±6σ events (real / placebo) | 1,854,384 / 1,766,058 (match rate 95.2 %; 71 % in the 1990s); 22 % of events alone in their minute |
63 +| Market-wide breadth, ΔΔ 5 min / 30 min | −1.04 (0.90) / −3.10 (1.01) ×10⁻⁴ (base rate ≈ 0.10) |
64 +| Market-wide breadth, pre-level-stratified 5 / 30 min | +19.5 (0.9) / +18.4 (1.0) ×10⁻⁴ (≈ 2 % of base rate) — upper bound |
65 +| Pre-event level, real − placebo | +40.3 (1.2) ×10⁻⁴ (shocks arrive in active minutes; differenced out) |
66 +| Network response (20 neighbours), 5 / 15 / 30 min | +64.0 (1.8) / +55.4 (1.8) / +49.6 (1.8) ×10⁻⁴; +51 (2.5) in singleton minutes; ticker-cluster SE 4.7 |
67 +| Network response by decade (5 min) | 1990s +28 → 2000s +30 → 2010s +81 → 2020s +90 (placebo contrast +200 / −41 / −44 / +68) |
68 +| Extreme minute returns (>1 %), 1990s → 2010s | 518 → 4.8 per 10,000 (÷108; 2002 → 2010s ÷10) |
65 69 | Epps: 1-min correlation, 1996 / 2015 / 2024 | 0.004 / 0.30 / 0.21 |
66 −| Severe co-jump minutes, calm / GFC / COVID | 3.5 / 9.3 / 6.4 per 10,000 ; COVID severe share 83.7 % vs 9.53 % pre-COVID |
67 −| Robustness (Table VIII) | 11 designs incl. placebo-network falsification, month-cluster bootstrap, post-decimalisation |
70 +| Severe co-jump minutes per 1,000, 2017–19 / 2008 / 2020 | 9.5 / 91 / 84 (1992–97: 128 — tick artefact, not comparable) |
71 +| Robustness | 11 designs (Table 8) + design diagnostics (Table 9) + network components (Table 10) |
68 72
69 73 ## 🗂 Repository layout
70 74
@@ -72,7 +76,7 @@ All headline numbers are macro-generated from machine-readable registries
72 76 intraday-fragility/
73 77 ├── config/ # all parameters (YAML) — no magic constants in code
74 78 ├── src/intraday_fragility/ # package: download, audit, cleaning, features, shocks, networks, econometrics, tables, figures
75 −├── scripts/ # 00_download_pilot … 16_robustness (numbered pipeline)
79 +├── scripts/ # 00_download_pilot … 17_revision_checks (numbered pipeline)
76 80 ├── tests/ # pytest (7 tests)
77 81 ├── data/ # manifests only (raw bars never committed)
78 82 ├── outputs/
@@ -128,7 +132,8 @@ licence — the raw data are therefore not redistributed (`docs/LIMITATIONS.md`)
128 132 4. **Co-jumps** (`13`–`14`): market-wide extreme-minute counts and severity regimes.
129 133 5. **Inference** (`src/…/econometrics.py`): event-clustered and month-cluster bootstrap standard errors.
130 134 6. **Robustness** (`16`): 11 designs — post-decimalisation, mid-day, ex-announcement, calm/crisis, |z| ≥ 8/12, ex-non-common, placebo network.
131 −7. **Reporting** (`15`, `src/…/tables.py`, `figures.py`): results macros, LaTeX tables, CVD-safe vector figures.
135 +7. **Design diagnostics** (`17`, added 2026-10-17, from stored event files): pre-event levels, co-arrival classes, pre-level-stratified estimator, ticker-cluster bootstrap, signed response, network-differential components by decade.
136 +8. **Reporting** (`15`): results macros, LaTeX tables, CVD-safe vector figures.
132 137
133 138 ## 🖼 Figures & tables inventory
134 139
@@ -142,14 +147,16 @@ licence — the raw data are therefore not redistributed (`docs/LIMITATIONS.md`)
142 147 | `figF4_network` | nearest-neighbour network response |
143 148 | `figF5_cojumps` | co-jump counts and severity regimes |
144 149
145 −Tables (`paper/tables/*.tex`): data audit, sample cascade, propagation estimates by horizon and decade, gradient, network, co-jumps, Epps, robustness battery (Table VIII).
150 +Tables (`paper/tables/*.tex`): sample construction (1), baseline propagation by horizon (2), design diagnostics (3), network mechanism (4), network components by decade (5), robustness battery (6); pilot summary and summary by decade in the online appendix.
146 151
147 152 ## ⚠️ Limitations
148 153
149 154 - Bars are trade-conditional (no zero-volume bars) and the venue scope (consolidated vs single venue) is not documented upstream.
150 −- Security-type classification relies on suffix heuristics; non-common securities are excluded in robustness only.
155 +- Security-type classification relies on suffix heuristics; ETFs cannot be identified; non-common securities are excluded in robustness only.
151 156 - Data licence unspecified on the dataset card: research use with citation, no redistribution of derived minute data.
152 −- Extended online appendix tables still to be finalised before journal submission.
157 +- Pre-2001 tick discreteness (52 % zero minute returns in 1996) precludes cross-regime comparisons of threshold-based series (extremes, co-jumps).
158 +- Neighbour sets are identified from one-minute residual correlations; the secular rise of the network channel combines transmission and measurement precision (see `docs/REVIEW-REPORT.md` §8 for the analyses needed to separate them, not executed: raw data required).
159 +- Real events arrive in minutes of elevated ambient activity; the market-wide response is bracketed (0 to ≈ 2 % of base rate), not pinned at zero.
153 160
154 161 ## 📝 Citation
155 162
@@ -161,7 +168,8 @@ Tables (`paper/tables/*.tex`): data audit, sample cascade, propagation estimates
161 168 type = {Working Paper},
162 169 number = {15},
163 170 year = {2026},
164 − month = {July}
171 + month = {October},
172 + note = {Version 1.1}
165 173 }
166 174 ```
167 175
@@ -176,5 +184,7 @@ Université du Québec en Outaouais (UQO), Gatineau, QC.
176 184 - Numbering: the project was first filed as `UQO_WP_001`; it is **Working Paper No. 15** since 2026-10-17 (No. 1 is the amenity-capitalisation paper).
177 185 - `paper/main.tex` (UQO cover, template of `Recherche-UQO/_modele-working-paper`) was created on 2026-10-17 from the IRFA manuscript; the pre-existing skeleton (“Preliminary — do not cite”) was discarded. Scientific content is identical to `main_irfa.tex`; only the cover, bibliography style (`aer`) and layout differ.
178 186 - Three PDFs are committed on purpose: `main.pdf` (working paper), `main_irfa.pdf` (journal), `main_irfa_anon.pdf` (anonymous review copy).
179 −- `outputs/estimates/` (707 MB) is excluded from git; regenerate with the pipeline.
187 +- `outputs/estimates/` (707 MB) is excluded from git; regenerate with the pipeline. `outputs/diagnostics/revision_checks.json` is a committed copy of the 2026-10-17 diagnostics registry.
188 +- **Revision 2026-10-17 (v1.1)**: full editorial-board review — `docs/REVIEW-REPORT.md` (diagnosis, methodology verdicts, literature, maturity) and `docs/CHANGELOG-REVISION.md` (audit trail of every correction, references added/removed, reanalyses). The IRFA package in `outputs/release/` was regenerated from v1.1; if the sections change again, regenerate it (`latexmk -pdf main_irfa.tex main_irfa_anon.tex`, appendix, highlights, then copy to `outputs/release/`).
189 +- Bibliography: 124 references, all verified against Crossref by DOI on 2026-10-17 (`outputs/diagnostics/bibliography/`).
180 190 - Original project brief: `docs/PROMPT_INITIAL_CLAUDE.md`; development conventions: `docs/CLAUDE.md`.
modified STATUS.md +17 −2
@@ -1,6 +1,21 @@
1 −# STATUS.md — Stages A–F2 complete; IRFA submission package assembled
1 +# STATUS.md — Stages A–F2 complete; v1.1 after editorial-board revision; IRFA package regenerated
2 +
3 +**Date:** 2026-10-17 (revision) — previous status 2026-07-13
4 +
5 +## Revision 2026-10-17 (v1.1)
6 +
7 +Full scientific review by the editorial board (`docs/REVIEW-REPORT.md`) with audit trail
8 +(`docs/CHANGELOG-REVISION.md`). Main changes: pre-event levels quantified (+40×10⁻⁴, the
9 +"no pre-event response" claim withdrawn); market-wide response bracketed between the ΔΔ (≈0) and a
10 +pre-level-stratified estimator (+19×10⁻⁴ ≈ 2 % of base rate); co-arrival disclosure (22 % singleton
11 +minutes) with estimates by class; network differential decomposed by decade (placebo contrast far
12 +from zero within decades) and its secular rise flagged as partly a measurement effect (Epps);
13 +co-jump severity not compared across the tick regime (1992–97 exceeds 2008 under the same
14 +standardisation); robustness section restricted to executed checks; 19 references added, 5 removed,
15 +124/124 Crossref-verified; `scripts/17_revision_checks.py` + Tables 9–10; macros 48 → 97; all four
16 +PDFs recompiled (0 errors, 0 undefined references); anonymous version checked (0 identifiers).
17 +Analyses recommended but not executed (raw data required): REVIEW-REPORT §8.
2 18
3 −**Date:** 2026-07-13
4 19
5 20 ## Completed work
6 21
modified appendix/internet_appendix.pdf +0 −0

Binary file not shown.

modified appendix/internet_appendix.tex +65 −18
@@ -22,7 +22,9 @@ originally sourced from Finnhub.io and aggregated into 411 monthly files
22 22 file's SHA-256 checksum was verified at download against the archive's
23 23 metadata and re-verified after the full download; the manifest (per-file
24 24 checksum, byte size, row count, column schema, timestamp bounds, symbol
25 −count, duplicate count) accompanies the replication package.
25 +count, duplicate count) accompanies the replication package. The archive
26 +carries no explicit licence; the minute data are used for research with
27 +attribution and are not redistributed.
26 28
27 29 \paragraph{Calendar and timezone.} Observed trading dates match the official
28 30 XNYS calendar exactly over the full sample: zero missing sessions, zero dates
@@ -52,12 +54,37 @@ coverage expansion (2020, over-the-counter names) is absorbed by fixed-rule
52 54 lagged universes.
53 55
54 56 \paragraph{Corporate actions.} A conservative overnight-ratio screen
55 −(canonical ratios $\{2,\dots,8,10,20,3/2\}$ and inverses, $\pm1.5\%$
56 −tolerance, both-day dollar volume $\ge \$50{,}000$, prior close $\ge \$1$)
57 +(canonical ratios 2 through 8, 10, 20, and 3/2, with their inverses,
58 +$\pm1.5\%$ tolerance, both-day dollar volume $\ge \$50{,}000$, prior close $\ge \$1$)
57 59 detects \NSplitEvents{} events (\NSplitFwd{} splits, \NSplitRev{} reverse
58 60 splits); event days $\pm1$ trading day are excluded from every return-based
59 61 statistic.
60 62
63 +\paragraph{Pilot audit.} Before the full download, three pilot months (June
64 +1992, June 2009, December 2025; 58.8~million bars) were audited. Trading-day
65 +counts match the exchange calendar in all three, including the
66 +December~24, 2025 early close; no OHLC-consistency violation, null value,
67 +negative volume, or zero-volume bar occurs; 14 duplicate ticker--minute pairs
68 +(all in inactive over-the-counter names) appear in December 2025. Liquid
69 +securities are densely covered throughout: the four most active tickers in
70 +December 2025 have complete 390-bar regular sessions, and large issuers
71 +already average 277--369 bars per day in June 1992. Volume units were
72 +verified against public daily volumes for benchmark securities.
73 +Figure~\ref{fig:app_pilot_coverage} summarizes raw data volume and per-day
74 +symbol counts, Figure~\ref{fig:app_pilot} the pilot intraday seasonality,
75 +and Table~\ref{tab:pilot_summary} pilot summary statistics for the 500
76 +highest dollar-volume securities in each pilot month.
77 +
78 +\begin{figure}[t]
79 +\centering
80 +\includegraphics[width=\textwidth]{figP1_coverage.pdf}
81 +\caption{\textbf{Dataset coverage.} Panel (a): compressed size of each
82 +monthly source file, 1992--2026 (all 411 files of the pinned snapshot). Panel
83 +(b): number of symbols with at least one bar per trading day in the three
84 +pilot months.}
85 +\label{fig:app_pilot_coverage}
86 +\end{figure}
87 +
61 88 \section{Additional descriptive evidence}
62 89 \label{app:descriptive}
63 90
@@ -68,7 +95,7 @@ stale-price share, and the time series of detected corporate-action events.
68 95 Table~\ref{tab:summary_by_decade} reports decade-level summary statistics for
69 96 the top-1000 universe.
70 97
71 −\begin{figure}[h]
98 +\begin{figure}[t]
72 99 \centering
73 100 \includegraphics[width=\textwidth]{figE2_survivorship.pdf}
74 101 \caption{\textbf{Universe dynamics.} (a) Active symbols per year; (b) entries
@@ -76,7 +103,7 @@ and exits. The shaded band marks the 2020 source-coverage expansion.}
76 103 \label{fig:app_survivorship}
77 104 \end{figure}
78 105
79 −\begin{figure}[h]
106 +\begin{figure}[t]
80 107 \centering
81 108 \includegraphics[width=\textwidth]{figE4_concentration.pdf}
82 109 \caption{\textbf{Concentration of trading.} (a) Share of aggregate dollar
@@ -85,7 +112,7 @@ volume.}
85 112 \label{fig:app_concentration}
86 113 \end{figure}
87 114
88 −\begin{figure}[h]
115 +\begin{figure}[t]
89 116 \centering
90 117 \includegraphics[width=.6\textwidth]{figE6_rv.pdf}
91 118 \caption{\textbf{Cross-section of realized volatility.} Median, interquartile
@@ -94,15 +121,16 @@ top-1000 universe.}
94 121 \label{fig:app_rv}
95 122 \end{figure}
96 123
97 −\begin{figure}[h]
124 +\begin{figure}[t]
98 125 \centering
99 126 \includegraphics[width=.6\textwidth]{figE7_staleness.pdf}
100 127 \caption{\textbf{Stale-price share.} Percentage of one-minute returns exactly
101 −zero, lagged top-1000 universe.}
128 +zero, lagged top-1000 universe (\ZeroRetNinetySix\% in 1996,
129 +\ZeroRetTwentyFour\% in 2024).}
102 130 \label{fig:app_staleness}
103 131 \end{figure}
104 132
105 −\begin{figure}[h]
133 +\begin{figure}[t]
106 134 \centering
107 135 \includegraphics[width=.6\textwidth]{figE9_splits.pdf}
108 136 \caption{\textbf{Detected corporate-action events per year.} Forward splits
@@ -118,11 +146,13 @@ zero, lagged top-1000 universe.}
118 146
119 147 Figure~\ref{fig:app_evolution} shows the year-by-year evolution of abnormal
120 148 market-wide breadth at $h=5$ and $h=30$ with 95\% day-block bootstrap bands:
121 −the market-wide null holds in essentially every year, not merely on average.
122 −Figure~\ref{fig:app_pilot} reproduces the pilot-sample intraday seasonality
123 −used in the initial data audit.
149 +the market-wide difference-in-differences is within two standard errors of
150 +zero in essentially every year, not merely on average. The design diagnostics
151 +(pre-event levels, co-arrival classes, pre-level-stratified estimates,
152 +alternative clustering, signed responses) and the decomposition of the
153 +network differential by decade are reported in the main text.
124 154
125 −\begin{figure}[h]
155 +\begin{figure}[t]
126 156 \centering
127 157 \includegraphics[width=.62\textwidth]{figF2_evolution.pdf}
128 158 \caption{\textbf{Evolution of abnormal market-wide propagation.}
@@ -131,7 +161,7 @@ day-block bootstrap intervals.}
131 161 \label{fig:app_evolution}
132 162 \end{figure}
133 163
134 −\begin{figure}[h]
164 +\begin{figure}[t]
135 165 \centering
136 166 \includegraphics[width=\textwidth]{figP2_seasonality.pdf}
137 167 \caption{\textbf{Pilot-sample intraday seasonality} (June 1992, June 2009,
@@ -157,9 +187,26 @@ labelling with true exchange closes, integrity and corporate-action flags);
157 187 universe construction with strictly lagged screens; descriptive aggregation;
158 188 shock detection and outcome measurement (\NEventsReal{} events,
159 189 \NEventsPlacebo{} placebos); network and co-jump measurement; and estimation.
160 −End-to-end wall-clock time is approximately six hours on a single
161 −consumer-grade machine (Apple Silicon, 4 parallel workers); peak memory is
162 −below 8~GB per worker because every stage operates month-by-month on
163 −partitioned columnar files.
190 +Every stage operates month-by-month on partitioned columnar files with four
191 +parallel workers on an Apple-silicon workstation; wall-clock times for each
192 +stage are recorded in the run logs of the replication package. The
193 +revision diagnostics (pre-event levels, co-arrival classes, stratified
194 +estimators, alternative clustering) are recomputed from the stored
195 +event-level files in a few minutes without access to the raw minute panel.
196 +
197 +\section{Analyses not executed}
198 +\label{app:notdone}
199 +
200 +The following checks are referred to in the main text as not executed, so
201 +that no claim rests on them: source thresholds $|z|\ge4$ and $|z|\ge5$;
202 +receiver threshold $|z|>3$; the top-1000 and top-3000 shock universes;
203 +neighbor sets estimated on 5- or 30-minute residual correlations or with a
204 +transaction-time covariance estimator; the pre-event ($t-5$) network
205 +contrast; permuted-network placebos preserving industry and liquidity
206 +composition; propagation measured inside scheduled-announcement windows; a
207 +Hawkes-type model of the co-jump minute series; and a reconciliation of
208 +bar volumes against consolidated-tape volumes to establish the venue scope
209 +of the source. Each requires re-running the detection stage on the raw
210 +minute panel or external data.
164 211
165 212 \end{document}
modified appendix/tables/tabP1_pilot_summary.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/02_pilot_figures_tables.py — do not edit numbers by hand.
2 −\begin{table}[t]
2 +\begin{table}[htbp]
3 3 \centering
4 4 \begin{adjustbox}{max width=\textwidth}
5 5 \begin{threeparttable}
modified appendix/tables/table2_summary_by_decade.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Summary statistics by decade, top-1000 universe}
5 5 \label{tab:summary_by_decade}
added docs/CHANGELOG-REVISION.md +75 −0
@@ -0,0 +1,75 @@
1 +# CHANGELOG-REVISION.md — Journal d'audit de la révision scientifique (editorial board, 2026-10-17)
2 +
3 +État de départ figé : tag git `pre-revision-20261017`, PDF `docs/main_pre-revision-20261017.pdf` (36 p., v1.0, July 2026).
4 +État d'arrivée : v1.1, October 2026 — `paper/main.pdf` (43 p.), `paper/main_irfa.pdf` (50 p.), `paper/main_irfa_anon.pdf` (50 p.),
5 +`appendix/internet_appendix.pdf` (8 p.). Rapport de l'editorial board : `docs/REVIEW-REPORT.md`.
6 +
7 +Convention : **Où** · **Avant** · **Après** · **Pourquoi** · **Preuve**.
8 +
9 +## 1. Corrections majeures (fond)
10 +
11 +| # | Où | Avant | Après | Pourquoi | Preuve |
12 +|---|---|---|---|---|---|
13 +| 1 | §5.5, §6.2, §1 (« falsification pré-événement ») | « real-event outcomes at t−5 equal placebo levels » (affirmé, jamais chiffré) | Niveaux pré-événement rapportés : breadth 1 044 vs 1 004 ×10⁻⁴ (écart +40,3, SE 1,2) ; intensité +0,074 (SE 0,012). L'affirmation d'égalité est retirée ; l'écart est présenté comme la base de différenciation et comme un diagnostic. | L'affirmation était fausse ; les chocs réels surviennent dans des minutes d'activité transversale élevée. | `scripts/17_revision_checks.py`, `outputs/tables/table9_design_checks.{tex,csv}` (Panel A) |
14 +| 2 | §5.1, §6.2, abstract, §1, §9 | Réponse marché « economically zero » (ΔΔ −1 à −3 ×10⁻⁴) | Réponse marché **bornée** entre ≈0 (ΔΔ) et ≈19 ×10⁻⁴ ≈ 2 % du taux de base (estimateur stratifié sur le niveau pré-événement, éq. (\ref{eq:strat})) ; le signe négatif du ΔΔ pooled attribué au retour à la moyenne des minutes à co-arrivées (3–5 événements : −8,3 (2,0) à h=30 ; minutes singleton : +1,7 (1,5) et +3,7 (1,8)) | La différenciation sur un niveau pré-événement élevé induit un biais de retour à la moyenne ; les deux estimateurs encadrent l'effet | Table 9 Panels B–C |
15 +| 3 | §4.3, §5.3, §6.2 | Événements qualifiés d'« isolés » sans préciser qu'ils partagent souvent leur minute | Divulgation : 22 % des événements sont seuls dans leur minute ; 2,1 événements source par minute d'événement en moyenne ; estimations séparées par classe de co-arrivée | Les filtres d'isolement plafonnent la part de |z|>4 à 2 %, pas le nombre d'événements |z|≥6 | Table 9 Panel B |
16 +| 4 | §4.6, §7.1, abstract, §1 | « network channel tripled » présenté comme fait de marché | Hausse maintenue (28→30→81→90 ×10⁻⁴) mais **explicitement confondue** avec la précision d'identification des voisins : les corrélations 1 min de 1996 (0,004) identifient mal les pairs économiques (effet Epps) ; décomposition en quatre composantes (Table 10) montrant que le contraste placebo est +200 ×10⁻⁴ (1990s) et −44 (2010s) | Lecture alternative « changement de mesure » non exclue ; analyse recommandée (voisins sur corrélations 5/30 min ou Hayashi–Yoshida) non exécutable sans données brutes | Table 10 ; `revision_checks.json` |
17 +| 5 | §7.1 | — | Différentiel réseau sur minutes singleton : +51 (2,5) vs +64 pooled → co-arrivées de paires corrélées ≈ 1/5 de l'effet, pas la source | Alternative mécanique (co-sauts simultanés + persistance de |z|) à écarter | Table 9 Panel B |
18 +| 6 | §7.3, §1, abstract, §9 | « tail synchronization migrated to the market-wide minute » ; « spike by orders of magnitude in 2008 and 2020 » | Comparaison inter-régime de tick **interdite** : 1992–1997 = 128 minutes sévères/1 000 et sévérité 7,1 % (> 2008 = 91 et 9,3 %) sous la même standardisation → discrétisation des prix. Claim restreinte à l'ère post-décimalisation : « roughly an order of magnitude » (2008 : 91 vs < 6 en 2006 ; 2020 : 83,7 vs 9,5 en 2017–19) | Figure F5 montre les années 1990 au-dessus de 2008 ; le texte l'ignorait | `outputs/figures/data/figF5_cojumps.csv` ; macros `\SevereEarlyNineties`, `\CoJumpSevEarlyNineties`, `\SevereGFC` |
19 +| 7 | §6.1, §1 | « a factor of 108 » sans mise en perspective | Facteur 108 conservé mais attribué en grande partie au tick (52 % de rendements minute nuls en 1996 vs 13 % en 2024) ; déclin post-décimalisation 2002→2010s ≈ 10× (50 → 4,8 par 10 000) | Honnêteté sur le biais de discrétisation | macros `\ZeroRetNinetySix`, `\ExtremeGtOneTwoThousandTwo` |
20 +| 8 | §6.1 | Epps : « information… now spreads within the minute » | Deux mécanismes non séparés : asynchronie/prix stagnants (Renò 2003 ; Hayashi–Yoshida 2005 ; Zhang 2011) et incorporation plus rapide ; Tóth–Kertész 2009 cité ; conséquence pour l'identification des voisins | Interprétation unilatérale | — |
21 +| 9 | §5.3, §6.2 | Taux de base « 7 ×10⁻² » / « ~700 ×10⁻⁴ » ; MDE « 2 ×10⁻⁴ = one security per 25 events » ; « 3 ×10⁻⁴ = per 17 events » | Taux de base mesuré 0,10 (1 004 ×10⁻⁴) ; récepteurs observables 427 (237 dans les 1990s) ; MDE 2,5 ×10⁻⁴ = un récepteur par 9 événements ; hausse voisin ≈ 6 % du taux de base (64/1 004) au lieu de « nine percent » | Chiffres erronés (arithmétique incohérente avec N≈427) | macros `\BaseBreadthShare`, `\NEligMean`, `\MDEFive`, `\MDEEventsPer` |
22 +| 10 | §4.3 | « on the order of one event per stock-week » | ≈ 2,5 par titre-semaine | 1 854 384 / (500 × 7 282/5) | macro `\EventsPerStockWeek` |
23 +| 11 | §5.2 | « match rate exceeds 95% throughout » | 95,2 % global ; **71 % dans les années 1990** ; ≥ 96 % ensuite | Faux pour les années 1990 | macros `\MatchRate*` |
24 +| 12 | §4.4 | Taux de base cumulé « about 53% at h=15 » | 58 % (fenêtres placebo) | Mesuré | macro `\CumBaseFifteen` |
25 +| 13 | §8 (Robustness) | Affirmations de tests non exécutés : « range-based measures, 5- and 15-minute aggregation », « N ∈ {500,1000,3000} », « ETFs… excluded », « source shocks at |z|≥4,5,6 and receiver thresholds |z|>2,3 », « sensitivity to window width and detection thresholds is tabulated », « Internet Appendix reports the full battery » | Section réécrite : seules les 11 lignes de Table 8 + Tables 9–10 sont décrites ; les tests non exécutés listés explicitement (§8, Annexe §D) | Intégrité (master prompt §11) : aucun de ces tests n'existe dans `outputs/` ni dans les scripts (`scripts/11` ne stocke que |z|≥6 et |z|>2) | `scripts/`, `outputs/estimates/` |
26 +| 14 | §2 (Literature) | Mentions de méthodes non utilisées : « local projections (Jordà) », « machine-learning extension (Gu–Kelly–Xiu, Sirignano–Cont, Easley et al.) », « permuted-network placebos », « transmitters, receivers, amplifiers, absorbers », « lead–lag falsification benchmark », « announcement-window propagation as a separate labeled object », « no headline result may be driven by ETFs » | Supprimées ou reformulées (ETF : « cannot be identified… stated rather than papered over ») ; 5 références devenues non citées retirées du .bib | Revue de littérature annonçant des analyses absentes | diff `02_literature.tex` |
27 +| 15 | §4.4, §7, abstract | « jump probability », « jumps beyond 4σ » pour des dépassements |z|>2 ou >4 | Terminologie « tail exceedance / extreme move » ; « co-jump minute » conservé avec avertissement (seuils, pas tests de sauts formels Lee–Mykland / Jacod–Todorov) | Un référé en économétrie des sauts objecterait | — |
28 +| 16 | §4.2 | Facteur saisonnier défini de façon ambiguë (« \overline{med} ») ; MAD=0 non mentionné | Éq. (4) réécrite avec u_{i,t}, médiane transversale, normalisation explicite ; exclusion MAD=0 (= filtre de stagnation, actif surtout avant 2001) ; justification Boudt–Croux–Laurent 2011 ; biais Scholes–Williams des bêtas | Équation incomplète / implémentation (`scripts/11`) non décrite | `scripts/11_detect_shocks_measure_propagation.py` |
29 +| 17 | §4.6 | Éq. réseau : non-voisins « same over j∉N » | Dénominateur explicite N_{t+h}−|N_i|−1 ; contraste pré-événement (t−5) absent du pipeline signalé comme limite | Définition incomplète | — |
30 +| 18 | §5.3 | Inférence : jour + mois | + bootstrap par titre (SE réseau 4,7 vs 1,8–1,9 ×10⁻⁴ ; marché insensible) | Dépendance intra-titre non traitée | Table 9 Panel D |
31 +| 19 | §6.2 | Signed outcome défini (éq. 7) mais jamais rapporté | Rapporté (Table 9 Panel E : bruité, jamais significatif) ; asymétrie du différentiel réseau par signe (66 vs 63) | Objet défini sans résultat | Table 9 |
32 +| 20 | §3 (Data) | Section pilote (2 figures + 1 table) dans le corps ; limites venue/licence/type de titre dispersées | Pilote déplacé en annexe (§A) ; §3 réorganisé : source, audit complet, univers (règles explicites), limites (venue, licence, types, discrétisation, dividendes non détectés mais sans effet intraday) | Lisibilité journal ; exhaustivité des limites | — |
33 +| 21 | Titre / §1 / §9 | Réponse à « More Fragile? » implicite | Réponse explicite en trois parties (faster : oui ; wider : non, « sharper » ; more fragile : seulement au sens borné, intra-ère post-décimalisation) | Cohérence titre–résultats | — |
34 +| 22 | Annexe §D (compute) | « approximately six hours on a single consumer-grade machine » | Retiré (non vérifiable : logs partiels) ; renvoi aux logs ; nouvelle section « Analyses not executed » | Intégrité | `logs/` |
35 +| 23 | IRFA : Data availability ; Acknowledgments | Licence non mentionnée ; « [To be completed.] » | Licence non spécifiée → pas de redistribution ; remerciements neutres (« All errors are the author's own. ») — à compléter par l'auteur | Paquet de soumission | `main_irfa.tex` |
36 +| 24 | highlights.tex, cover_letter.tex, title_page.tex | « ~56-fold » (périmé vs 108 dans le texte) ; « no market-wide propagation » ; `\today` | Cohérents avec v1.1 ; date October 2026 | Incohérence interne | `outputs/release/` |
37 +
38 +## 2. Corrections éditoriales (forme)
39 +- Abstract réécrit (3 versions synchronisées : `\WPabstract`, `main_irfa.tex`, `main_irfa_anon.tex`) ; `\WPversion` 1.0 → 1.1 ; `\WPdate` July → October 2026.
40 +- Listes numérotées (`enumerate`) de §4.3 et §5.2 converties en prose (standard UQO).
41 +- Équations : numérotation et références complètes (éq. (1)–(10) + (\ref{eq:strat})) ; équation réseau scindée en `align` (overfull 23 pt corrigé).
42 +- Tableaux : placement `[htbp]` dans tous les fragments générés et dans les scripts générateurs ; `\clearpage` avant les références (les tableaux 7–10 étaient rejetés après la bibliographie).
43 +- Commentaires LaTeX périmés (« to be added after Stages B–D ») retirés.
44 +- Légendes : Fig. E8 (attenuation asynchrone), Fig. F4 (« other-receiver »), Fig. F5 (avertissement pré-2001), Fig. F1 (« difference-in-differences »).
45 +- Compilation : 0 erreur, 0 citation/référence indéfinie, overfull max 7,5 pt (elsarticle, tableaux ajustés), 0 pt (article).
46 +- Version anonyme : recherche plein texte Boucher / Simon-Pierre / UQO / Outaouais / Gatineau / spboucher / acknowledg / uqo.ca → 0 occurrence ; métadonnée auteur = « [Author identity withheld for review] ».
47 +
48 +## 3. Références
49 +- **110 références existantes vérifiées** une par une par résolution de leur DOI sur Crossref (2026-10-17, `outputs/diagnostics/bibliography/crossref_check_existing.json`) : titre (similarité > 0,85 pour 110/110), premier auteur, année (online-first vs print acceptée). Écarts résiduels = diacritiques/casse (Pástor, O'Hara, Aït-Sahalia, Jordà, Antón, Ben-David, Comerton-Forde, Roşu) et une coquille Crossref (« Anderson » pour Andersen et al. 2003). Aucune correction bibliographique nécessaire.
50 +- **5 références retirées** (non citées après suppression des mentions de méthodes non utilisées) : gu2020empirical, sirignano2019universal, easley2021microstructure, jorda2005estimation, hasbrouck1991measuring.
51 +- **19 références ajoutées**, toutes vérifiées sur Crossref (`crossref_new_candidates.json`, DOI dans le .bib) : Epps/asynchronie — Tóth & Kertész 2009 (QF, 10.1080/14697680802595668), Renò 2003 (IJTAF, 10.1142/S0219024903001839), Hayashi & Yoshida 2005 (Bernoulli, 10.3150/bj/1116340299), Zhang 2011 (JoE, 10.1016/j.jeconom.2010.03.012), Lo & MacKinlay 1990 (JoE, 10.1016/0304-4076(90)90098-E), Scholes & Williams 1977 (JFE, 10.1016/0304-405X(77)90041-1), Harris 1991 (RFS, 10.1093/rfs/4.3.389), Hansen & Lunde 2006 (JBES, 10.1198/073500106000000071) ; co-sauts — Jacod & Todorov 2009 (AoS, 10.1214/08-AOS624), Gilder–Shackleton–Taylor 2014 (JBF, 10.1016/j.jbankfin.2013.04.025), Bollerslev–Todorov–Li 2013 (JoE, 10.1016/j.jeconom.2012.08.014), Bormetti et al. 2015 (QF, 10.1080/14697688.2014.996586), Calcagnile et al. 2018 (QF, 10.1080/14697688.2017.1403141), Dungey et al. 2018 (JoE, 10.1016/j.jeconom.2017.09.002) ; saisonnalité/dérives — Boudt–Croux–Laurent 2011 (JEF, 10.1016/j.jempfin.2010.11.005), Christensen–Oomen–Renò 2022 (JoE, 10.1016/j.jeconom.2020.11.004) ; annonces/structure — Bollerslev–Li–Xue 2018 (REStud, 10.1093/restud/rdy003), Bessembinder et al. 2016 (JFE, 10.1016/j.jfineco.2016.02.011), Johnson et al. 2013 (Sci. Rep., 10.1038/srep02627).
52 +- Total : **124 références citées, 124 entrées .bib, 124/124 vérifiées**.
53 +
54 +## 4. Réanalyses exécutées (toutes depuis `outputs/estimates/events*/`, sans données brutes)
55 +| Script | Durée | Sortie | Contenu |
56 +|---|---|---|---|
57 +| `scripts/17_revision_checks.py` (nouveau) | ≈ 2 min (Apple Silicon) | `outputs/estimates/revision_checks.json` (+ copie `outputs/diagnostics/`), `outputs/tables/table9_design_checks.{tex,csv}`, `outputs/tables/table10_network_components.{tex,csv}`, copies `paper/tables/` | niveaux pré-événement ; taux de base ; récepteurs observables ; co-arrivées ; ΔΔ et différentiel réseau par classe de co-arrivée ; estimateur stratifié ; bootstrap par titre ; réponse signée ; différentiel par signe ; composantes réseau par décennie ; taux d'appariement par décennie |
58 +| `scripts/15_build_results_macros.py` (étendu) | < 5 s | `paper/results_macros.tex` (48 → 97 macros) | nouvelles macros pour tous les chiffres ci-dessus + divulgations co-sauts 1990s, stagnation, extrêmes 2002 |
59 +| Vérification bibliographique (`/tmp/bibcheck/*.py`, non versionnée) | ≈ 1 min | `outputs/diagnostics/bibliography/*.json` | Crossref |
60 +
61 +Note de reproductibilité : l'ordre de sortie de `group_by` de polars n'est pas stable, ce qui faisait varier les SE bootstrap de ±0,1 ×10⁻⁴ d'une exécution à l'autre (explique Table 4 SE 0,90 vs Table 8 SE 0,87 pour le même estimateur). `scripts/17` trie les clusters avant tirage ; les scripts 12/14/16 d'origine n'ont pas été modifiés ni relancés (données brutes requises pour les étapes amont, et les tables existantes sont conservées telles quelles).
62 +
63 +## 5. Chiffres du texte modifiés (ancien → nouveau, justification)
64 +- Taux de base breadth : 7 ×10⁻² → 0,10 (mesuré, placebo h=5).
65 +- Récepteurs observables : implicite ~200 → 427 (237 en 1990s).
66 +- MDE : 2 ×10⁻⁴ / « per 25 events » → 2,5 ×10⁻⁴ / « per 9 events ».
67 +- Hausse du voisin en % du taux de base : « nine percent » → « six percent ».
68 +- Événements par titre-semaine : « one » → 2,5.
69 +- Match rate : « > 95% throughout » → 95,2 % ; 71 % (1990s) ; ≥ 96 % après.
70 +- Base cumulée h=15 : 53 % → 58 %.
71 +- Co-sauts : « orders of magnitude » → « roughly an order of magnitude » (2008 : 91 ; 2020 : 83,7 ; calme : < 6–9,5 par 1 000).
72 +- Tous les autres chiffres (ΔΔ, différentiel réseau, Epps, extrêmes, splits, audit) sont inchangés.
73 +
74 +## 6. Éléments non modifiés faute de données (voir REVIEW-REPORT §6)
75 +Seuils |z|≥4/5 et |z|>3 ; univers top-1000/3000 ; voisins sur corrélations 5/30 min ou Hayashi–Yoshida ; contraste réseau à t−5 ; placebos de réseau permutés ; propagation dans les fenêtres d'annonces ; identification des ETF ; réconciliation des volumes (venue) ; modèle de Hawkes des minutes de co-saut.
modified docs/DECISION_LOG.md +14 −0
@@ -88,3 +88,17 @@ bibliography; highlights, cover letter, declarations, data-availability produced
88 88 (outputs/release/). Provenance framed as Finnhub.io-sourced aggregation with pinned
89 89 public snapshot for reproducibility (user directive). JF-format main.tex retained
90 90 for reference. Ref: paper/main_irfa.tex, outputs/release/.
91 +
92 +**2026-10-17 — Editorial-board revision (v1.1).**
93 +Decisions: (1) report pre-event levels instead of asserting equality; (2) present the market-wide
94 +response as a bracket [ΔΔ, pre-level-stratified] rather than "economically zero"; (3) disclose
95 +co-arrivals and report estimates by class; (4) decompose the network differential by decade and
96 +flag the Epps/measurement confound; (5) forbid cross-tick-regime comparisons of threshold series;
97 +(6) describe only executed robustness checks; (7) replace "jump" by "tail exceedance" for |z|>2;
98 +(8) add ticker-cluster bootstrap; (9) make bootstrap draws deterministic in scripts/17 (sort
99 +clusters) — scripts 12/14/16 left unchanged (raw data needed to rerun; SE differences ±0.1×10⁻⁴
100 +documented); (10) table placement [htbp] + \clearpage before references; (11) move pilot audit
101 +to the online appendix. Alternatives considered: re-matching placebos on pre-event level (needs
102 +raw panel); dropping the decade comparison of the network channel (kept with explicit caveat).
103 +Pre-specified: no (review-driven). Refs: docs/REVIEW-REPORT.md, docs/CHANGELOG-REVISION.md,
104 +scripts/17_revision_checks.py. IRFA Guide for Authors not re-checked on this date (author action).
modified docs/EMPIRICAL_DESIGN.md +14 −6
@@ -27,9 +27,17 @@ multiple-testing corrections within outcome families.
27 27
28 28 ## Status
29 29 - [x] Stage A pilot audit (2026-07-13)
30 −- [ ] Stage B full download — in progress
31 −- [ ] Stage C full raw audit
32 −- [ ] Stage D clean panels
33 −- [ ] Stage E descriptive evidence
34 −- [ ] Stage F shocks & propagation
35 −- [ ] Stage G paper
30 +- [x] Stage B full download (2026-07-13)
31 +- [x] Stage C full raw audit
32 +- [x] Stage D clean panels
33 +- [x] Stage E descriptive evidence
34 +- [x] Stage F shocks & propagation (baseline, F2 network/cumulative/co-jumps, robustness)
35 +- [x] Stage G paper (v1.0 2026-07-13; v1.1 revised 2026-10-17)
36 +- [x] Revision diagnostics from stored events (scripts/17, 2026-10-17)
37 +
38 +Implemented vs planned (as of 2026-10-17): implemented — residual shocks |z|≥6 with isolation,
39 +placebo timing, ΔΔ, day/month/ticker bootstrap, lagged top-20 correlation neighbours, cumulative
40 +breadth, co-jump minutes, pre-event levels, co-arrival classes, stratified estimator. Not implemented
41 +— two-way clustering regressions, local projections with fixed effects, permuted networks, future
42 +network weights, multiple-testing corrections, sector proxies (no security master), macro-window
43 +propagation as a separate object, ML extension. See paper §8 and REVIEW-REPORT §8.
modified docs/LIMITATIONS.md +26 −0
@@ -41,3 +41,29 @@ Zero-volume bars do not exist in the data: a missing minute means "no reported t
41 41 - One-minute returns of sparse names are contaminated by bid–ask bounce and stale prices; robust/quantile methods and liquidity screens are required.
42 42 - Common-information contamination (macro news moving everything at once) is the central identification threat; OHLCV alone cannot label news — scheduled-announcement windows and placebo designs stand in (CLAUDE.md §12).
43 43 - Nothing in the data supports direct causal claims about algorithmic/high-frequency trading; only association with market-structure eras, with explicit competing explanations.
44 +
45 +## 7. Additions from the 2026-10-17 review (v1.1)
46 +
47 +- **Ambient-activity selection.** Real ±6σ events occur in minutes whose cross-sectional breadth is
48 + 40×10⁻⁴ (4 % of base rate) higher than at matched placebo minutes, although the market factor is
49 + calm at t. The ΔΔ removes the level but inherits mean reversion (negative pooled estimates at long
50 + horizons come from co-arrival minutes); the pre-level-stratified estimator is an upper bound
51 + (+19×10⁻⁴). The market-wide response is therefore bracketed, not pinned at zero.
52 +- **Co-arrivals.** Only 22 % of events are alone in their minute (2.1 source events per event minute).
53 + The network differential survives in singleton minutes (+51 vs +64 ×10⁻⁴).
54 +- **Neighbour identification.** Neighbour sets come from one-minute residual correlations in month
55 + m−1; with the 1996 Epps effect (mean correlation 0.004) they are a noisy selection of economic
56 + peers. The 1990s→2020s rise of the network differential combines transmission and measurement
57 + precision; neighbour sets on 5/30-minute or Hayashi–Yoshida correlations are required to separate
58 + them (not executed; raw data).
59 +- **Placebo network contrast.** ≈0 on average but +200 (1990s), −41 (2000s), −44 (2010s), +68 (2020s)
60 + ×10⁻⁴: every network statistic must be a real-minus-placebo difference.
61 +- **Tick regime.** 52 % of top-1000 one-minute returns are exactly zero in 1996 (13 % in 2024);
62 + MAD=0 securities are dropped; threshold-based series (extreme minutes, co-jump minutes, severity)
63 + are not comparable across 2001. The early 1990s show more "severe" minutes than 2008 under the same
64 + standardisation.
65 +- **Universe composition 2020–21.** Dollar-volume ranking admits leveraged ETPs and distressed
66 + issuers; the |r|>5 % spike in 2020 is partly compositional; ETFs cannot be identified.
67 +- **Not executed.** Lower source thresholds, higher receiver threshold, top-1000/3000 universes,
68 + permuted networks, announcement-window propagation, pre-event network contrast, venue
69 + reconciliation, formal jump/co-jump tests (see paper §8 and Online Appendix §D).
modified docs/LITERATURE_MAP.md +34 −0
@@ -119,3 +119,37 @@ glosten2021etf, coles2022index, bollerslev2020realized.
119 119
120 120 Full metadata (volume/issue/pages/DOI + abstract-based summaries) in the agent
121 121 verification reports of 2026-07-13 and in `paper/bibliography/references.bib`.
122 +
123 +## Verification pass of 2026-10-17 (editorial-board revision)
124 +
125 +All 110 entries above re-verified by DOI resolution on Crossref
126 +(`outputs/diagnostics/bibliography/crossref_check_existing.json`): title similarity > 0.85 for
127 +110/110, first author and year consistent (diacritics/online-first differences only). Five entries
128 +no longer cited after pruning the literature review were removed from `references.bib`
129 +(gu2020empirical, sirignano2019universal, easley2021microstructure, jorda2005estimation,
130 +hasbrouck1991measuring). Nineteen entries added, each verified on Crossref
131 +(`crossref_new_candidates.json`), DOI recorded in the .bib:
132 +
133 +| Key | Citation | Stream | Use in the paper |
134 +|---|---|---|---|
135 +| toth2009epps | Tóth & Kertész (2009), QF 9(7), 793–802 | 8,16 | Epps decline 1993–2003; denser trading explains only part |
136 +| reno2003closer | Renò (2003), IJTAF 6(1), 87–102 | 16 | Asynchronous trading reproduces Epps |
137 +| hayashi2005covariance | Hayashi & Yoshida (2005), Bernoulli 11(2), 359–379 | 16 | Estimator robust to asynchronicity (recommended extension) |
138 +| zhang2011estimating | Zhang (2011), J. Econometrics 160(1), 33–47 | 16 | Epps and microstructure noise |
139 +| lo1990nonsynchronous | Lo & MacKinlay (1990), J. Econometrics 45(1–2), 181–211 | 8 | Nonsynchronous trading |
140 +| scholes1977estimating | Scholes & Williams (1977), JFE 5(3), 309–327 | 11 | Beta bias from nonsynchronous data |
141 +| harris1991stock | Harris (1991), RFS 4(3), 389–415 | 16 | Price discreteness/clustering |
142 +| hansen2006realized | Hansen & Lunde (2006), JBES 24(2), 127–161 | 16 | Microstructure noise in HF second moments |
143 +| jacod2009testing | Jacod & Todorov (2009), Ann. Stat. 37(4), 1792–1838 | 9 | Formal common-jump test |
144 +| gilder2014cojumps | Gilder, Shackleton & Taylor (2014), JBF 40, 443–459 | 9 | Co-jumps among U.S. stocks |
145 +| bollerslev2013jump | Bollerslev, Todorov & Li (2013), J. Econometrics 172(2), 307–324 | 9 | Jump tails, extreme dependence |
146 +| bormetti2015modelling | Bormetti et al. (2015), QF 15(7), 1137–1156 | 9,10 | Hawkes factor model of systemic co-jumps |
147 +| calcagnile2018collective | Calcagnile et al. (2018), QF 18(2), 237–247 | 5,9 | Collective synchronization events, HF era |
148 +| dungey2018testing | Dungey et al. (2018), J. Econometrics 202(1), 18–44 | 10 | Mutually exciting jumps at HF |
149 +| boudt2011robust | Boudt, Croux & Laurent (2011), JEF 18(2), 353–367 | 9 | Robust periodicity for jump detection |
150 +| christensen2022drift | Christensen, Oomen & Renò (2022), J. Econometrics 227(2), 461–497 | 9 | Drift bursts vs jumps |
151 +| bollerslev2018volume | Bollerslev, Li & Xue (2018), REStud 85(4), 2005–2041 | 9 | Announcements, volume, volatility |
152 +| bessembinder2016liquidity | Bessembinder et al. (2016), JFE 121(1), 142–166 | 4 | Resiliency around predictable trades |
153 +| johnson2013abrupt | Johnson et al. (2013), Sci. Rep. 3, 2627 | 5 | Sub-second fractures beyond human response |
154 +
155 +Total after revision: 124 cited references, 124/124 verified. Novelty scan to be repeated before submission.
added docs/REVIEW-REPORT.md +76 −0
@@ -0,0 +1,76 @@
1 +# REVIEW-REPORT.md — Rapport de l'editorial board (microstructure, haute fréquence, contagion/co-sauts, économétrie financière)
2 +
3 +**Papier :** WP 15 — *Faster, Wider, More Fragile? The Evolution of Intraday Shock Propagation in U.S. Equity Markets, 1992–2026*
4 +**Révision :** 2026-10-17, v1.0 (July 2026, 36 p.) → v1.1 (October 2026, 43 p. ; version IRFA 50 p.)
5 +**Journal d'audit :** `docs/CHANGELOG-REVISION.md` · **Réanalyses :** `scripts/17_revision_checks.py`
6 +
7 +## 1. Appréciation générale
8 +
9 +Le papier assemble un panel minute exceptionnel (7,69 G barres, 81 306 tickers, 1992–2026, titres radiés inclus), l'audite avec un soin rare (calendrier XNYS exact, DST, 100 violations OHLC, 10 128 doublons, 6 944 événements de fractionnement) et pose une question économique nette. Le dispositif — chocs résiduels ±6 MAD, filtres d'isolement, placebos appariés intra-titre, différence de différences, bootstrap par jour, contraste voisins/non-voisins sur réseau retardé — est transparent et entièrement généré par scripts. La version 1.0 souffrait cependant de trois défauts que des rapporteurs d'IRFA/JFM/JFEC auraient relevés immédiatement : (i) plusieurs affirmations de falsification ou de robustesse n'étaient pas étayées par des sorties existantes (égalité des niveaux pré-événement, tests |z|≥4/5, |z|>3, top-1000/3000, mesures range-based, exclusion des ETF, placebos de réseau permutés, projections locales, extension ML) ; (ii) le récit « la fragilité a migré vers la minute » ignorait que, sous la même standardisation, les années 1990 produisent plus de minutes « sévères » que 2008 (artefact de discrétisation des prix) ; (iii) l'« triplement du canal réseau » était présenté sans discuter que les corrélations 1 min des années 1990 (0,004 en 1996) identifient mal les voisins. La révision corrige ces points, ajoute deux tableaux de diagnostics recalculés depuis les fichiers d'événements stockés, encadre la réponse marché entre deux estimateurs, et réécrit les sections dans un anglais académique sobre.
10 +
11 +## 2. Diagnostic par section
12 +
13 +| Section | Forces | Faiblesses / erreurs (v1.0) | Traitement (v1.1) |
14 +|---|---|---|---|
15 +| Abstract | Trois faits clairs, chiffres macro-générés | « economically zero » non borné ; « orders of magnitude » ; « transmit the systemic » non démontré | Réécrit : borne 0–2 % du taux de base ; « roughly ten times » ; caveat Epps sur le triplement |
16 +| 1 Introduction | Tension théorique bien posée ; contributions explicites | Pas de réponse explicite au titre ; « pre-event window shows no abnormal response » faux ; « factor of 108 » sans mention du tick | Réponse en trois parties ; niveaux pré-événement divulgués ; déclin post-décimalisation ≈ 10× |
17 +| 2 Littérature | 110 références vérifiées, organisées par courants | Annonce de méthodes non utilisées (local projections, ML, permuted networks, rôles transmetteurs/absorbeurs, benchmark lead–lag) ; absence des travaux sur asynchronie (Renò, Hayashi–Yoshida, Zhang, Tóth–Kertész), tests de co-sauts (Jacod–Todorov, Gilder et al.), synchronisation collective (Calcagnile et al., Bormetti et al.), périodicité robuste (Boudt et al.), dérives (Christensen–Oomen–Renò 2022) | Élaguée et enrichie (19 ajouts vérifiés, 5 retraits) ; nouvelle sous-section « Lead–lag, asynchronous trading, and the Epps effect » |
18 +| 3 Données | Audit exemplaire | Section pilote dans le corps ; venue, licence, types de titres, dividendes, discrétisation dispersés ; commentaires LaTeX périmés | Pilote → annexe ; sous-sections Source / Audit / Universes / Limitations ; règles d'univers explicitées |
19 +| 4 Mesure | Discipline de retard stricte ; définitions complètes | Facteur saisonnier mal écrit ; exclusion MAD=0 non documentée ; « one event per stock-week » (réalité 2,5) ; « 7 % » taux de base (réalité 10 %) ; « jump » pour |z|>2 ; dénominateur non-voisins implicite ; co-arrivées non divulguées | Éq. (4) réécrite ; filtre de stagnation documenté ; terminologie « tail exceedance » ; éq. (9)–(10) ; co-arrivées 22 % / 2,1 |
20 +| 5 Design | Estimand DiD clair, placebo bien construit | « match rate > 95 % throughout » (71 % en 1990s) ; MDE et arithmétique « per 25/17 events » fausses ; pas de clustering par titre ; « falsification » pré-événement affirmée | Estimateur stratifié (éq. 11) ; MDE 2,5 ×10⁻⁴ = 1 récepteur/9 événements ; bootstrap par titre ; batterie de diagnostics reformulée |
21 +| 6 Résultats | Null puissant, uniforme par décennie | ΔΔ négatif significatif à h=30 (t≈−3) non expliqué ; Epps interprété unilatéralement ; extrêmes 2020 sans caveat de composition (ETF à levier) | Table 9 : retour à la moyenne des minutes à co-arrivées ; borne 0–19 ×10⁻⁴ ; Epps : deux mécanismes ; caveat composition |
22 +| 7 Mécanismes | Différentiel réseau +64 (t≈34) ; co-sauts | Triplement sans caveat de mesure ; contraste placebo « ≈ 0 » faux par décennie (+200 / −44) ; 1990s > 2008 sur la série sévère ignoré ; « nine percent » (réalité 6 %) | Table 10 (composantes) ; singleton +51 ; deux lectures explicitées ; comparaison inter-régime interdite ; sévérité « roughly an order of magnitude » |
23 +| 8 Robustesse | Table 8 solide (11 designs) | Six familles de tests revendiquées sans sortie | Seuls les tests exécutés décrits ; liste des non-exécutés (§8 et Annexe §D) |
24 +| 9 Conclusion | Implication pour la surveillance | « Fragility migrated from securities to minutes » surinterprété | Réponse bornée au titre ; extensions hiérarchisées |
25 +| Annexe | Audit étendu, figures E2–E9 | « six hours » non vérifiable ; tabP1 dupliquée | Pilote regroupé ; section « Analyses not executed » |
26 +
27 +## 3. Méthodologie — verdict point par point (master prompt §3)
28 +
29 +- **Spécification justifiée** : oui pour le DiD placebo ; la version 1.0 omettait le biais de retour à la moyenne induit par la sélection sur un niveau ambiant élevé. Corrigé par l'estimateur stratifié et l'analyse par co-arrivée (Table 9).
30 +- **Hypothèses d'estimation explicites** : bêta mensuel retardé, MAD, facteur saisonnier médian transversal — désormais entièrement écrits ; biais Scholes–Williams mentionné et absorbé par le placebo.
31 +- **Variables définies** : oui ; « jump » remplacé par « tail exceedance » ; dénominateurs explicites.
32 +- **Effets fixes** : non applicables (design en moyennes appariées) ; le placebo intra-titre × minute-du-jour × mois joue ce rôle. Mentionné.
33 +- **Dépendance temporelle/transversale** : bootstrap par jour (78 % des événements partagent leur minute avec un autre événement) ; ajout du bootstrap par titre (SE réseau 1,9 → 4,7 ×10⁻⁴, t > 13) et du bootstrap par mois existant.
34 +- **Erreurs-types** : correctes (ratio de sommes rééchantillonné par grappes) ; non-déterminisme de l'ordre `group_by` corrigé dans `scripts/17` (documenté pour 12/14/16).
35 +- **Construction de l'échantillon** : transparente ; taux d'appariement par décennie désormais rapporté.
36 +- **Robustesse face aux vraies menaces** : tick (post-2002), annonces, crise/calme, seuils supérieurs, non-common, co-arrivées, clustering, placebo réseau. Menaces non traitées faute de données brutes : identification des voisins (Epps), ETF, venue.
37 +- **Interprétation des coefficients** : magnitudes rapportées en % du taux de base mesuré (6 % pour les voisins ; ≤ 2 % marché).
38 +- **Significativité statistique vs économique** : distinguée (ΔΔ −3,1 (1,1) à h=30 significatif mais économiquement nul et attribué au retour à la moyenne).
39 +- **Langage causal** : « predictive propagation », « association » ; le triplement est présenté comme « at least tripled along the measured network » avec deux lectures.
40 +
41 +## 4. Qualité des données — verdict
42 +
43 +Source HF non officielle (agrégation de Finnhub) : provenance documentée, snapshot épinglé par checksum, **licence absente** → non-redistribution (désormais dans §3, dans la déclaration de disponibilité des données IRFA et dans la checklist). Venue consolidée vs unique : inconnue (prints post-clôture suggèrent inclusion hors bourse) — explicité. Survivorship : titres radiés présents (22–26 % de persistance 1992→2009/2025, 800–1 600 sorties/an) mais capture jusqu'au dernier jour non vérifiable — explicité. Fractionnements : écran overnight conservateur, fenêtre ±1 jour ; dividendes non détectés mais sans effet sur des rendements intrajournaliers — explicité. Barres manquantes = absence de transaction ; dénominateurs sur récepteurs observables (427 en moyenne, 237 dans les 1990s) — chiffré. Discrétisation : 52 % de rendements minute nuls en 1996 (13 % en 2024) ; MAD=0 → exclusion ; aucune comparaison inter-régime sur les séries de seuil.
44 +
45 +## 5. Littérature
46 +
47 +- 110 références existantes : toutes vérifiées sur Crossref par DOI (titre, premier auteur, année) — aucune correction nécessaire (écarts = diacritiques, « Anderson » = coquille Crossref).
48 +- 19 ajouts (DOI dans `references.bib`, liste dans CHANGELOG §3) : Epps/asynchronie, tests de co-sauts, synchronisation collective, périodicité robuste, dérives, annonces, résilience, « machine ecology ».
49 +- 5 retraits (non cités après élagage).
50 +- Total 124, 124/124 vérifiées. Les suggestions Kyle–Obizhaeva 2016, Barrot–Kaniel–Sraer 2016 et Menkveld–Zoican 2017 ont été vérifiées sur Crossref mais non ajoutées (pas de phrase qu'elles étayeraient sans forcer le propos).
51 +
52 +## 6. Résultats — incohérences trouvées et corrigées
53 +Voir CHANGELOG §1 (24 entrées) et §5. Les plus importantes : niveaux pré-événement (égalité affirmée, écart +40 ×10⁻⁴ mesuré) ; taux de base 7 % → 10 % ; MDE ; « nine percent » → « six percent » ; match rate ; co-sauts 1990s ; highlights « 56-fold » périmés ; `\today` sur la page titre IRFA.
54 +
55 +## 7. Réanalyses exécutées
56 +`scripts/17_revision_checks.py` (≈ 2 min, depuis `outputs/estimates/events*/`, sans données brutes) → Table 9 (diagnostics de design, 5 panels) et Table 10 (composantes du différentiel réseau par décennie) ; 49 nouvelles macros ; résultats : pré-événement +40,3 (1,2) ; stratifié +19,5 (0,9) à h=5 ; singleton ΔΔ +1,7 (1,7) / réseau +51 (2,5) ; co-arrivées 3–5 : −8,3 (2,0) à h=30 ; SE par titre 4,7 ; signé jamais significatif ; différentiel par signe 66/63 ; contraste placebo par décennie +200/−41/−44/+68.
57 +
58 +## 8. Analyses recommandées, NON exécutées (données brutes ou externes requises)
59 +1. **Voisins sur corrélations résiduelles 5/30 min ou estimateur Hayashi–Yoshida** (priorité 1 : tranche entre « transmission plus forte » et « réseau mieux mesuré »).
60 +2. **Contraste réseau à t−5** (anticipation / nouvelles communes graduelles).
61 +3. Seuils source |z|≥4/5 ; seuil récepteur |z|>3 ; univers top-1000/3000.
62 +4. Placebos de réseau permutés préservant secteur/liquidité.
63 +5. Propagation dans les fenêtres d'annonces (objet séparé).
64 +6. Identification des ETF (security master externe) et exclusion ; composition top-1000 en 2020–21.
65 +7. Réconciliation des volumes avec le consolidated tape (venue).
66 +8. Tests formels de sauts (Lee–Mykland) et de co-sauts (Jacod–Todorov) sur un sous-échantillon ; modèle de Hawkes des minutes de co-saut.
67 +9. Appariement des placebos sur le niveau pré-événement (au lieu de la stratification ex post).
68 +10. Scan de nouveauté 2018–2026 à refaire juste avant soumission.
69 +
70 +## 9. Limites restantes
71 +Pas de cotes/profondeur/flux ; identité par ticker ; venue et licence inconnues ; discrétisation pré-2001 ; réseau identifié sur corrélations 1 min ; co-arrivées fréquentes ; série de co-sauts non attribuée (nouvelles vs endogène) ; non-déterminisme résiduel des SE dans les scripts 12/14/16 (±0,1 ×10⁻⁴).
72 +
73 +## 10. Maturité pour publication
74 +- **Revue cible : IRFA** — adéquation bonne (empirique, données massives, microstructure/contagion). Paquet `outputs/release/` régénéré (manuscrit 50 p., anonyme, highlights, lettre, page titre, annexe en ligne). Remerciements/financement à compléter par l'auteur ; scan de nouveauté à refaire ; vérifier le Guide for Authors du jour.
75 +- **Risques de rejet** : (a) un rapporteur exigera la robustesse de la hausse du canal réseau à l'identification des voisins (analyse 1 ci-dessus) — le papier l'annonce honnêtement mais ne la fournit pas ; (b) source de données non standard (pas TAQ/CRSP) et venue inconnue ; (c) 50 pages en elsarticle — envisager de déplacer Tables 9–10 ou les figures descriptives en annexe si la revue impose une limite ; (d) la question du titre reçoit une réponse nuancée (« more fragile » seulement au sens borné) — c'est une force scientifique mais un rapporteur pourrait juger le titre vendeur ; un sous-titre ou la conservation du point d'interrogation (actuel) suffit.
76 +- **Verdict** : publiable après révision mineure à modérée dans IRFA/JFM ; pour JFEC/JoE, l'analyse 1 et des tests formels de co-sauts seraient nécessaires. Ne pas déposer sur SSRN avant d'avoir fait l'analyse 1 si l'auteur souhaite maintenir le « triplement » comme résultat phare.
added outputs/diagnostics/bibliography/crossref_check_existing.json +2897 −0
@@ -0,0 +1,2897 @@
1 +[
2 + {
3 + "key": "chordia2000commonality",
4 + "doi": "10.1016/S0304-405X(99)00057-4",
5 + "bib_title": "Commonality in liquidity",
6 + "bib_year": "2000",
7 + "bib_journal": "Journal of Financial Economics",
8 + "cr_title": "Commonality in liquidity",
9 + "cr_journal": "Journal of Financial Economics",
10 + "cr_year_issued": 2000,
11 + "cr_year_print": 2000,
12 + "cr_authors": [
13 + "Chordia",
14 + "Roll",
15 + "Subrahmanyam"
16 + ],
17 + "cr_volume": "56",
18 + "cr_issue": "1",
19 + "cr_page": "3-28",
20 + "cr_doi": "10.1016/s0304-405x(99)00057-4",
21 + "bib_volume": "56",
22 + "bib_number": "1",
23 + "bib_pages": "3--28",
24 + "title_sim": 1.0,
25 + "first_author_ok": true,
26 + "year_ok": true,
27 + "status": "OK"
28 + },
29 + {
30 + "key": "hasbrouck2001common",
31 + "doi": "10.1016/S0304-405X(00)00091-X",
32 + "bib_title": "Common factors in prices, order flows, and liquidity",
33 + "bib_year": "2001",
34 + "bib_journal": "Journal of Financial Economics",
35 + "cr_title": "Common factors in prices, order flows, and liquidity",
36 + "cr_journal": "Journal of Financial Economics",
37 + "cr_year_issued": 2001,
38 + "cr_year_print": 2001,
39 + "cr_authors": [
40 + "Hasbrouck"
41 + ],
42 + "cr_volume": "59",
43 + "cr_issue": "3",
44 + "cr_page": "383-411",
45 + "cr_doi": "10.1016/s0304-405x(00)00091-x",
46 + "bib_volume": "59",
47 + "bib_number": "3",
48 + "bib_pages": "383--411",
49 + "title_sim": 1.0,
50 + "first_author_ok": true,
51 + "year_ok": true,
52 + "status": "OK"
53 + },
54 + {
55 + "key": "amihud2002illiquidity",
56 + "doi": "10.1016/S1386-4181(01)00024-6",
57 + "bib_title": "Illiquidity and stock returns: cross-section and time-series effects",
58 + "bib_year": "2002",
59 + "bib_journal": "Journal of Financial Markets",
60 + "cr_title": "Illiquidity and stock returns: cross-section and time-series effects",
61 + "cr_journal": "Journal of Financial Markets",
62 + "cr_year_issued": 2002,
63 + "cr_year_print": 2002,
64 + "cr_authors": [
65 + "Amihud"
66 + ],
67 + "cr_volume": "5",
68 + "cr_issue": "1",
69 + "cr_page": "31-56",
70 + "cr_doi": "10.1016/s1386-4181(01)00024-6",
71 + "bib_volume": "5",
72 + "bib_number": "1",
73 + "bib_pages": "31--56",
74 + "title_sim": 1.0,
75 + "first_author_ok": true,
76 + "year_ok": true,
77 + "status": "OK"
78 + },
79 + {
80 + "key": "corwin2012simple",
81 + "doi": "10.1111/j.1540-6261.2012.01729.x",
82 + "bib_title": "A Simple Way to Estimate Bid-Ask Spreads from Daily High and Low Prices",
83 + "bib_year": "2012",
84 + "bib_journal": "Journal of Finance",
85 + "cr_title": "A Simple Way to Estimate Bid\u2010Ask Spreads from Daily High and Low Prices",
86 + "cr_journal": "The Journal of Finance",
87 + "cr_year_issued": 2012,
88 + "cr_year_print": 2012,
89 + "cr_authors": [
90 + "CORWIN",
91 + "SCHULTZ"
92 + ],
93 + "cr_volume": "67",
94 + "cr_issue": "2",
95 + "cr_page": "719-760",
96 + "cr_doi": "10.1111/j.1540-6261.2012.01729.x",
97 + "bib_volume": "67",
98 + "bib_number": "2",
99 + "bib_pages": "719--760",
100 + "title_sim": 0.986,
101 + "first_author_ok": true,
102 + "year_ok": true,
103 + "status": "OK"
104 + },
105 + {
106 + "key": "roll1984simple",
107 + "doi": "10.1111/j.1540-6261.1984.tb03897.x",
108 + "bib_title": "A Simple Implicit Measure of the Effective Bid-Ask Spread in an Efficient Market",
109 + "bib_year": "1984",
110 + "bib_journal": "Journal of Finance",
111 + "cr_title": "A Simple Implicit Measure of the Effective Bid\u2010Ask Spread in an Efficient Market",
112 + "cr_journal": "The Journal of Finance",
113 + "cr_year_issued": 1984,
114 + "cr_year_print": 1984,
115 + "cr_authors": [
116 + "ROLL"
117 + ],
118 + "cr_volume": "39",
119 + "cr_issue": "4",
120 + "cr_page": "1127-1139",
121 + "cr_doi": "10.1111/j.1540-6261.1984.tb03897.x",
122 + "bib_volume": "39",
123 + "bib_number": "4",
124 + "bib_pages": "1127--1139",
125 + "title_sim": 0.988,
126 + "first_author_ok": true,
127 + "year_ok": true,
128 + "status": "OK"
129 + },
130 + {
131 + "key": "pastor2003liquidity",
132 + "doi": "10.1086/374184",
133 + "bib_title": "Liquidity Risk and Expected Stock Returns",
134 + "bib_year": "2003",
135 + "bib_journal": "Journal of Political Economy",
136 + "cr_title": "Liquidity Risk and Expected Stock Returns",
137 + "cr_journal": "Journal of Political Economy",
138 + "cr_year_issued": 2003,
139 + "cr_year_print": 2003,
140 + "cr_authors": [
141 + "P\u00e1stor",
142 + "Stambaugh"
143 + ],
144 + "cr_volume": "111",
145 + "cr_issue": "3",
146 + "cr_page": "642-685",
147 + "cr_doi": "10.1086/374184",
148 + "bib_volume": "111",
149 + "bib_number": "3",
150 + "bib_pages": "642--685",
151 + "title_sim": 1.0,
152 + "first_author_ok": false,
153 + "year_ok": true,
154 + "status": "CHECK"
155 + },
156 + {
157 + "key": "acharya2005asset",
158 + "doi": "10.1016/j.jfineco.2004.06.007",
159 + "bib_title": "Asset pricing with liquidity risk",
160 + "bib_year": "2005",
161 + "bib_journal": "Journal of Financial Economics",
162 + "cr_title": "Asset pricing with liquidity risk",
163 + "cr_journal": "Journal of Financial Economics",
164 + "cr_year_issued": 2005,
165 + "cr_year_print": 2005,
166 + "cr_authors": [
167 + "ACHARYA",
168 + "PEDERSEN"
169 + ],
170 + "cr_volume": "77",
171 + "cr_issue": "2",
172 + "cr_page": "375-410",
173 + "cr_doi": "10.1016/j.jfineco.2004.06.007",
174 + "bib_volume": "77",
175 + "bib_number": "2",
176 + "bib_pages": "375--410",
177 + "title_sim": 1.0,
178 + "first_author_ok": true,
179 + "year_ok": true,
180 + "status": "OK"
181 + },
182 + {
183 + "key": "karolyi2012understanding",
184 + "doi": "10.1016/j.jfineco.2011.12.008",
185 + "bib_title": "Understanding commonality in liquidity around the world",
186 + "bib_year": "2012",
187 + "bib_journal": "Journal of Financial Economics",
188 + "cr_title": "Understanding commonality in liquidity around the world",
189 + "cr_journal": "Journal of Financial Economics",
190 + "cr_year_issued": 2012,
191 + "cr_year_print": 2012,
192 + "cr_authors": [
193 + "Karolyi",
194 + "Lee",
195 + "van Dijk"
196 + ],
197 + "cr_volume": "105",
198 + "cr_issue": "1",
199 + "cr_page": "82-112",
200 + "cr_doi": "10.1016/j.jfineco.2011.12.008",
201 + "bib_volume": "105",
202 + "bib_number": "1",
203 + "bib_pages": "82--112",
204 + "title_sim": 1.0,
205 + "first_author_ok": true,
206 + "year_ok": true,
207 + "status": "OK"
208 + },
209 + {
210 + "key": "parkinson1980extreme",
211 + "doi": "10.1086/296071",
212 + "bib_title": "The Extreme Value Method for Estimating the Variance of the Rate of Return",
213 + "bib_year": "1980",
214 + "bib_journal": "Journal of Business",
215 + "cr_title": "The Extreme Value Method for Estimating the Variance of the Rate of Return",
216 + "cr_journal": "The Journal of Business",
217 + "cr_year_issued": 1980,
218 + "cr_year_print": 1980,
219 + "cr_authors": [
220 + "Parkinson"
221 + ],
222 + "cr_volume": "53",
223 + "cr_issue": "1",
224 + "cr_page": "61",
225 + "cr_doi": "10.1086/296071",
226 + "bib_volume": "53",
227 + "bib_number": "1",
228 + "bib_pages": "61--65",
229 + "title_sim": 1.0,
230 + "first_author_ok": true,
231 + "year_ok": true,
232 + "status": "OK"
233 + },
234 + {
235 + "key": "garman1980estimation",
236 + "doi": "10.1086/296072",
237 + "bib_title": "On the Estimation of Security Price Volatilities from Historical Data",
238 + "bib_year": "1980",
239 + "bib_journal": "Journal of Business",
240 + "cr_title": "On the Estimation of Security Price Volatilities from Historical Data",
241 + "cr_journal": "The Journal of Business",
242 + "cr_year_issued": 1980,
243 + "cr_year_print": 1980,
244 + "cr_authors": [
245 + "Garman",
246 + "Klass"
247 + ],
248 + "cr_volume": "53",
249 + "cr_issue": "1",
250 + "cr_page": "67",
251 + "cr_doi": "10.1086/296072",
252 + "bib_volume": "53",
253 + "bib_number": "1",
254 + "bib_pages": "67--78",
255 + "title_sim": 1.0,
256 + "first_author_ok": true,
257 + "year_ok": true,
258 + "status": "OK"
259 + },
260 + {
261 + "key": "goyenko2009liquidity",
262 + "doi": "10.1016/j.jfineco.2008.06.002",
263 + "bib_title": "Do liquidity measures measure liquidity?",
264 + "bib_year": "2009",
265 + "bib_journal": "Journal of Financial Economics",
266 + "cr_title": "Do liquidity measures measure liquidity?\u2606",
267 + "cr_journal": "Journal of Financial Economics",
268 + "cr_year_issued": 2009,
269 + "cr_year_print": 2009,
270 + "cr_authors": [
271 + "Goyenko",
272 + "Holden",
273 + "Trzcinka"
274 + ],
275 + "cr_volume": "92",
276 + "cr_issue": "2",
277 + "cr_page": "153-181",
278 + "cr_doi": "10.1016/j.jfineco.2008.06.002",
279 + "bib_volume": "92",
280 + "bib_number": "2",
281 + "bib_pages": "153--181",
282 + "title_sim": 0.988,
283 + "first_author_ok": true,
284 + "year_ok": true,
285 + "status": "OK"
286 + },
287 + {
288 + "key": "hendershott2011algorithmic",
289 + "doi": "10.1111/j.1540-6261.2010.01624.x",
290 + "bib_title": "Does Algorithmic Trading Improve Liquidity?",
291 + "bib_year": "2011",
292 + "bib_journal": "Journal of Finance",
293 + "cr_title": "Does Algorithmic Trading Improve Liquidity?",
294 + "cr_journal": "The Journal of Finance",
295 + "cr_year_issued": 2011,
296 + "cr_year_print": 2011,
297 + "cr_authors": [
298 + "HENDERSHOTT",
299 + "JONES",
300 + "MENKVELD"
301 + ],
302 + "cr_volume": "66",
303 + "cr_issue": "1",
304 + "cr_page": "1-33",
305 + "cr_doi": "10.1111/j.1540-6261.2010.01624.x",
306 + "bib_volume": "66",
307 + "bib_number": "1",
308 + "bib_pages": "1--33",
309 + "title_sim": 1.0,
310 + "first_author_ok": true,
311 + "year_ok": true,
312 + "status": "OK"
313 + },
314 + {
315 + "key": "kirilenko2017flash",
316 + "doi": "10.1111/jofi.12498",
317 + "bib_title": "The Flash Crash: High-Frequency Trading in an Electronic Market",
318 + "bib_year": "2017",
319 + "bib_journal": "Journal of Finance",
320 + "cr_title": "The Flash Crash: High\u2010Frequency Trading in an Electronic Market",
321 + "cr_journal": "The Journal of Finance",
322 + "cr_year_issued": 2017,
323 + "cr_year_print": 2017,
324 + "cr_authors": [
325 + "KIRILENKO",
326 + "KYLE",
327 + "SAMADI",
328 + "TUZUN"
329 + ],
330 + "cr_volume": "72",
331 + "cr_issue": "3",
332 + "cr_page": "967-998",
333 + "cr_doi": "10.1111/jofi.12498",
334 + "bib_volume": "72",
335 + "bib_number": "3",
336 + "bib_pages": "967--998",
337 + "title_sim": 0.984,
338 + "first_author_ok": true,
339 + "year_ok": true,
340 + "status": "OK"
341 + },
342 + {
343 + "key": "menkveld2013high",
344 + "doi": "10.1016/j.finmar.2013.06.006",
345 + "bib_title": "High Frequency Trading and the New Market Makers",
346 + "bib_year": "2013",
347 + "bib_journal": "Journal of Financial Markets",
348 + "cr_title": "High frequency trading and the new market makers",
349 + "cr_journal": "Journal of Financial Markets",
350 + "cr_year_issued": 2013,
351 + "cr_year_print": 2013,
352 + "cr_authors": [
353 + "Menkveld"
354 + ],
355 + "cr_volume": "16",
356 + "cr_issue": "4",
357 + "cr_page": "712-740",
358 + "cr_doi": "10.1016/j.finmar.2013.06.006",
359 + "bib_volume": "16",
360 + "bib_number": "4",
361 + "bib_pages": "712--740",
362 + "title_sim": 1.0,
363 + "first_author_ok": true,
364 + "year_ok": true,
365 + "status": "OK"
366 + },
367 + {
368 + "key": "brogaard2014high",
369 + "doi": "10.1093/rfs/hhu032",
370 + "bib_title": "High-Frequency Trading and Price Discovery",
371 + "bib_year": "2014",
372 + "bib_journal": "Review of Financial Studies",
373 + "cr_title": "High-Frequency Trading and Price Discovery",
374 + "cr_journal": "Review of Financial Studies",
375 + "cr_year_issued": 2014,
376 + "cr_year_print": 2014,
377 + "cr_authors": [
378 + "Brogaard",
379 + "Hendershott",
380 + "Riordan"
381 + ],
382 + "cr_volume": "27",
383 + "cr_issue": "8",
384 + "cr_page": "2267-2306",
385 + "cr_doi": "10.1093/rfs/hhu032",
386 + "bib_volume": "27",
387 + "bib_number": "8",
388 + "bib_pages": "2267--2306",
389 + "title_sim": 1.0,
390 + "first_author_ok": true,
391 + "year_ok": true,
392 + "status": "OK"
393 + },
394 + {
395 + "key": "budish2015high",
396 + "doi": "10.1093/qje/qjv027",
397 + "bib_title": "The High-Frequency Trading Arms Race: Frequent Batch Auctions as a Market Design Response",
398 + "bib_year": "2015",
399 + "bib_journal": "Quarterly Journal of Economics",
400 + "cr_title": "The High-Frequency Trading Arms Race: Frequent Batch Auctions as a Market Design Response *",
401 + "cr_journal": "The Quarterly Journal of Economics",
402 + "cr_year_issued": 2015,
403 + "cr_year_print": 2015,
404 + "cr_authors": [
405 + "Budish",
406 + "Cramton",
407 + "Shim"
408 + ],
409 + "cr_volume": "130",
410 + "cr_issue": "4",
411 + "cr_page": "1547-1621",
412 + "cr_doi": "10.1093/qje/qjv027",
413 + "bib_volume": "130",
414 + "bib_number": "4",
415 + "bib_pages": "1547--1621",
416 + "title_sim": 0.989,
417 + "first_author_ok": true,
418 + "year_ok": true,
419 + "status": "OK"
420 + },
421 + {
422 + "key": "ohara2015high",
423 + "doi": "10.1016/j.jfineco.2015.01.003",
424 + "bib_title": "High Frequency Market Microstructure",
425 + "bib_year": "2015",
426 + "bib_journal": "Journal of Financial Economics",
427 + "cr_title": "High frequency market microstructure",
428 + "cr_journal": "Journal of Financial Economics",
429 + "cr_year_issued": 2015,
430 + "cr_year_print": 2015,
431 + "cr_authors": [
432 + "O\u2019Hara"
433 + ],
434 + "cr_volume": "116",
435 + "cr_issue": "2",
436 + "cr_page": "257-270",
437 + "cr_doi": "10.1016/j.jfineco.2015.01.003",
438 + "bib_volume": "116",
439 + "bib_number": "2",
440 + "bib_pages": "257--270",
441 + "title_sim": 1.0,
442 + "first_author_ok": false,
443 + "year_ok": true,
444 + "status": "CHECK"
445 + },
446 + {
447 + "key": "bendavid2018etfs",
448 + "doi": "10.1111/jofi.12727",
449 + "bib_title": "Do {ETFs} Increase Volatility?",
450 + "bib_year": "2018",
451 + "bib_journal": "Journal of Finance",
452 + "cr_title": "Do ETFs Increase Volatility?",
453 + "cr_journal": "The Journal of Finance",
454 + "cr_year_issued": 2018,
455 + "cr_year_print": 2018,
456 + "cr_authors": [
457 + "BEN\u2010DAVID",
458 + "FRANZONI",
459 + "MOUSSAWI"
460 + ],
461 + "cr_volume": "73",
462 + "cr_issue": "6",
463 + "cr_page": "2471-2535",
464 + "cr_doi": "10.1111/jofi.12727",
465 + "bib_volume": "73",
466 + "bib_number": "6",
467 + "bib_pages": "2471--2535",
468 + "title_sim": 1.0,
469 + "first_author_ok": false,
470 + "year_ok": true,
471 + "status": "CHECK"
472 + },
473 + {
474 + "key": "madhavan2012exchange",
475 + "doi": "10.2469/faj.v68.n4.6",
476 + "bib_title": "Exchange-Traded Funds, Market Structure, and the Flash Crash",
477 + "bib_year": "2012",
478 + "bib_journal": "Financial Analysts Journal",
479 + "cr_title": "Exchange-Traded Funds, Market Structure, and the Flash Crash",
480 + "cr_journal": "Financial Analysts Journal",
481 + "cr_year_issued": 2012,
482 + "cr_year_print": 2012,
483 + "cr_authors": [
484 + "Madhavan"
485 + ],
486 + "cr_volume": "68",
487 + "cr_issue": "4",
488 + "cr_page": "20-35",
489 + "cr_doi": "10.2469/faj.v68.n4.6",
490 + "bib_volume": "68",
491 + "bib_number": "4",
492 + "bib_pages": "20--35",
493 + "title_sim": 1.0,
494 + "first_author_ok": true,
495 + "year_ok": true,
496 + "status": "OK"
497 + },
498 + {
499 + "key": "boehmer2021algorithmic",
500 + "doi": "10.1017/S0022109020000782",
501 + "bib_title": "Algorithmic Trading and Market Quality: International Evidence",
502 + "bib_year": "2021",
503 + "bib_journal": "Journal of Financial and Quantitative Analysis",
504 + "cr_title": "Algorithmic Trading and Market Quality: International Evidence",
505 + "cr_journal": "Journal of Financial and Quantitative Analysis",
506 + "cr_year_issued": 2020,
507 + "cr_year_print": 2021,
508 + "cr_authors": [
509 + "Boehmer",
510 + "Fong",
511 + "Wu"
512 + ],
513 + "cr_volume": "56",
514 + "cr_issue": "8",
515 + "cr_page": "2659-2688",
516 + "cr_doi": "10.1017/s0022109020000782",
517 + "bib_volume": "56",
518 + "bib_number": "8",
519 + "bib_pages": "2659--2688",
520 + "title_sim": 1.0,
521 + "first_author_ok": true,
522 + "year_ok": true,
523 + "status": "OK"
524 + },
525 + {
526 + "key": "hasbrouck2013low",
527 + "doi": "10.1016/j.finmar.2013.05.003",
528 + "bib_title": "Low-Latency Trading",
529 + "bib_year": "2013",
530 + "bib_journal": "Journal of Financial Markets",
531 + "cr_title": "Low-latency trading",
532 + "cr_journal": "Journal of Financial Markets",
533 + "cr_year_issued": 2013,
534 + "cr_year_print": 2013,
535 + "cr_authors": [
536 + "Hasbrouck",
537 + "Saar"
538 + ],
539 + "cr_volume": "16",
540 + "cr_issue": "4",
541 + "cr_page": "646-679",
542 + "cr_doi": "10.1016/j.finmar.2013.05.003",
543 + "bib_volume": "16",
544 + "bib_number": "4",
545 + "bib_pages": "646--679",
546 + "title_sim": 1.0,
547 + "first_author_ok": true,
548 + "year_ok": true,
549 + "status": "OK"
550 + },
551 + {
552 + "key": "baldauf2020high",
553 + "doi": "10.1111/jofi.12882",
554 + "bib_title": "High-Frequency Trading and Market Performance",
555 + "bib_year": "2020",
556 + "bib_journal": "Journal of Finance",
557 + "cr_title": "High\u2010Frequency Trading and Market Performance",
558 + "cr_journal": "The Journal of Finance",
559 + "cr_year_issued": 2020,
560 + "cr_year_print": 2020,
561 + "cr_authors": [
562 + "BALDAUF",
563 + "MOLLNER"
564 + ],
565 + "cr_volume": "75",
566 + "cr_issue": "3",
567 + "cr_page": "1495-1526",
568 + "cr_doi": "10.1111/jofi.12882",
569 + "bib_volume": "75",
570 + "bib_number": "3",
571 + "bib_pages": "1495--1526",
572 + "title_sim": 0.978,
573 + "first_author_ok": true,
574 + "year_ok": true,
575 + "status": "OK"
576 + },
577 + {
578 + "key": "diebold2014network",
579 + "doi": "10.1016/j.jeconom.2014.04.012",
580 + "bib_title": "On the Network Topology of Variance Decompositions: Measuring the Connectedness of Financial Firms",
581 + "bib_year": "2014",
582 + "bib_journal": "Journal of Econometrics",
583 + "cr_title": "On the network topology of variance decompositions: Measuring the connectedness of financial firms",
584 + "cr_journal": "Journal of Econometrics",
585 + "cr_year_issued": 2014,
586 + "cr_year_print": 2014,
587 + "cr_authors": [
588 + "Diebold",
589 + "Y\u0131lmaz"
590 + ],
591 + "cr_volume": "182",
592 + "cr_issue": "1",
593 + "cr_page": "119-134",
594 + "cr_doi": "10.1016/j.jeconom.2014.04.012",
595 + "bib_volume": "182",
596 + "bib_number": "1",
597 + "bib_pages": "119--134",
598 + "title_sim": 1.0,
599 + "first_author_ok": true,
600 + "year_ok": true,
601 + "status": "OK"
602 + },
603 + {
604 + "key": "diebold2012better",
605 + "doi": "10.1016/j.ijforecast.2011.02.006",
606 + "bib_title": "Better to Give than to Receive: Predictive Directional Measurement of Volatility Spillovers",
607 + "bib_year": "2012",
608 + "bib_journal": "International Journal of Forecasting",
609 + "cr_title": "Better to give than to receive: Predictive directional measurement of volatility spillovers",
610 + "cr_journal": "International Journal of Forecasting",
611 + "cr_year_issued": 2012,
612 + "cr_year_print": 2012,
613 + "cr_authors": [
614 + "Diebold",
615 + "Yilmaz"
616 + ],
617 + "cr_volume": "28",
618 + "cr_issue": "1",
619 + "cr_page": "57-66",
620 + "cr_doi": "10.1016/j.ijforecast.2011.02.006",
621 + "bib_volume": "28",
622 + "bib_number": "1",
623 + "bib_pages": "57--66",
624 + "title_sim": 1.0,
625 + "first_author_ok": true,
626 + "year_ok": true,
627 + "status": "OK"
628 + },
629 + {
630 + "key": "billio2012econometric",
631 + "doi": "10.1016/j.jfineco.2011.12.010",
632 + "bib_title": "Econometric Measures of Connectedness and Systemic Risk in the Finance and Insurance Sectors",
633 + "bib_year": "2012",
634 + "bib_journal": "Journal of Financial Economics",
635 + "cr_title": "Econometric measures of connectedness and systemic risk in the finance and insurance sectors",
636 + "cr_journal": "Journal of Financial Economics",
637 + "cr_year_issued": 2012,
638 + "cr_year_print": 2012,
639 + "cr_authors": [
640 + "Billio",
641 + "Getmansky",
642 + "Lo",
643 + "Pelizzon"
644 + ],
645 + "cr_volume": "104",
646 + "cr_issue": "3",
647 + "cr_page": "535-559",
648 + "cr_doi": "10.1016/j.jfineco.2011.12.010",
649 + "bib_volume": "104",
650 + "bib_number": "3",
651 + "bib_pages": "535--559",
652 + "title_sim": 1.0,
653 + "first_author_ok": true,
654 + "year_ok": true,
655 + "status": "OK"
656 + },
657 + {
658 + "key": "barigozzi2019nets",
659 + "doi": "10.1002/jae.2676",
660 + "bib_title": "{NETS}: Network Estimation for Time Series",
661 + "bib_year": "2019",
662 + "bib_journal": "Journal of Applied Econometrics",
663 + "cr_title": "NETS: Network estimation for time series",
664 + "cr_journal": "Journal of Applied Econometrics",
665 + "cr_year_issued": 2019,
666 + "cr_year_print": 2019,
667 + "cr_authors": [
668 + "Barigozzi",
669 + "Brownlees"
670 + ],
671 + "cr_volume": "34",
672 + "cr_issue": "3",
673 + "cr_page": "347-364",
674 + "cr_doi": "10.1002/jae.2676",
675 + "bib_volume": "34",
676 + "bib_number": "3",
677 + "bib_pages": "347--364",
678 + "title_sim": 1.0,
679 + "first_author_ok": true,
680 + "year_ok": true,
681 + "status": "OK"
682 + },
683 + {
684 + "key": "acemoglu2012network",
685 + "doi": "10.3982/ECTA9623",
686 + "bib_title": "The Network Origins of Aggregate Fluctuations",
687 + "bib_year": "2012",
688 + "bib_journal": "Econometrica",
689 + "cr_title": "The Network Origins of Aggregate Fluctuations",
690 + "cr_journal": "Econometrica",
691 + "cr_year_issued": 2012,
692 + "cr_year_print": null,
693 + "cr_authors": [],
694 + "cr_volume": "80",
695 + "cr_issue": "5",
696 + "cr_page": "1977-2016",
697 + "cr_doi": "10.3982/ecta9623",
698 + "bib_volume": "80",
699 + "bib_number": "5",
700 + "bib_pages": "1977--2016",
701 + "title_sim": 1.0,
702 + "first_author_ok": false,
703 + "year_ok": true,
704 + "status": "CHECK"
705 + },
706 + {
707 + "key": "aitsahalia2015modeling",
708 + "doi": "10.1016/j.jfineco.2015.03.002",
709 + "bib_title": "Modeling Financial Contagion Using Mutually Exciting Jump Processes",
710 + "bib_year": "2015",
711 + "bib_journal": "Journal of Financial Economics",
712 + "cr_title": "Modeling financial contagion using mutually exciting jump processes",
713 + "cr_journal": "Journal of Financial Economics",
714 + "cr_year_issued": 2015,
715 + "cr_year_print": 2015,
716 + "cr_authors": [
717 + "A\u00eft-Sahalia",
718 + "Cacho-Diaz",
719 + "Laeven"
720 + ],
721 + "cr_volume": "117",
722 + "cr_issue": "3",
723 + "cr_page": "585-606",
724 + "cr_doi": "10.1016/j.jfineco.2015.03.002",
725 + "bib_volume": "117",
726 + "bib_number": "3",
727 + "bib_pages": "585--606",
728 + "title_sim": 1.0,
729 + "first_author_ok": false,
730 + "year_ok": true,
731 + "status": "CHECK"
732 + },
733 + {
734 + "key": "bacry2015hawkes",
735 + "doi": "10.1142/S2382626615500057",
736 + "bib_title": "Hawkes Processes in Finance",
737 + "bib_year": "2015",
738 + "bib_journal": "Market Microstructure and Liquidity",
739 + "cr_title": "Hawkes Processes in Finance",
740 + "cr_journal": "Market Microstructure and Liquidity",
741 + "cr_year_issued": 2015,
742 + "cr_year_print": 2015,
743 + "cr_authors": [
744 + "Bacry",
745 + "Mastromatteo",
746 + "Muzy"
747 + ],
748 + "cr_volume": "01",
749 + "cr_issue": "01",
750 + "cr_page": "1550005",
751 + "cr_doi": "10.1142/s2382626615500057",
752 + "bib_volume": "1",
753 + "bib_number": "1",
754 + "bib_pages": "1550005",
755 + "title_sim": 1.0,
756 + "first_author_ok": true,
757 + "year_ok": true,
758 + "status": "OK"
759 + },
760 + {
761 + "key": "hawkes1971spectra",
762 + "doi": "10.1093/biomet/58.1.83",
763 + "bib_title": "Spectra of Some Self-Exciting and Mutually Exciting Point Processes",
764 + "bib_year": "1971",
765 + "bib_journal": "Biometrika",
766 + "cr_title": "Spectra of some self-exciting and mutually exciting point processes",
767 + "cr_journal": "Biometrika",
768 + "cr_year_issued": 1971,
769 + "cr_year_print": 1971,
770 + "cr_authors": [
771 + "HAWKES"
772 + ],
773 + "cr_volume": "58",
774 + "cr_issue": "1",
775 + "cr_page": "83-90",
776 + "cr_doi": "10.1093/biomet/58.1.83",
777 + "bib_volume": "58",
778 + "bib_number": "1",
779 + "bib_pages": "83--90",
780 + "title_sim": 1.0,
781 + "first_author_ok": true,
782 + "year_ok": true,
783 + "status": "OK"
784 + },
785 + {
786 + "key": "elliott2014financial",
787 + "doi": "10.1257/aer.104.10.3115",
788 + "bib_title": "Financial Networks and Contagion",
789 + "bib_year": "2014",
790 + "bib_journal": "American Economic Review",
791 + "cr_title": "Financial Networks and Contagion",
792 + "cr_journal": "American Economic Review",
793 + "cr_year_issued": 2014,
794 + "cr_year_print": 2014,
795 + "cr_authors": [
796 + "Elliott",
797 + "Golub",
798 + "Jackson"
799 + ],
800 + "cr_volume": "104",
801 + "cr_issue": "10",
802 + "cr_page": "3115-3153",
803 + "cr_doi": "10.1257/aer.104.10.3115",
804 + "bib_volume": "104",
805 + "bib_number": "10",
806 + "bib_pages": "3115--3153",
807 + "title_sim": 1.0,
808 + "first_author_ok": true,
809 + "year_ok": true,
810 + "status": "OK"
811 + },
812 + {
813 + "key": "forbes2002no",
814 + "doi": "10.1111/0022-1082.00494",
815 + "bib_title": "No Contagion, Only Interdependence: Measuring Stock Market Comovements",
816 + "bib_year": "2002",
817 + "bib_journal": "Journal of Finance",
818 + "cr_title": "No Contagion, Only Interdependence: Measuring Stock Market Comovements",
819 + "cr_journal": "The Journal of Finance",
820 + "cr_year_issued": 2002,
821 + "cr_year_print": 2002,
822 + "cr_authors": [
823 + "Forbes",
824 + "Rigobon"
825 + ],
826 + "cr_volume": "57",
827 + "cr_issue": "5",
828 + "cr_page": "2223-2261",
829 + "cr_doi": "10.1111/0022-1082.00494",
830 + "bib_volume": "57",
831 + "bib_number": "5",
832 + "bib_pages": "2223--2261",
833 + "title_sim": 1.0,
834 + "first_author_ok": true,
835 + "year_ok": true,
836 + "status": "OK"
837 + },
838 + {
839 + "key": "hautsch2015financial",
840 + "doi": "10.1093/rof/rfu010",
841 + "bib_title": "Financial Network Systemic Risk Contributions",
842 + "bib_year": "2015",
843 + "bib_journal": "Review of Finance",
844 + "cr_title": "Financial Network Systemic Risk Contributions",
845 + "cr_journal": "Review of Finance",
846 + "cr_year_issued": 2014,
847 + "cr_year_print": 2015,
848 + "cr_authors": [
849 + "Hautsch",
850 + "Schaumburg",
851 + "Schienle"
852 + ],
853 + "cr_volume": "19",
854 + "cr_issue": "2",
855 + "cr_page": "685-738",
856 + "cr_doi": "10.1093/rof/rfu010",
857 + "bib_volume": "19",
858 + "bib_number": "2",
859 + "bib_pages": "685--738",
860 + "title_sim": 1.0,
861 + "first_author_ok": true,
862 + "year_ok": true,
863 + "status": "OK"
864 + },
865 + {
866 + "key": "bekaert2014global",
867 + "doi": "10.1111/jofi.12203",
868 + "bib_title": "The Global Crisis and Equity Market Contagion",
869 + "bib_year": "2014",
870 + "bib_journal": "Journal of Finance",
871 + "cr_title": "The Global Crisis and Equity Market Contagion",
872 + "cr_journal": "The Journal of Finance",
873 + "cr_year_issued": 2014,
874 + "cr_year_print": 2014,
875 + "cr_authors": [
876 + "BEKAERT",
877 + "EHRMANN",
878 + "FRATZSCHER",
879 + "MEHL"
880 + ],
881 + "cr_volume": "69",
882 + "cr_issue": "6",
883 + "cr_page": "2597-2649",
884 + "cr_doi": "10.1111/jofi.12203",
885 + "bib_volume": "69",
886 + "bib_number": "6",
887 + "bib_pages": "2597--2649",
888 + "title_sim": 1.0,
889 + "first_author_ok": true,
890 + "year_ok": true,
891 + "status": "OK"
892 + },
893 + {
894 + "key": "lo1990contrarian",
895 + "doi": "10.1093/rfs/3.2.175",
896 + "bib_title": "When Are Contrarian Profits Due to Stock Market Overreaction?",
897 + "bib_year": "1990",
898 + "bib_journal": "Review of Financial Studies",
899 + "cr_title": "When Are Contrarian Profits Due to Stock Market Overreaction?",
900 + "cr_journal": "Review of Financial Studies",
901 + "cr_year_issued": 1990,
902 + "cr_year_print": 1990,
903 + "cr_authors": [
904 + "Lo",
905 + "MacKinlay"
906 + ],
907 + "cr_volume": "3",
908 + "cr_issue": "2",
909 + "cr_page": "175-205",
910 + "cr_doi": "10.1093/rfs/3.2.175",
911 + "bib_volume": "3",
912 + "bib_number": "2",
913 + "bib_pages": "175--205",
914 + "title_sim": 1.0,
915 + "first_author_ok": true,
916 + "year_ok": true,
917 + "status": "OK"
918 + },
919 + {
920 + "key": "hou2007industry",
921 + "doi": "10.1093/revfin/hhm003",
922 + "bib_title": "Industry Information Diffusion and the Lead-Lag Effect in Stock Returns",
923 + "bib_year": "2007",
924 + "bib_journal": "Review of Financial Studies",
925 + "cr_title": "Industry Information Diffusion and the Lead-lag Effect in Stock Returns",
926 + "cr_journal": "Review of Financial Studies",
927 + "cr_year_issued": 2007,
928 + "cr_year_print": 2007,
929 + "cr_authors": [
930 + "Hou"
931 + ],
932 + "cr_volume": "20",
933 + "cr_issue": "4",
934 + "cr_page": "1113-1138",
935 + "cr_doi": "10.1093/revfin/hhm003",
936 + "bib_volume": "20",
937 + "bib_number": "4",
938 + "bib_pages": "1113--1138",
939 + "title_sim": 1.0,
940 + "first_author_ok": true,
941 + "year_ok": true,
942 + "status": "OK"
943 + },
944 + {
945 + "key": "andersen1997intraday",
946 + "doi": "10.1016/S0927-5398(97)00004-2",
947 + "bib_title": "Intraday Periodicity and Volatility Persistence in Financial Markets",
948 + "bib_year": "1997",
949 + "bib_journal": "Journal of Empirical Finance",
950 + "cr_title": "Intraday periodicity and volatility persistence in financial markets",
951 + "cr_journal": "Journal of Empirical Finance",
952 + "cr_year_issued": 1997,
953 + "cr_year_print": 1997,
954 + "cr_authors": [
955 + "Andersen",
956 + "Bollerslev"
957 + ],
958 + "cr_volume": "4",
959 + "cr_issue": "2-3",
960 + "cr_page": "115-158",
961 + "cr_doi": "10.1016/s0927-5398(97)00004-2",
962 + "bib_volume": "4",
963 + "bib_number": "2--3",
964 + "bib_pages": "115--158",
965 + "title_sim": 1.0,
966 + "first_author_ok": true,
967 + "year_ok": true,
968 + "status": "OK"
969 + },
970 + {
971 + "key": "lee2008jumps",
972 + "doi": "10.1093/rfs/hhm056",
973 + "bib_title": "Jumps in Financial Markets: A New Nonparametric Test and Jump Dynamics",
974 + "bib_year": "2008",
975 + "bib_journal": "Review of Financial Studies",
976 + "cr_title": "Jumps in Financial Markets: A New Nonparametric Test and Jump Dynamics",
977 + "cr_journal": "Review of Financial Studies",
978 + "cr_year_issued": 2007,
979 + "cr_year_print": 2008,
980 + "cr_authors": [
981 + "Lee",
982 + "Mykland"
983 + ],
984 + "cr_volume": "21",
985 + "cr_issue": "6",
986 + "cr_page": "2535-2563",
987 + "cr_doi": "10.1093/rfs/hhm056",
988 + "bib_volume": "21",
989 + "bib_number": "6",
990 + "bib_pages": "2535--2563",
991 + "title_sim": 1.0,
992 + "first_author_ok": true,
993 + "year_ok": true,
994 + "status": "OK"
995 + },
996 + {
997 + "key": "barndorffnielsen2004power",
998 + "doi": "10.1093/jjfinec/nbh001",
999 + "bib_title": "Power and Bipower Variation with Stochastic Volatility and Jumps",
1000 + "bib_year": "2004",
1001 + "bib_journal": "Journal of Financial Econometrics",
1002 + "cr_title": "Power and Bipower Variation with Stochastic Volatility and Jumps",
1003 + "cr_journal": "Journal of Financial Econometrics",
1004 + "cr_year_issued": 2004,
1005 + "cr_year_print": 2004,
1006 + "cr_authors": [
1007 + "Barndorff-Nielsen"
1008 + ],
1009 + "cr_volume": "2",
1010 + "cr_issue": "1",
1011 + "cr_page": "1-37",
1012 + "cr_doi": "10.1093/jjfinec/nbh001",
1013 + "bib_volume": "2",
1014 + "bib_number": "1",
1015 + "bib_pages": "1--37",
1016 + "title_sim": 1.0,
1017 + "first_author_ok": true,
1018 + "year_ok": true,
1019 + "status": "OK"
1020 + },
1021 + {
1022 + "key": "aitsahalia2009testing",
1023 + "doi": "10.1214/07-AOS568",
1024 + "bib_title": "Testing for Jumps in a Discretely Observed Process",
1025 + "bib_year": "2009",
1026 + "bib_journal": "Annals of Statistics",
1027 + "cr_title": "Testing for jumps in a discretely observed process",
1028 + "cr_journal": "The Annals of Statistics",
1029 + "cr_year_issued": 2009,
1030 + "cr_year_print": 2009,
1031 + "cr_authors": [
1032 + "A\u00eft-Sahalia",
1033 + "Jacod"
1034 + ],
1035 + "cr_volume": "37",
1036 + "cr_issue": "1",
1037 + "cr_page": "184-222",
1038 + "cr_doi": "10.1214/07-aos568",
1039 + "bib_volume": "37",
1040 + "bib_number": "1",
1041 + "bib_pages": "184--222",
1042 + "title_sim": 1.0,
1043 + "first_author_ok": false,
1044 + "year_ok": true,
1045 + "status": "CHECK"
1046 + },
1047 + {
1048 + "key": "gu2020empirical",
1049 + "doi": "10.1093/rfs/hhaa009",
1050 + "bib_title": "Empirical Asset Pricing via Machine Learning",
1051 + "bib_year": "2020",
1052 + "bib_journal": "Review of Financial Studies",
1053 + "cr_title": "Empirical Asset Pricing via Machine Learning",
1054 + "cr_journal": "The Review of Financial Studies",
1055 + "cr_year_issued": 2020,
1056 + "cr_year_print": 2020,
1057 + "cr_authors": [
1058 + "Gu",
1059 + "Kelly",
1060 + "Xiu"
1061 + ],
1062 + "cr_volume": "33",
1063 + "cr_issue": "5",
1064 + "cr_page": "2223-2273",
1065 + "cr_doi": "10.1093/rfs/hhaa009",
1066 + "bib_volume": "33",
1067 + "bib_number": "5",
1068 + "bib_pages": "2223--2273",
1069 + "title_sim": 1.0,
1070 + "first_author_ok": true,
1071 + "year_ok": true,
1072 + "status": "OK"
1073 + },
1074 + {
1075 + "key": "easley2012flow",
1076 + "doi": "10.1093/rfs/hhs053",
1077 + "bib_title": "Flow Toxicity and Liquidity in a High-Frequency World",
1078 + "bib_year": "2012",
1079 + "bib_journal": "Review of Financial Studies",
1080 + "cr_title": "Flow Toxicity and Liquidity in a High-frequency World",
1081 + "cr_journal": "Review of Financial Studies",
1082 + "cr_year_issued": 2012,
1083 + "cr_year_print": 2012,
1084 + "cr_authors": [
1085 + "Easley",
1086 + "L\u00f3pez de Prado",
1087 + "O'Hara"
1088 + ],
1089 + "cr_volume": "25",
1090 + "cr_issue": "5",
1091 + "cr_page": "1457-1493",
1092 + "cr_doi": "10.1093/rfs/hhs053",
1093 + "bib_volume": "25",
1094 + "bib_number": "5",
1095 + "bib_pages": "1457--1493",
1096 + "title_sim": 1.0,
1097 + "first_author_ok": true,
1098 + "year_ok": true,
1099 + "status": "OK"
1100 + },
1101 + {
1102 + "key": "andersen2014vpin",
1103 + "doi": "10.1016/j.finmar.2013.05.005",
1104 + "bib_title": "{VPIN} and the Flash Crash",
1105 + "bib_year": "2014",
1106 + "bib_journal": "Journal of Financial Markets",
1107 + "cr_title": "VPIN and the flash crash",
1108 + "cr_journal": "Journal of Financial Markets",
1109 + "cr_year_issued": 2014,
1110 + "cr_year_print": 2014,
1111 + "cr_authors": [
1112 + "Andersen",
1113 + "Bondarenko"
1114 + ],
1115 + "cr_volume": "17",
1116 + "cr_issue": null,
1117 + "cr_page": "1-46",
1118 + "cr_doi": "10.1016/j.finmar.2013.05.005",
1119 + "bib_volume": "17",
1120 + "bib_number": "",
1121 + "bib_pages": "1--46",
1122 + "title_sim": 1.0,
1123 + "first_author_ok": true,
1124 + "year_ok": true,
1125 + "status": "OK"
1126 + },
1127 + {
1128 + "key": "epps1979comovements",
1129 + "doi": "10.1080/01621459.1979.10482508",
1130 + "bib_title": "Comovements in Stock Prices in the Very Short Run",
1131 + "bib_year": "1979",
1132 + "bib_journal": "Journal of the American Statistical Association",
1133 + "cr_title": "Comovements in Stock Prices in the Very Short Run",
1134 + "cr_journal": "Journal of the American Statistical Association",
1135 + "cr_year_issued": 1979,
1136 + "cr_year_print": 1979,
1137 + "cr_authors": [
1138 + "Epps"
1139 + ],
1140 + "cr_volume": "74",
1141 + "cr_issue": "366a",
1142 + "cr_page": "291-298",
1143 + "cr_doi": "10.1080/01621459.1979.10482508",
1144 + "bib_volume": "74",
1145 + "bib_number": "366",
1146 + "bib_pages": "291--298",
1147 + "title_sim": 1.0,
1148 + "first_author_ok": true,
1149 + "year_ok": true,
1150 + "status": "OK"
1151 + },
1152 + {
1153 + "key": "jorda2005estimation",
1154 + "doi": "10.1257/0002828053828518",
1155 + "bib_title": "Estimation and Inference of Impulse Responses by Local Projections",
1156 + "bib_year": "2005",
1157 + "bib_journal": "American Economic Review",
1158 + "cr_title": "Estimation and Inference of Impulse Responses by Local Projections",
1159 + "cr_journal": "American Economic Review",
1160 + "cr_year_issued": 2005,
1161 + "cr_year_print": 2005,
1162 + "cr_authors": [
1163 + "Jord\u00e0"
1164 + ],
1165 + "cr_volume": "95",
1166 + "cr_issue": "1",
1167 + "cr_page": "161-182",
1168 + "cr_doi": "10.1257/0002828053828518",
1169 + "bib_volume": "95",
1170 + "bib_number": "1",
1171 + "bib_pages": "161--182",
1172 + "title_sim": 1.0,
1173 + "first_author_ok": false,
1174 + "year_ok": true,
1175 + "status": "CHECK"
1176 + },
1177 + {
1178 + "key": "kyle2001contagion",
1179 + "doi": "10.1111/0022-1082.00373",
1180 + "bib_title": "Contagion as a Wealth Effect",
1181 + "bib_year": "2001",
1182 + "bib_journal": "Journal of Finance",
1183 + "cr_title": "Contagion as a Wealth Effect",
1184 + "cr_journal": "The Journal of Finance",
1185 + "cr_year_issued": 2001,
1186 + "cr_year_print": 2001,
1187 + "cr_authors": [
1188 + "Kyle",
1189 + "Xiong"
1190 + ],
1191 + "cr_volume": "56",
1192 + "cr_issue": "4",
1193 + "cr_page": "1401-1440",
1194 + "cr_doi": "10.1111/0022-1082.00373",
1195 + "bib_volume": "56",
1196 + "bib_number": "4",
1197 + "bib_pages": "1401--1440",
1198 + "title_sim": 1.0,
1199 + "first_author_ok": true,
1200 + "year_ok": true,
1201 + "status": "OK"
1202 + },
1203 + {
1204 + "key": "kodres2002rational",
1205 + "doi": "10.1111/1540-6261.00441",
1206 + "bib_title": "A Rational Expectations Model of Financial Contagion",
1207 + "bib_year": "2002",
1208 + "bib_journal": "Journal of Finance",
1209 + "cr_title": "A Rational Expectations Model of Financial Contagion",
1210 + "cr_journal": "The Journal of Finance",
1211 + "cr_year_issued": 2002,
1212 + "cr_year_print": 2002,
1213 + "cr_authors": [
1214 + "Kodres",
1215 + "Pritsker"
1216 + ],
1217 + "cr_volume": "57",
1218 + "cr_issue": "2",
1219 + "cr_page": "769-799",
1220 + "cr_doi": "10.1111/1540-6261.00441",
1221 + "bib_volume": "57",
1222 + "bib_number": "2",
1223 + "bib_pages": "769--799",
1224 + "title_sim": 1.0,
1225 + "first_author_ok": true,
1226 + "year_ok": true,
1227 + "status": "OK"
1228 + },
1229 + {
1230 + "key": "yuan2005asymmetric",
1231 + "doi": "10.1111/j.1540-6261.2005.00733.x",
1232 + "bib_title": "Asymmetric Price Movements and Borrowing Constraints: A Rational Expectations Equilibrium Model of Crises, Contagion, and Confusion",
1233 + "bib_year": "2005",
1234 + "bib_journal": "Journal of Finance",
1235 + "cr_title": "Asymmetric Price Movements and Borrowing Constraints: A Rational Expectations Equilibrium Model of Crises, Contagion, and Confusion",
1236 + "cr_journal": "The Journal of Finance",
1237 + "cr_year_issued": 2005,
1238 + "cr_year_print": 2005,
1239 + "cr_authors": [
1240 + "YUAN"
1241 + ],
1242 + "cr_volume": "60",
1243 + "cr_issue": "1",
1244 + "cr_page": "379-411",
1245 + "cr_doi": "10.1111/j.1540-6261.2005.00733.x",
1246 + "bib_volume": "60",
1247 + "bib_number": "1",
1248 + "bib_pages": "379--411",
1249 + "title_sim": 1.0,
1250 + "first_author_ok": true,
1251 + "year_ok": true,
1252 + "status": "OK"
1253 + },
1254 + {
1255 + "key": "brunnermeier2009market",
1256 + "doi": "10.1093/rfs/hhn098",
1257 + "bib_title": "Market Liquidity and Funding Liquidity",
1258 + "bib_year": "2009",
1259 + "bib_journal": "Review of Financial Studies",
1260 + "cr_title": "Market Liquidity and Funding Liquidity",
1261 + "cr_journal": "Review of Financial Studies",
1262 + "cr_year_issued": 2008,
1263 + "cr_year_print": 2009,
1264 + "cr_authors": [
1265 + "Brunnermeier",
1266 + "Pedersen"
1267 + ],
1268 + "cr_volume": "22",
1269 + "cr_issue": "6",
1270 + "cr_page": "2201-2238",
1271 + "cr_doi": "10.1093/rfs/hhn098",
1272 + "bib_volume": "22",
1273 + "bib_number": "6",
1274 + "bib_pages": "2201--2238",
1275 + "title_sim": 1.0,
1276 + "first_author_ok": true,
1277 + "year_ok": true,
1278 + "status": "OK"
1279 + },
1280 + {
1281 + "key": "cespa2014illiquidity",
1282 + "doi": "10.1093/rfs/hhu016",
1283 + "bib_title": "Illiquidity Contagion and Liquidity Crashes",
1284 + "bib_year": "2014",
1285 + "bib_journal": "Review of Financial Studies",
1286 + "cr_title": "Illiquidity Contagion and Liquidity Crashes",
1287 + "cr_journal": "Review of Financial Studies",
1288 + "cr_year_issued": 2014,
1289 + "cr_year_print": 2014,
1290 + "cr_authors": [
1291 + "Cespa",
1292 + "Foucault"
1293 + ],
1294 + "cr_volume": "27",
1295 + "cr_issue": "6",
1296 + "cr_page": "1615-1660",
1297 + "cr_doi": "10.1093/rfs/hhu016",
1298 + "bib_volume": "27",
1299 + "bib_number": "6",
1300 + "bib_pages": "1615--1660",
1301 + "title_sim": 1.0,
1302 + "first_author_ok": true,
1303 + "year_ok": true,
1304 + "status": "OK"
1305 + },
1306 + {
1307 + "key": "coval2007asset",
1308 + "doi": "10.1016/j.jfineco.2006.09.007",
1309 + "bib_title": "Asset Fire Sales (and Purchases) in Equity Markets",
1310 + "bib_year": "2007",
1311 + "bib_journal": "Journal of Financial Economics",
1312 + "cr_title": "Asset fire sales (and purchases) in equity markets",
1313 + "cr_journal": "Journal of Financial Economics",
1314 + "cr_year_issued": 2007,
1315 + "cr_year_print": 2007,
1316 + "cr_authors": [
1317 + "Coval",
1318 + "Stafford"
1319 + ],
1320 + "cr_volume": "86",
1321 + "cr_issue": "2",
1322 + "cr_page": "479-512",
1323 + "cr_doi": "10.1016/j.jfineco.2006.09.007",
1324 + "bib_volume": "86",
1325 + "bib_number": "2",
1326 + "bib_pages": "479--512",
1327 + "title_sim": 1.0,
1328 + "first_author_ok": true,
1329 + "year_ok": true,
1330 + "status": "OK"
1331 + },
1332 + {
1333 + "key": "greenwood2011stock",
1334 + "doi": "10.1016/j.jfineco.2011.06.003",
1335 + "bib_title": "Stock Price Fragility",
1336 + "bib_year": "2011",
1337 + "bib_journal": "Journal of Financial Economics",
1338 + "cr_title": "Stock price fragility",
1339 + "cr_journal": "Journal of Financial Economics",
1340 + "cr_year_issued": 2011,
1341 + "cr_year_print": 2011,
1342 + "cr_authors": [
1343 + "Greenwood",
1344 + "Thesmar"
1345 + ],
1346 + "cr_volume": "102",
1347 + "cr_issue": "3",
1348 + "cr_page": "471-490",
1349 + "cr_doi": "10.1016/j.jfineco.2011.06.003",
1350 + "bib_volume": "102",
1351 + "bib_number": "3",
1352 + "bib_pages": "471--490",
1353 + "title_sim": 1.0,
1354 + "first_author_ok": true,
1355 + "year_ok": true,
1356 + "status": "OK"
1357 + },
1358 + {
1359 + "key": "anton2014connected",
1360 + "doi": "10.1111/jofi.12149",
1361 + "bib_title": "Connected Stocks",
1362 + "bib_year": "2014",
1363 + "bib_journal": "Journal of Finance",
1364 + "cr_title": "Connected Stocks",
1365 + "cr_journal": "The Journal of Finance",
1366 + "cr_year_issued": 2014,
1367 + "cr_year_print": 2014,
1368 + "cr_authors": [
1369 + "ANT\u00d3N",
1370 + "POLK"
1371 + ],
1372 + "cr_volume": "69",
1373 + "cr_issue": "3",
1374 + "cr_page": "1099-1127",
1375 + "cr_doi": "10.1111/jofi.12149",
1376 + "bib_volume": "69",
1377 + "bib_number": "3",
1378 + "bib_pages": "1099--1127",
1379 + "title_sim": 1.0,
1380 + "first_author_ok": false,
1381 + "year_ok": true,
1382 + "status": "CHECK"
1383 + },
1384 + {
1385 + "key": "lou2012flow",
1386 + "doi": "10.1093/rfs/hhs103",
1387 + "bib_title": "A Flow-Based Explanation for Return Predictability",
1388 + "bib_year": "2012",
1389 + "bib_journal": "Review of Financial Studies",
1390 + "cr_title": "A Flow-Based Explanation for Return Predictability",
1391 + "cr_journal": "Review of Financial Studies",
1392 + "cr_year_issued": 2012,
1393 + "cr_year_print": 2012,
1394 + "cr_authors": [
1395 + "Lou"
1396 + ],
1397 + "cr_volume": "25",
1398 + "cr_issue": "12",
1399 + "cr_page": "3457-3489",
1400 + "cr_doi": "10.1093/rfs/hhs103",
1401 + "bib_volume": "25",
1402 + "bib_number": "12",
1403 + "bib_pages": "3457--3489",
1404 + "title_sim": 1.0,
1405 + "first_author_ok": true,
1406 + "year_ok": true,
1407 + "status": "OK"
1408 + },
1409 + {
1410 + "key": "greenwood2015vulnerable",
1411 + "doi": "10.1016/j.jfineco.2014.11.006",
1412 + "bib_title": "Vulnerable Banks",
1413 + "bib_year": "2015",
1414 + "bib_journal": "Journal of Financial Economics",
1415 + "cr_title": "Vulnerable banks",
1416 + "cr_journal": "Journal of Financial Economics",
1417 + "cr_year_issued": 2015,
1418 + "cr_year_print": 2015,
1419 + "cr_authors": [
1420 + "Greenwood",
1421 + "Landier",
1422 + "Thesmar"
1423 + ],
1424 + "cr_volume": "115",
1425 + "cr_issue": "3",
1426 + "cr_page": "471-485",
1427 + "cr_doi": "10.1016/j.jfineco.2014.11.006",
1428 + "bib_volume": "115",
1429 + "bib_number": "3",
1430 + "bib_pages": "471--485",
1431 + "title_sim": 1.0,
1432 + "first_author_ok": true,
1433 + "year_ok": true,
1434 + "status": "OK"
1435 + },
1436 + {
1437 + "key": "duarte2021fire",
1438 + "doi": "10.1111/jofi.13010",
1439 + "bib_title": "Fire-Sale Spillovers and Systemic Risk",
1440 + "bib_year": "2021",
1441 + "bib_journal": "Journal of Finance",
1442 + "cr_title": "Fire\u2010Sale Spillovers and Systemic Risk",
1443 + "cr_journal": "The Journal of Finance",
1444 + "cr_year_issued": 2021,
1445 + "cr_year_print": 2021,
1446 + "cr_authors": [
1447 + "DUARTE",
1448 + "EISENBACH"
1449 + ],
1450 + "cr_volume": "76",
1451 + "cr_issue": "3",
1452 + "cr_page": "1251-1294",
1453 + "cr_doi": "10.1111/jofi.13010",
1454 + "bib_volume": "76",
1455 + "bib_number": "3",
1456 + "bib_pages": "1251--1294",
1457 + "title_sim": 0.974,
1458 + "first_author_ok": true,
1459 + "year_ok": true,
1460 + "status": "OK"
1461 + },
1462 + {
1463 + "key": "glasserman2016contagion",
1464 + "doi": "10.1257/jel.20151228",
1465 + "bib_title": "Contagion in Financial Networks",
1466 + "bib_year": "2016",
1467 + "bib_journal": "Journal of Economic Literature",
1468 + "cr_title": "Contagion in Financial Networks",
1469 + "cr_journal": "Journal of Economic Literature",
1470 + "cr_year_issued": 2016,
1471 + "cr_year_print": 2016,
1472 + "cr_authors": [
1473 + "Glasserman",
1474 + "Young"
1475 + ],
1476 + "cr_volume": "54",
1477 + "cr_issue": "3",
1478 + "cr_page": "779-831",
1479 + "cr_doi": "10.1257/jel.20151228",
1480 + "bib_volume": "54",
1481 + "bib_number": "3",
1482 + "bib_pages": "779--831",
1483 + "title_sim": 1.0,
1484 + "first_author_ok": true,
1485 + "year_ok": true,
1486 + "status": "OK"
1487 + },
1488 + {
1489 + "key": "bollerslev2008risk",
1490 + "doi": "10.1016/j.jeconom.2008.01.006",
1491 + "bib_title": "Risk, Jumps, and Diversification",
1492 + "bib_year": "2008",
1493 + "bib_journal": "Journal of Econometrics",
1494 + "cr_title": "Risk, jumps, and diversification",
1495 + "cr_journal": "Journal of Econometrics",
1496 + "cr_year_issued": 2008,
1497 + "cr_year_print": 2008,
1498 + "cr_authors": [
1499 + "Bollerslev",
1500 + "Law",
1501 + "Tauchen"
1502 + ],
1503 + "cr_volume": "144",
1504 + "cr_issue": "1",
1505 + "cr_page": "234-256",
1506 + "cr_doi": "10.1016/j.jeconom.2008.01.006",
1507 + "bib_volume": "144",
1508 + "bib_number": "1",
1509 + "bib_pages": "234--256",
1510 + "title_sim": 1.0,
1511 + "first_author_ok": true,
1512 + "year_ok": true,
1513 + "status": "OK"
1514 + },
1515 + {
1516 + "key": "das2004systemic",
1517 + "doi": "10.1111/j.1540-6261.2004.00717.x",
1518 + "bib_title": "Systemic Risk and International Portfolio Choice",
1519 + "bib_year": "2004",
1520 + "bib_journal": "Journal of Finance",
1521 + "cr_title": "Systemic Risk and International Portfolio Choice",
1522 + "cr_journal": "The Journal of Finance",
1523 + "cr_year_issued": 2004,
1524 + "cr_year_print": 2004,
1525 + "cr_authors": [
1526 + "DAS",
1527 + "UPPAL"
1528 + ],
1529 + "cr_volume": "59",
1530 + "cr_issue": "6",
1531 + "cr_page": "2809-2834",
1532 + "cr_doi": "10.1111/j.1540-6261.2004.00717.x",
1533 + "bib_volume": "59",
1534 + "bib_number": "6",
1535 + "bib_pages": "2809--2834",
1536 + "title_sim": 1.0,
1537 + "first_author_ok": true,
1538 + "year_ok": true,
1539 + "status": "OK"
1540 + },
1541 + {
1542 + "key": "lahaye2011jumps",
1543 + "doi": "10.1002/jae.1149",
1544 + "bib_title": "Jumps, Cojumps and Macro Announcements",
1545 + "bib_year": "2011",
1546 + "bib_journal": "Journal of Applied Econometrics",
1547 + "cr_title": "Jumps, cojumps and macro announcements",
1548 + "cr_journal": "Journal of Applied Econometrics",
1549 + "cr_year_issued": 2010,
1550 + "cr_year_print": 2011,
1551 + "cr_authors": [
1552 + "Lahaye",
1553 + "Laurent",
1554 + "Neely"
1555 + ],
1556 + "cr_volume": "26",
1557 + "cr_issue": "6",
1558 + "cr_page": "893-921",
1559 + "cr_doi": "10.1002/jae.1149",
1560 + "bib_volume": "26",
1561 + "bib_number": "6",
1562 + "bib_pages": "893--921",
1563 + "title_sim": 1.0,
1564 + "first_author_ok": true,
1565 + "year_ok": true,
1566 + "status": "OK"
1567 + },
1568 + {
1569 + "key": "caporin2017systemic",
1570 + "doi": "10.1016/j.jfineco.2017.06.016",
1571 + "bib_title": "Systemic Co-jumps",
1572 + "bib_year": "2017",
1573 + "bib_journal": "Journal of Financial Economics",
1574 + "cr_title": "Systemic co-jumps",
1575 + "cr_journal": "Journal of Financial Economics",
1576 + "cr_year_issued": 2017,
1577 + "cr_year_print": 2017,
1578 + "cr_authors": [
1579 + "Caporin",
1580 + "Kolokolov",
1581 + "Ren\u00f2"
1582 + ],
1583 + "cr_volume": "126",
1584 + "cr_issue": "3",
1585 + "cr_page": "563-591",
1586 + "cr_doi": "10.1016/j.jfineco.2017.06.016",
1587 + "bib_volume": "126",
1588 + "bib_number": "3",
1589 + "bib_pages": "563--591",
1590 + "title_sim": 1.0,
1591 + "first_author_ok": true,
1592 + "year_ok": true,
1593 + "status": "OK"
1594 + },
1595 + {
1596 + "key": "bajgrowicz2016jumps",
1597 + "doi": "10.1287/mnsc.2015.2234",
1598 + "bib_title": "Jumps in High-Frequency Data: Spurious Detections, Dynamics, and News",
1599 + "bib_year": "2016",
1600 + "bib_journal": "Management Science",
1601 + "cr_title": "Jumps in High-Frequency Data: Spurious Detections, Dynamics, and News",
1602 + "cr_journal": "Management Science",
1603 + "cr_year_issued": 2016,
1604 + "cr_year_print": 2016,
1605 + "cr_authors": [
1606 + "Bajgrowicz",
1607 + "Scaillet",
1608 + "Treccani"
1609 + ],
1610 + "cr_volume": "62",
1611 + "cr_issue": "8",
1612 + "cr_page": "2198-2217",
1613 + "cr_doi": "10.1287/mnsc.2015.2234",
1614 + "bib_volume": "62",
1615 + "bib_number": "8",
1616 + "bib_pages": "2198--2217",
1617 + "title_sim": 1.0,
1618 + "first_author_ok": true,
1619 + "year_ok": true,
1620 + "status": "OK"
1621 + },
1622 + {
1623 + "key": "christensen2014fact",
1624 + "doi": "10.1016/j.jfineco.2014.07.007",
1625 + "bib_title": "Fact or Friction: Jumps at Ultra High Frequency",
1626 + "bib_year": "2014",
1627 + "bib_journal": "Journal of Financial Economics",
1628 + "cr_title": "Fact or friction: Jumps at ultra high frequency",
1629 + "cr_journal": "Journal of Financial Economics",
1630 + "cr_year_issued": 2014,
1631 + "cr_year_print": 2014,
1632 + "cr_authors": [
1633 + "Christensen",
1634 + "Oomen",
1635 + "Podolskij"
1636 + ],
1637 + "cr_volume": "114",
1638 + "cr_issue": "3",
1639 + "cr_page": "576-599",
1640 + "cr_doi": "10.1016/j.jfineco.2014.07.007",
1641 + "bib_volume": "114",
1642 + "bib_number": "3",
1643 + "bib_pages": "576--599",
1644 + "title_sim": 1.0,
1645 + "first_author_ok": true,
1646 + "year_ok": true,
1647 + "status": "OK"
1648 + },
1649 + {
1650 + "key": "brogaard2018high",
1651 + "doi": "10.1016/j.jfineco.2018.02.002",
1652 + "bib_title": "High Frequency Trading and Extreme Price Movements",
1653 + "bib_year": "2018",
1654 + "bib_journal": "Journal of Financial Economics",
1655 + "cr_title": "High frequency trading and extreme price movements",
1656 + "cr_journal": "Journal of Financial Economics",
1657 + "cr_year_issued": 2018,
1658 + "cr_year_print": 2018,
1659 + "cr_authors": [
1660 + "Brogaard",
1661 + "Carrion",
1662 + "Moyaert",
1663 + "Riordan",
1664 + "Shkilko",
1665 + "Sokolov"
1666 + ],
1667 + "cr_volume": "128",
1668 + "cr_issue": "2",
1669 + "cr_page": "253-265",
1670 + "cr_doi": "10.1016/j.jfineco.2018.02.002",
1671 + "bib_volume": "128",
1672 + "bib_number": "2",
1673 + "bib_pages": "253--265",
1674 + "title_sim": 1.0,
1675 + "first_author_ok": true,
1676 + "year_ok": true,
1677 + "status": "OK"
1678 + },
1679 + {
1680 + "key": "menkveld2019flash",
1681 + "doi": "10.1287/mnsc.2018.3040",
1682 + "bib_title": "The Flash Crash: A Cautionary Tale About Highly Fragmented Markets",
1683 + "bib_year": "2019",
1684 + "bib_journal": "Management Science",
1685 + "cr_title": "The Flash Crash: A Cautionary Tale About Highly Fragmented Markets",
1686 + "cr_journal": "Management Science",
1687 + "cr_year_issued": 2019,
1688 + "cr_year_print": 2019,
1689 + "cr_authors": [
1690 + "Menkveld",
1691 + "Yueshen"
1692 + ],
1693 + "cr_volume": "65",
1694 + "cr_issue": "10",
1695 + "cr_page": "4470-4488",
1696 + "cr_doi": "10.1287/mnsc.2018.3040",
1697 + "bib_volume": "65",
1698 + "bib_number": "10",
1699 + "bib_pages": "4470--4488",
1700 + "title_sim": 1.0,
1701 + "first_author_ok": true,
1702 + "year_ok": true,
1703 + "status": "OK"
1704 + },
1705 + {
1706 + "key": "andersen2003micro",
1707 + "doi": "10.1257/000282803321455151",
1708 + "bib_title": "Micro Effects of Macro Announcements: Real-Time Price Discovery in Foreign Exchange",
1709 + "bib_year": "2003",
1710 + "bib_journal": "American Economic Review",
1711 + "cr_title": "Micro Effects of Macro Announcements: Real-Time Price Discovery in Foreign Exchange",
1712 + "cr_journal": "American Economic Review",
1713 + "cr_year_issued": 2003,
1714 + "cr_year_print": 2003,
1715 + "cr_authors": [
1716 + "Anderson",
1717 + "Bollerslev",
1718 + "Diebold",
1719 + "Vega"
1720 + ],
1721 + "cr_volume": "93",
1722 + "cr_issue": "1",
1723 + "cr_page": "38-62",
1724 + "cr_doi": "10.1257/000282803321455151",
1725 + "bib_volume": "93",
1726 + "bib_number": "1",
1727 + "bib_pages": "38--62",
1728 + "title_sim": 1.0,
1729 + "first_author_ok": false,
1730 + "year_ok": true,
1731 + "status": "CHECK"
1732 + },
1733 + {
1734 + "key": "andersen2007real",
1735 + "doi": "10.1016/j.jinteco.2007.02.004",
1736 + "bib_title": "Real-Time Price Discovery in Global Stock, Bond and Foreign Exchange Markets",
1737 + "bib_year": "2007",
1738 + "bib_journal": "Journal of International Economics",
1739 + "cr_title": "Real-time price discovery in global stock, bond and foreign exchange markets",
1740 + "cr_journal": "Journal of International Economics",
1741 + "cr_year_issued": 2007,
1742 + "cr_year_print": 2007,
1743 + "cr_authors": [
1744 + "Andersen",
1745 + "Bollerslev",
1746 + "Diebold",
1747 + "Vega"
1748 + ],
1749 + "cr_volume": "73",
1750 + "cr_issue": "2",
1751 + "cr_page": "251-277",
1752 + "cr_doi": "10.1016/j.jinteco.2007.02.004",
1753 + "bib_volume": "73",
1754 + "bib_number": "2",
1755 + "bib_pages": "251--277",
1756 + "title_sim": 1.0,
1757 + "first_author_ok": true,
1758 + "year_ok": true,
1759 + "status": "OK"
1760 + },
1761 + {
1762 + "key": "lucca2015pre",
1763 + "doi": "10.1111/jofi.12196",
1764 + "bib_title": "The Pre-{FOMC} Announcement Drift",
1765 + "bib_year": "2015",
1766 + "bib_journal": "Journal of Finance",
1767 + "cr_title": "The Pre\u2010FOMC Announcement Drift",
1768 + "cr_journal": "The Journal of Finance",
1769 + "cr_year_issued": 2015,
1770 + "cr_year_print": 2015,
1771 + "cr_authors": [
1772 + "LUCCA",
1773 + "MOENCH"
1774 + ],
1775 + "cr_volume": "70",
1776 + "cr_issue": "1",
1777 + "cr_page": "329-371",
1778 + "cr_doi": "10.1111/jofi.12196",
1779 + "bib_volume": "70",
1780 + "bib_number": "1",
1781 + "bib_pages": "329--371",
1782 + "title_sim": 0.968,
1783 + "first_author_ok": true,
1784 + "year_ok": true,
1785 + "status": "OK"
1786 + },
1787 + {
1788 + "key": "savor2013how",
1789 + "doi": "10.1017/S002210901300015X",
1790 + "bib_title": "How Much Do Investors Care About Macroeconomic Risk? {E}vidence from Scheduled Economic Announcements",
1791 + "bib_year": "2013",
1792 + "bib_journal": "Journal of Financial and Quantitative Analysis",
1793 + "cr_title": "How Much Do Investors Care About Macroeconomic Risk? Evidence from Scheduled Economic Announcements",
1794 + "cr_journal": "Journal of Financial and Quantitative Analysis",
1795 + "cr_year_issued": 2013,
1796 + "cr_year_print": 2013,
1797 + "cr_authors": [
1798 + "Savor",
1799 + "Wilson"
1800 + ],
1801 + "cr_volume": "48",
1802 + "cr_issue": "2",
1803 + "cr_page": "343-375",
1804 + "cr_doi": "10.1017/s002210901300015x",
1805 + "bib_volume": "48",
1806 + "bib_number": "2",
1807 + "bib_pages": "343--375",
1808 + "title_sim": 1.0,
1809 + "first_author_ok": true,
1810 + "year_ok": true,
1811 + "status": "OK"
1812 + },
1813 + {
1814 + "key": "engle1990meteor",
1815 + "doi": "10.2307/2938189",
1816 + "bib_title": "Meteor Showers or Heat Waves? {H}eteroskedastic Intra-Daily Volatility in the Foreign Exchange Market",
1817 + "bib_year": "1990",
1818 + "bib_journal": "Econometrica",
1819 + "cr_title": "Meteor Showers or Heat Waves? Heteroskedastic Intra-Daily Volatility in the Foreign Exchange Market",
1820 + "cr_journal": "Econometrica",
1821 + "cr_year_issued": 1990,
1822 + "cr_year_print": 1990,
1823 + "cr_authors": [
1824 + "Engle",
1825 + "Ito",
1826 + "Lin"
1827 + ],
1828 + "cr_volume": "58",
1829 + "cr_issue": "3",
1830 + "cr_page": "525",
1831 + "cr_doi": "10.2307/2938189",
1832 + "bib_volume": "58",
1833 + "bib_number": "3",
1834 + "bib_pages": "525--542",
1835 + "title_sim": 1.0,
1836 + "first_author_ok": true,
1837 + "year_ok": true,
1838 + "status": "OK"
1839 + },
1840 + {
1841 + "key": "king1990transmission",
1842 + "doi": "10.1093/rfs/3.1.5",
1843 + "bib_title": "Transmission of Volatility between Stock Markets",
1844 + "bib_year": "1990",
1845 + "bib_journal": "Review of Financial Studies",
1846 + "cr_title": "Transmission of Volatility between Stock Markets",
1847 + "cr_journal": "Review of Financial Studies",
1848 + "cr_year_issued": 1990,
1849 + "cr_year_print": 1990,
1850 + "cr_authors": [
1851 + "King",
1852 + "Wadhwani"
1853 + ],
1854 + "cr_volume": "3",
1855 + "cr_issue": "1",
1856 + "cr_page": "5-33",
1857 + "cr_doi": "10.1093/rfs/3.1.5",
1858 + "bib_volume": "3",
1859 + "bib_number": "1",
1860 + "bib_pages": "5--33",
1861 + "title_sim": 1.0,
1862 + "first_author_ok": true,
1863 + "year_ok": true,
1864 + "status": "OK"
1865 + },
1866 + {
1867 + "key": "hamao1990correlations",
1868 + "doi": "10.1093/rfs/3.2.281",
1869 + "bib_title": "Correlations in Price Changes and Volatility across International Stock Markets",
1870 + "bib_year": "1990",
1871 + "bib_journal": "Review of Financial Studies",
1872 + "cr_title": "Correlations in Price Changes and Volatility across International Stock Markets",
1873 + "cr_journal": "Review of Financial Studies",
1874 + "cr_year_issued": 1990,
1875 + "cr_year_print": 1990,
1876 + "cr_authors": [
1877 + "Hamao",
1878 + "Masulis",
1879 + "Ng"
1880 + ],
1881 + "cr_volume": "3",
1882 + "cr_issue": "2",
1883 + "cr_page": "281-307",
1884 + "cr_doi": "10.1093/rfs/3.2.281",
1885 + "bib_volume": "3",
1886 + "bib_number": "2",
1887 + "bib_pages": "281--307",
1888 + "title_sim": 1.0,
1889 + "first_author_ok": true,
1890 + "year_ok": true,
1891 + "status": "OK"
1892 + },
1893 + {
1894 + "key": "cohen2008economic",
1895 + "doi": "10.1111/j.1540-6261.2008.01379.x",
1896 + "bib_title": "Economic Links and Predictable Returns",
1897 + "bib_year": "2008",
1898 + "bib_journal": "Journal of Finance",
1899 + "cr_title": "Economic Links and Predictable Returns",
1900 + "cr_journal": "The Journal of Finance",
1901 + "cr_year_issued": 2008,
1902 + "cr_year_print": 2008,
1903 + "cr_authors": [
1904 + "COHEN",
1905 + "FRAZZINI"
1906 + ],
1907 + "cr_volume": "63",
1908 + "cr_issue": "4",
1909 + "cr_page": "1977-2011",
1910 + "cr_doi": "10.1111/j.1540-6261.2008.01379.x",
1911 + "bib_volume": "63",
1912 + "bib_number": "4",
1913 + "bib_pages": "1977--2011",
1914 + "title_sim": 1.0,
1915 + "first_author_ok": true,
1916 + "year_ok": true,
1917 + "status": "OK"
1918 + },
1919 + {
1920 + "key": "menzly2010market",
1921 + "doi": "10.1111/j.1540-6261.2010.01578.x",
1922 + "bib_title": "Market Segmentation and Cross-Predictability of Returns",
1923 + "bib_year": "2010",
1924 + "bib_journal": "Journal of Finance",
1925 + "cr_title": "Market Segmentation and Cross\u2010predictability of Returns",
1926 + "cr_journal": "The Journal of Finance",
1927 + "cr_year_issued": 2010,
1928 + "cr_year_print": 2010,
1929 + "cr_authors": [
1930 + "MENZLY",
1931 + "OZBAS"
1932 + ],
1933 + "cr_volume": "65",
1934 + "cr_issue": "4",
1935 + "cr_page": "1555-1580",
1936 + "cr_doi": "10.1111/j.1540-6261.2010.01578.x",
1937 + "bib_volume": "65",
1938 + "bib_number": "4",
1939 + "bib_pages": "1555--1580",
1940 + "title_sim": 0.982,
1941 + "first_author_ok": true,
1942 + "year_ok": true,
1943 + "status": "OK"
1944 + },
1945 + {
1946 + "key": "barrot2016input",
1947 + "doi": "10.1093/qje/qjw018",
1948 + "bib_title": "Input Specificity and the Propagation of Idiosyncratic Shocks in Production Networks",
1949 + "bib_year": "2016",
1950 + "bib_journal": "Quarterly Journal of Economics",
1951 + "cr_title": "Input Specificity and the Propagation of Idiosyncratic Shocks in Production Networks\n*",
1952 + "cr_journal": "The Quarterly Journal of Economics",
1953 + "cr_year_issued": 2016,
1954 + "cr_year_print": 2016,
1955 + "cr_authors": [
1956 + "Barrot",
1957 + "Sauvagnat"
1958 + ],
1959 + "cr_volume": "131",
1960 + "cr_issue": "3",
1961 + "cr_page": "1543-1592",
1962 + "cr_doi": "10.1093/qje/qjw018",
1963 + "bib_volume": "131",
1964 + "bib_number": "3",
1965 + "bib_pages": "1543--1592",
1966 + "title_sim": 0.988,
1967 + "first_author_ok": true,
1968 + "year_ok": true,
1969 + "status": "OK"
1970 + },
1971 + {
1972 + "key": "herskovic2018networks",
1973 + "doi": "10.1111/jofi.12684",
1974 + "bib_title": "Networks in Production: Asset Pricing Implications",
1975 + "bib_year": "2018",
1976 + "bib_journal": "Journal of Finance",
1977 + "cr_title": "Networks in Production: Asset Pricing Implications",
1978 + "cr_journal": "The Journal of Finance",
1979 + "cr_year_issued": 2018,
1980 + "cr_year_print": 2018,
1981 + "cr_authors": [
1982 + "HERSKOVIC"
1983 + ],
1984 + "cr_volume": "73",
1985 + "cr_issue": "4",
1986 + "cr_page": "1785-1818",
1987 + "cr_doi": "10.1111/jofi.12684",
1988 + "bib_volume": "73",
1989 + "bib_number": "4",
1990 + "bib_pages": "1785--1818",
1991 + "title_sim": 1.0,
1992 + "first_author_ok": true,
1993 + "year_ok": true,
1994 + "status": "OK"
1995 + },
1996 + {
1997 + "key": "chinco2019sparse",
1998 + "doi": "10.1111/jofi.12733",
1999 + "bib_title": "Sparse Signals in the Cross-Section of Returns",
2000 + "bib_year": "2019",
2001 + "bib_journal": "Journal of Finance",
2002 + "cr_title": "Sparse Signals in the Cross\u2010Section of Returns",
2003 + "cr_journal": "The Journal of Finance",
2004 + "cr_year_issued": 2018,
2005 + "cr_year_print": 2019,
2006 + "cr_authors": [
2007 + "CHINCO",
2008 + "CLARK\u2010JOSEPH",
2009 + "YE"
2010 + ],
2011 + "cr_volume": "74",
2012 + "cr_issue": "1",
2013 + "cr_page": "449-492",
2014 + "cr_doi": "10.1111/jofi.12733",
2015 + "bib_volume": "74",
2016 + "bib_number": "1",
2017 + "bib_pages": "449--492",
2018 + "title_sim": 0.978,
2019 + "first_author_ok": true,
2020 + "year_ok": true,
2021 + "status": "OK"
2022 + },
2023 + {
2024 + "key": "huth2014high",
2025 + "doi": "10.1016/j.jempfin.2014.01.003",
2026 + "bib_title": "High Frequency Lead/lag Relationships --- Empirical Facts",
2027 + "bib_year": "2014",
2028 + "bib_journal": "Journal of Empirical Finance",
2029 + "cr_title": "High frequency lead/lag relationships \u2014 Empirical facts",
2030 + "cr_journal": "Journal of Empirical Finance",
2031 + "cr_year_issued": 2014,
2032 + "cr_year_print": 2014,
2033 + "cr_authors": [
2034 + "Huth",
2035 + "Abergel"
2036 + ],
2037 + "cr_volume": "26",
2038 + "cr_issue": null,
2039 + "cr_page": "41-58",
2040 + "cr_doi": "10.1016/j.jempfin.2014.01.003",
2041 + "bib_volume": "26",
2042 + "bib_number": "",
2043 + "bib_pages": "41--58",
2044 + "title_sim": 0.964,
2045 + "first_author_ok": true,
2046 + "year_ok": true,
2047 + "status": "OK"
2048 + },
2049 + {
2050 + "key": "hameed2010stock",
2051 + "doi": "10.1111/j.1540-6261.2009.01529.x",
2052 + "bib_title": "Stock Market Declines and Liquidity",
2053 + "bib_year": "2010",
2054 + "bib_journal": "Journal of Finance",
2055 + "cr_title": "Stock Market Declines and Liquidity",
2056 + "cr_journal": "The Journal of Finance",
2057 + "cr_year_issued": 2010,
2058 + "cr_year_print": 2010,
2059 + "cr_authors": [
2060 + "HAMEED",
2061 + "KANG",
2062 + "VISWANATHAN"
2063 + ],
2064 + "cr_volume": "65",
2065 + "cr_issue": "1",
2066 + "cr_page": "257-293",
2067 + "cr_doi": "10.1111/j.1540-6261.2009.01529.x",
2068 + "bib_volume": "65",
2069 + "bib_number": "1",
2070 + "bib_pages": "257--293",
2071 + "title_sim": 1.0,
2072 + "first_author_ok": true,
2073 + "year_ok": true,
2074 + "status": "OK"
2075 + },
2076 + {
2077 + "key": "nagel2012evaporating",
2078 + "doi": "10.1093/rfs/hhs066",
2079 + "bib_title": "Evaporating Liquidity",
2080 + "bib_year": "2012",
2081 + "bib_journal": "Review of Financial Studies",
2082 + "cr_title": "Evaporating Liquidity",
2083 + "cr_journal": "Review of Financial Studies",
2084 + "cr_year_issued": 2012,
2085 + "cr_year_print": 2012,
2086 + "cr_authors": [
2087 + "Nagel"
2088 + ],
2089 + "cr_volume": "25",
2090 + "cr_issue": "7",
2091 + "cr_page": "2005-2039",
2092 + "cr_doi": "10.1093/rfs/hhs066",
2093 + "bib_volume": "25",
2094 + "bib_number": "7",
2095 + "bib_pages": "2005--2039",
2096 + "title_sim": 1.0,
2097 + "first_author_ok": true,
2098 + "year_ok": true,
2099 + "status": "OK"
2100 + },
2101 + {
2102 + "key": "comertonforde2010time",
2103 + "doi": "10.1111/j.1540-6261.2009.01530.x",
2104 + "bib_title": "Time Variation in Liquidity: The Role of Market-Maker Inventories and Revenues",
2105 + "bib_year": "2010",
2106 + "bib_journal": "Journal of Finance",
2107 + "cr_title": "Time Variation in Liquidity: The Role of Market\u2010Maker Inventories and Revenues",
2108 + "cr_journal": "The Journal of Finance",
2109 + "cr_year_issued": 2010,
2110 + "cr_year_print": 2010,
2111 + "cr_authors": [
2112 + "COMERTON\u2010FORDE",
2113 + "HENDERSHOTT",
2114 + "JONES",
2115 + "MOULTON",
2116 + "SEASHOLES"
2117 + ],
2118 + "cr_volume": "65",
2119 + "cr_issue": "1",
2120 + "cr_page": "295-331",
2121 + "cr_doi": "10.1111/j.1540-6261.2009.01530.x",
2122 + "bib_volume": "65",
2123 + "bib_number": "1",
2124 + "bib_pages": "295--331",
2125 + "title_sim": 0.987,
2126 + "first_author_ok": false,
2127 + "year_ok": true,
2128 + "status": "CHECK"
2129 + },
2130 + {
2131 + "key": "coughenour2004common",
2132 + "doi": "10.1016/j.jfineco.2003.05.006",
2133 + "bib_title": "Common Market Makers and Commonality in Liquidity",
2134 + "bib_year": "2004",
2135 + "bib_journal": "Journal of Financial Economics",
2136 + "cr_title": "Common market makers and commonality in liquidity",
2137 + "cr_journal": "Journal of Financial Economics",
2138 + "cr_year_issued": 2004,
2139 + "cr_year_print": 2004,
2140 + "cr_authors": [
2141 + "Coughenour",
2142 + "Saad"
2143 + ],
2144 + "cr_volume": "73",
2145 + "cr_issue": "1",
2146 + "cr_page": "37-69",
2147 + "cr_doi": "10.1016/j.jfineco.2003.05.006",
2148 + "bib_volume": "73",
2149 + "bib_number": "1",
2150 + "bib_pages": "37--69",
2151 + "title_sim": 1.0,
2152 + "first_author_ok": true,
2153 + "year_ok": true,
2154 + "status": "OK"
2155 + },
2156 + {
2157 + "key": "andrade2008trading",
2158 + "doi": "10.1016/j.jfineco.2007.04.005",
2159 + "bib_title": "Trading Imbalances, Predictable Reversals, and Cross-Stock Price Pressure",
2160 + "bib_year": "2008",
2161 + "bib_journal": "Journal of Financial Economics",
2162 + "cr_title": "Trading imbalances, predictable reversals, and cross-stock price pressure",
2163 + "cr_journal": "Journal of Financial Economics",
2164 + "cr_year_issued": 2008,
2165 + "cr_year_print": 2008,
2166 + "cr_authors": [
2167 + "Andrade",
2168 + "Chang",
2169 + "Seasholes"
2170 + ],
2171 + "cr_volume": "88",
2172 + "cr_issue": "2",
2173 + "cr_page": "406-423",
2174 + "cr_doi": "10.1016/j.jfineco.2007.04.005",
2175 + "bib_volume": "88",
2176 + "bib_number": "2",
2177 + "bib_pages": "406--423",
2178 + "title_sim": 1.0,
2179 + "first_author_ok": true,
2180 + "year_ok": true,
2181 + "status": "OK"
2182 + },
2183 + {
2184 + "key": "hasbrouck1991measuring",
2185 + "doi": "10.1111/j.1540-6261.1991.tb03749.x",
2186 + "bib_title": "Measuring the Information Content of Stock Trades",
2187 + "bib_year": "1991",
2188 + "bib_journal": "Journal of Finance",
2189 + "cr_title": "Measuring the Information Content of Stock Trades",
2190 + "cr_journal": "The Journal of Finance",
2191 + "cr_year_issued": 1991,
2192 + "cr_year_print": 1991,
2193 + "cr_authors": [
2194 + "HASBROUCK"
2195 + ],
2196 + "cr_volume": "46",
2197 + "cr_issue": "1",
2198 + "cr_page": "179-207",
2199 + "cr_doi": "10.1111/j.1540-6261.1991.tb03749.x",
2200 + "bib_volume": "46",
2201 + "bib_number": "1",
2202 + "bib_pages": "179--207",
2203 + "title_sim": 1.0,
2204 + "first_author_ok": true,
2205 + "year_ok": true,
2206 + "status": "OK"
2207 + },
2208 + {
2209 + "key": "boudt2014intraday",
2210 + "doi": "10.1016/j.finmar.2013.05.004",
2211 + "bib_title": "Intraday Liquidity Dynamics and News Releases around Price Jumps: Evidence from the {DJIA} Stocks",
2212 + "bib_year": "2014",
2213 + "bib_journal": "Journal of Financial Markets",
2214 + "cr_title": "Intraday liquidity dynamics and news releases around price jumps: Evidence from the DJIA stocks",
2215 + "cr_journal": "Journal of Financial Markets",
2216 + "cr_year_issued": 2014,
2217 + "cr_year_print": 2014,
2218 + "cr_authors": [
2219 + "Boudt",
2220 + "Petitjean"
2221 + ],
2222 + "cr_volume": "17",
2223 + "cr_issue": null,
2224 + "cr_page": "121-149",
2225 + "cr_doi": "10.1016/j.finmar.2013.05.004",
2226 + "bib_volume": "17",
2227 + "bib_number": "",
2228 + "bib_pages": "121--149",
2229 + "title_sim": 1.0,
2230 + "first_author_ok": true,
2231 + "year_ok": true,
2232 + "status": "OK"
2233 + },
2234 + {
2235 + "key": "bollerslev2011tails",
2236 + "doi": "10.1111/j.1540-6261.2011.01695.x",
2237 + "bib_title": "Tails, Fears, and Risk Premia",
2238 + "bib_year": "2011",
2239 + "bib_journal": "Journal of Finance",
2240 + "cr_title": "Tails, Fears, and Risk Premia",
2241 + "cr_journal": "The Journal of Finance",
2242 + "cr_year_issued": 2011,
2243 + "cr_year_print": 2011,
2244 + "cr_authors": [
2245 + "BOLLERSLEV",
2246 + "TODOROV"
2247 + ],
2248 + "cr_volume": "66",
2249 + "cr_issue": "6",
2250 + "cr_page": "2165-2211",
2251 + "cr_doi": "10.1111/j.1540-6261.2011.01695.x",
2252 + "bib_volume": "66",
2253 + "bib_number": "6",
2254 + "bib_pages": "2165--2211",
2255 + "title_sim": 1.0,
2256 + "first_author_ok": true,
2257 + "year_ok": true,
2258 + "status": "OK"
2259 + },
2260 + {
2261 + "key": "bollerslev2016roughing",
2262 + "doi": "10.1016/j.jfineco.2016.02.001",
2263 + "bib_title": "Roughing up beta: Continuous versus discontinuous betas and the cross section of expected stock returns",
2264 + "bib_year": "2016",
2265 + "bib_journal": "Journal of Financial Economics",
2266 + "cr_title": "Roughing up beta: Continuous versus discontinuous betas and the cross section of expected stock returns",
2267 + "cr_journal": "Journal of Financial Economics",
2268 + "cr_year_issued": 2016,
2269 + "cr_year_print": 2016,
2270 + "cr_authors": [
2271 + "Bollerslev",
2272 + "Li",
2273 + "Todorov"
2274 + ],
2275 + "cr_volume": "120",
2276 + "cr_issue": "3",
2277 + "cr_page": "464-490",
2278 + "cr_doi": "10.1016/j.jfineco.2016.02.001",
2279 + "bib_volume": "120",
2280 + "bib_number": "3",
2281 + "bib_pages": "464--490",
2282 + "title_sim": 1.0,
2283 + "first_author_ok": true,
2284 + "year_ok": true,
2285 + "status": "OK"
2286 + },
2287 + {
2288 + "key": "pelger2020understanding",
2289 + "doi": "10.1111/jofi.12898",
2290 + "bib_title": "Understanding Systematic Risk: A High-Frequency Approach",
2291 + "bib_year": "2020",
2292 + "bib_journal": "Journal of Finance",
2293 + "cr_title": "Understanding Systematic Risk: A High\u2010Frequency Approach",
2294 + "cr_journal": "The Journal of Finance",
2295 + "cr_year_issued": 2020,
2296 + "cr_year_print": 2020,
2297 + "cr_authors": [
2298 + "PELGER"
2299 + ],
2300 + "cr_volume": "75",
2301 + "cr_issue": "4",
2302 + "cr_page": "2179-2220",
2303 + "cr_doi": "10.1111/jofi.12898",
2304 + "bib_volume": "75",
2305 + "bib_number": "4",
2306 + "bib_pages": "2179--2220",
2307 + "title_sim": 0.982,
2308 + "first_author_ok": true,
2309 + "year_ok": true,
2310 + "status": "OK"
2311 + },
2312 + {
2313 + "key": "aitsahalia2019principal",
2314 + "doi": "10.1080/01621459.2017.1401542",
2315 + "bib_title": "Principal Component Analysis of High-Frequency Data",
2316 + "bib_year": "2019",
2317 + "bib_journal": "Journal of the American Statistical Association",
2318 + "cr_title": "Principal Component Analysis of High-Frequency Data",
2319 + "cr_journal": "Journal of the American Statistical Association",
2320 + "cr_year_issued": 2018,
2321 + "cr_year_print": 2019,
2322 + "cr_authors": [
2323 + "A\u00eft-Sahalia",
2324 + "Xiu"
2325 + ],
2326 + "cr_volume": "114",
2327 + "cr_issue": "525",
2328 + "cr_page": "287-303",
2329 + "cr_doi": "10.1080/01621459.2017.1401542",
2330 + "bib_volume": "114",
2331 + "bib_number": "525",
2332 + "bib_pages": "287--303",
2333 + "title_sim": 1.0,
2334 + "first_author_ok": false,
2335 + "year_ok": true,
2336 + "status": "CHECK"
2337 + },
2338 + {
2339 + "key": "sirignano2019universal",
2340 + "doi": "10.1080/14697688.2019.1622295",
2341 + "bib_title": "Universal features of price formation in financial markets: perspectives from deep learning",
2342 + "bib_year": "2019",
2343 + "bib_journal": "Quantitative Finance",
2344 + "cr_title": "Universal features of price formation in financial markets: perspectives from deep learning",
2345 + "cr_journal": "Quantitative Finance",
2346 + "cr_year_issued": 2019,
2347 + "cr_year_print": 2019,
2348 + "cr_authors": [
2349 + "Sirignano",
2350 + "Cont"
2351 + ],
2352 + "cr_volume": "19",
2353 + "cr_issue": "9",
2354 + "cr_page": "1449-1459",
2355 + "cr_doi": "10.1080/14697688.2019.1622295",
2356 + "bib_volume": "19",
2357 + "bib_number": "9",
2358 + "bib_pages": "1449--1459",
2359 + "title_sim": 1.0,
2360 + "first_author_ok": true,
2361 + "year_ok": true,
2362 + "status": "OK"
2363 + },
2364 + {
2365 + "key": "easley2021microstructure",
2366 + "doi": "10.1093/rfs/hhaa078",
2367 + "bib_title": "Microstructure in the Machine Age",
2368 + "bib_year": "2021",
2369 + "bib_journal": "Review of Financial Studies",
2370 + "cr_title": "Microstructure in the Machine Age",
2371 + "cr_journal": "The Review of Financial Studies",
2372 + "cr_year_issued": 2020,
2373 + "cr_year_print": 2021,
2374 + "cr_authors": [
2375 + "Easley",
2376 + "L\u00f3pez de Prado",
2377 + "O\u2019Hara",
2378 + "Zhang"
2379 + ],
2380 + "cr_volume": "34",
2381 + "cr_issue": "7",
2382 + "cr_page": "3316-3363",
2383 + "cr_doi": "10.1093/rfs/hhaa078",
2384 + "bib_volume": "34",
2385 + "bib_number": "7",
2386 + "bib_pages": "3316--3363",
2387 + "title_sim": 1.0,
2388 + "first_author_ok": true,
2389 + "year_ok": true,
2390 + "status": "OK"
2391 + },
2392 + {
2393 + "key": "baltussen2019indexing",
2394 + "doi": "10.1016/j.jfineco.2018.07.016",
2395 + "bib_title": "Indexing and stock market serial dependence around the world",
2396 + "bib_year": "2019",
2397 + "bib_journal": "Journal of Financial Economics",
2398 + "cr_title": "Indexing and stock market serial dependence around the world",
2399 + "cr_journal": "Journal of Financial Economics",
2400 + "cr_year_issued": 2019,
2401 + "cr_year_print": 2019,
2402 + "cr_authors": [
2403 + "Baltussen",
2404 + "van Bekkum",
2405 + "Da"
2406 + ],
2407 + "cr_volume": "132",
2408 + "cr_issue": "1",
2409 + "cr_page": "26-48",
2410 + "cr_doi": "10.1016/j.jfineco.2018.07.016",
2411 + "bib_volume": "132",
2412 + "bib_number": "1",
2413 + "bib_pages": "26--48",
2414 + "title_sim": 1.0,
2415 + "first_author_ok": true,
2416 + "year_ok": true,
2417 + "status": "OK"
2418 + },
2419 + {
2420 + "key": "da2018exchange",
2421 + "doi": "10.1111/eufm.12137",
2422 + "bib_title": "Exchange traded funds and asset return correlations",
2423 + "bib_year": "2018",
2424 + "bib_journal": "European Financial Management",
2425 + "cr_title": "Exchange traded funds and asset return correlations",
2426 + "cr_journal": "European Financial Management",
2427 + "cr_year_issued": 2017,
2428 + "cr_year_print": 2018,
2429 + "cr_authors": [
2430 + "Da",
2431 + "Shive"
2432 + ],
2433 + "cr_volume": "24",
2434 + "cr_issue": "1",
2435 + "cr_page": "136-168",
2436 + "cr_doi": "10.1111/eufm.12137",
2437 + "bib_volume": "24",
2438 + "bib_number": "1",
2439 + "bib_pages": "136--168",
2440 + "title_sim": 1.0,
2441 + "first_author_ok": true,
2442 + "year_ok": true,
2443 + "status": "OK"
2444 + },
2445 + {
2446 + "key": "israeli2017dark",
2447 + "doi": "10.1007/s11142-017-9400-8",
2448 + "bib_title": "Is there a dark side to exchange traded funds? An information perspective",
2449 + "bib_year": "2017",
2450 + "bib_journal": "Review of Accounting Studies",
2451 + "cr_title": "Is there a dark side to exchange traded funds? An information perspective",
2452 + "cr_journal": "Review of Accounting Studies",
2453 + "cr_year_issued": 2017,
2454 + "cr_year_print": 2017,
2455 + "cr_authors": [
2456 + "Israeli",
2457 + "Lee",
2458 + "Sridharan"
2459 + ],
2460 + "cr_volume": "22",
2461 + "cr_issue": "3",
2462 + "cr_page": "1048-1083",
2463 + "cr_doi": "10.1007/s11142-017-9400-8",
2464 + "bib_volume": "22",
2465 + "bib_number": "3",
2466 + "bib_pages": "1048--1083",
2467 + "title_sim": 1.0,
2468 + "first_author_ok": true,
2469 + "year_ok": true,
2470 + "status": "OK"
2471 + },
2472 + {
2473 + "key": "glosten2021etf",
2474 + "doi": "10.1287/mnsc.2019.3427",
2475 + "bib_title": "{ETF} Activity and Informational Efficiency of Underlying Securities",
2476 + "bib_year": "2021",
2477 + "bib_journal": "Management Science",
2478 + "cr_title": "ETF Activity and Informational Efficiency of Underlying Securities",
2479 + "cr_journal": "Management Science",
2480 + "cr_year_issued": 2021,
2481 + "cr_year_print": 2021,
2482 + "cr_authors": [
2483 + "Glosten",
2484 + "Nallareddy",
2485 + "Zou"
2486 + ],
2487 + "cr_volume": "67",
2488 + "cr_issue": "1",
2489 + "cr_page": "22-47",
2490 + "cr_doi": "10.1287/mnsc.2019.3427",
2491 + "bib_volume": "67",
2492 + "bib_number": "1",
2493 + "bib_pages": "22--47",
2494 + "title_sim": 1.0,
2495 + "first_author_ok": true,
2496 + "year_ok": true,
2497 + "status": "OK"
2498 + },
2499 + {
2500 + "key": "coles2022index",
2501 + "doi": "10.1016/j.jfineco.2022.05.007",
2502 + "bib_title": "On index investing",
2503 + "bib_year": "2022",
2504 + "bib_journal": "Journal of Financial Economics",
2505 + "cr_title": "On index investing",
2506 + "cr_journal": "Journal of Financial Economics",
2507 + "cr_year_issued": 2022,
2508 + "cr_year_print": 2022,
2509 + "cr_authors": [
2510 + "Coles",
2511 + "Heath",
2512 + "Ringgenberg"
2513 + ],
2514 + "cr_volume": "145",
2515 + "cr_issue": "3",
2516 + "cr_page": "665-683",
2517 + "cr_doi": "10.1016/j.jfineco.2022.05.007",
2518 + "bib_volume": "145",
2519 + "bib_number": "3",
2520 + "bib_pages": "665--683",
2521 + "title_sim": 1.0,
2522 + "first_author_ok": true,
2523 + "year_ok": true,
2524 + "status": "OK"
2525 + },
2526 + {
2527 + "key": "bollerslev2020realized",
2528 + "doi": "10.3982/ecta17056",
2529 + "bib_title": "Realized Semicovariances",
2530 + "bib_year": "2020",
2531 + "bib_journal": "Econometrica",
2532 + "cr_title": "Realized Semicovariances",
2533 + "cr_journal": "Econometrica",
2534 + "cr_year_issued": 2020,
2535 + "cr_year_print": null,
2536 + "cr_authors": [
2537 + "Bollerslev",
2538 + "Li",
2539 + "Patton",
2540 + "Quaedvlieg"
2541 + ],
2542 + "cr_volume": "88",
2543 + "cr_issue": "4",
2544 + "cr_page": "1515-1551",
2545 + "cr_doi": "10.3982/ecta17056",
2546 + "bib_volume": "88",
2547 + "bib_number": "4",
2548 + "bib_pages": "1515--1551",
2549 + "title_sim": 1.0,
2550 + "first_author_ok": true,
2551 + "year_ok": true,
2552 + "status": "OK"
2553 + },
2554 + {
2555 + "key": "biais2015equilibrium",
2556 + "doi": "10.1016/j.jfineco.2015.03.004",
2557 + "bib_title": "Equilibrium fast trading",
2558 + "bib_year": "2015",
2559 + "bib_journal": "Journal of Financial Economics",
2560 + "cr_title": "Equilibrium fast trading",
2561 + "cr_journal": "Journal of Financial Economics",
2562 + "cr_year_issued": 2015,
2563 + "cr_year_print": 2015,
2564 + "cr_authors": [
2565 + "Biais",
2566 + "Foucault",
2567 + "Moinas"
2568 + ],
2569 + "cr_volume": "116",
2570 + "cr_issue": "2",
2571 + "cr_page": "292-313",
2572 + "cr_doi": "10.1016/j.jfineco.2015.03.004",
2573 + "bib_volume": "116",
2574 + "bib_number": "2",
2575 + "bib_pages": "292--313",
2576 + "title_sim": 1.0,
2577 + "first_author_ok": true,
2578 + "year_ok": true,
2579 + "status": "OK"
2580 + },
2581 + {
2582 + "key": "foucault2016news",
2583 + "doi": "10.1111/jofi.12302",
2584 + "bib_title": "News Trading and Speed",
2585 + "bib_year": "2016",
2586 + "bib_journal": "Journal of Finance",
2587 + "cr_title": "News Trading and Speed",
2588 + "cr_journal": "The Journal of Finance",
2589 + "cr_year_issued": 2016,
2590 + "cr_year_print": 2016,
2591 + "cr_authors": [
2592 + "FOUCAULT",
2593 + "HOMBERT",
2594 + "RO\u015eU"
2595 + ],
2596 + "cr_volume": "71",
2597 + "cr_issue": "1",
2598 + "cr_page": "335-382",
2599 + "cr_doi": "10.1111/jofi.12302",
2600 + "bib_volume": "71",
2601 + "bib_number": "1",
2602 + "bib_pages": "335--382",
2603 + "title_sim": 1.0,
2604 + "first_author_ok": false,
2605 + "year_ok": true,
2606 + "status": "CHECK"
2607 + },
2608 + {
2609 + "key": "foucault2017toxic",
2610 + "doi": "10.1093/rfs/hhw103",
2611 + "bib_title": "Toxic Arbitrage",
2612 + "bib_year": "2017",
2613 + "bib_journal": "Review of Financial Studies",
2614 + "cr_title": "Toxic Arbitrage",
2615 + "cr_journal": "The Review of Financial Studies",
2616 + "cr_year_issued": 2016,
2617 + "cr_year_print": 2017,
2618 + "cr_authors": [
2619 + "Foucault",
2620 + "Kozhan",
2621 + "Tham"
2622 + ],
2623 + "cr_volume": "30",
2624 + "cr_issue": "4",
2625 + "cr_page": "1053-1094",
2626 + "cr_doi": "10.1093/rfs/hhw103",
2627 + "bib_volume": "30",
2628 + "bib_number": "4",
2629 + "bib_pages": "1053--1094",
2630 + "title_sim": 1.0,
2631 + "first_author_ok": true,
2632 + "year_ok": true,
2633 + "status": "OK"
2634 + },
2635 + {
2636 + "key": "aquilina2022quantifying",
2637 + "doi": "10.1093/qje/qjab032",
2638 + "bib_title": "Quantifying the High-Frequency Trading ``Arms Race''",
2639 + "bib_year": "2022",
2640 + "bib_journal": "Quarterly Journal of Economics",
2641 + "cr_title": "Quantifying the High-Frequency Trading \u201cArms Race\u201d",
2642 + "cr_journal": "The Quarterly Journal of Economics",
2643 + "cr_year_issued": 2021,
2644 + "cr_year_print": 2021,
2645 + "cr_authors": [
2646 + "Aquilina",
2647 + "Budish",
2648 + "O\u2019Neill"
2649 + ],
2650 + "cr_volume": "137",
2651 + "cr_issue": "1",
2652 + "cr_page": "493-564",
2653 + "cr_doi": "10.1093/qje/qjab032",
2654 + "bib_volume": "137",
2655 + "bib_number": "1",
2656 + "bib_pages": "493--564",
2657 + "title_sim": 0.98,
2658 + "first_author_ok": true,
2659 + "year_ok": false,
2660 + "status": "CHECK"
2661 + },
2662 + {
2663 + "key": "weller2018does",
2664 + "doi": "10.1093/rfs/hhx137",
2665 + "bib_title": "Does Algorithmic Trading Reduce Information Acquisition?",
2666 + "bib_year": "2018",
2667 + "bib_journal": "Review of Financial Studies",
2668 + "cr_title": "Does Algorithmic Trading Reduce Information Acquisition?",
2669 + "cr_journal": "The Review of Financial Studies",
2670 + "cr_year_issued": 2017,
2671 + "cr_year_print": 2018,
2672 + "cr_authors": [
2673 + "Weller"
2674 + ],
2675 + "cr_volume": "31",
2676 + "cr_issue": "6",
2677 + "cr_page": "2184-2226",
2678 + "cr_doi": "10.1093/rfs/hhx137",
2679 + "bib_volume": "31",
2680 + "bib_number": "6",
2681 + "bib_pages": "2184--2226",
2682 + "title_sim": 1.0,
2683 + "first_author_ok": true,
2684 + "year_ok": true,
2685 + "status": "OK"
2686 + },
2687 + {
2688 + "key": "vankervel2019high",
2689 + "doi": "10.1111/jofi.12759",
2690 + "bib_title": "High-Frequency Trading around Large Institutional Orders",
2691 + "bib_year": "2019",
2692 + "bib_journal": "Journal of Finance",
2693 + "cr_title": "High\u2010Frequency Trading around Large Institutional Orders",
2694 + "cr_journal": "The Journal of Finance",
2695 + "cr_year_issued": 2019,
2696 + "cr_year_print": 2019,
2697 + "cr_authors": [
2698 + "VAN KERVEL",
2699 + "MENKVELD"
2700 + ],
2701 + "cr_volume": "74",
2702 + "cr_issue": "3",
2703 + "cr_page": "1091-1137",
2704 + "cr_doi": "10.1111/jofi.12759",
2705 + "bib_volume": "74",
2706 + "bib_number": "3",
2707 + "bib_pages": "1091--1137",
2708 + "title_sim": 0.982,
2709 + "first_author_ok": true,
2710 + "year_ok": true,
2711 + "status": "OK"
2712 + },
2713 + {
2714 + "key": "chaboud2014rise",
2715 + "doi": "10.1111/jofi.12186",
2716 + "bib_title": "Rise of the Machines: Algorithmic Trading in the Foreign Exchange Market",
2717 + "bib_year": "2014",
2718 + "bib_journal": "Journal of Finance",
2719 + "cr_title": "Rise of the Machines: Algorithmic Trading in the Foreign Exchange Market",
2720 + "cr_journal": "The Journal of Finance",
2721 + "cr_year_issued": 2014,
2722 + "cr_year_print": 2014,
2723 + "cr_authors": [
2724 + "CHABOUD",
2725 + "CHIQUOINE",
2726 + "HJALMARSSON",
2727 + "VEGA"
2728 + ],
2729 + "cr_volume": "69",
2730 + "cr_issue": "5",
2731 + "cr_page": "2045-2084",
2732 + "cr_doi": "10.1111/jofi.12186",
2733 + "bib_volume": "69",
2734 + "bib_number": "5",
2735 + "bib_pages": "2045--2084",
2736 + "title_sim": 1.0,
2737 + "first_author_ok": true,
2738 + "year_ok": true,
2739 + "status": "OK"
2740 + },
2741 + {
2742 + "key": "menkveld2016economics",
2743 + "doi": "10.1146/annurev-financial-121415-033010",
2744 + "bib_title": "The Economics of High-Frequency Trading: Taking Stock",
2745 + "bib_year": "2016",
2746 + "bib_journal": "Annual Review of Financial Economics",
2747 + "cr_title": "The Economics of High-Frequency Trading: Taking Stock",
2748 + "cr_journal": "Annual Review of Financial Economics",
2749 + "cr_year_issued": 2016,
2750 + "cr_year_print": 2016,
2751 + "cr_authors": [
2752 + "Menkveld"
2753 + ],
2754 + "cr_volume": "8",
2755 + "cr_issue": "1",
2756 + "cr_page": "1-24",
2757 + "cr_doi": "10.1146/annurev-financial-121415-033010",
2758 + "bib_volume": "8",
2759 + "bib_number": "1",
2760 + "bib_pages": "1--24",
2761 + "title_sim": 1.0,
2762 + "first_author_ok": true,
2763 + "year_ok": true,
2764 + "status": "OK"
2765 + },
2766 + {
2767 + "key": "hendershott2011automation",
2768 + "doi": "10.1016/j.finmar.2011.02.003",
2769 + "bib_title": "Automation, speed, and stock market quality: The {NYSE}'s Hybrid",
2770 + "bib_year": "2011",
2771 + "bib_journal": "Journal of Financial Markets",
2772 + "cr_title": "Automation, speed, and stock market quality: The NYSE's Hybrid",
2773 + "cr_journal": "Journal of Financial Markets",
2774 + "cr_year_issued": 2011,
2775 + "cr_year_print": 2011,
2776 + "cr_authors": [
2777 + "Hendershott",
2778 + "Moulton"
2779 + ],
2780 + "cr_volume": "14",
2781 + "cr_issue": "4",
2782 + "cr_page": "568-604",
2783 + "cr_doi": "10.1016/j.finmar.2011.02.003",
2784 + "bib_volume": "14",
2785 + "bib_number": "4",
2786 + "bib_pages": "568--604",
2787 + "title_sim": 1.0,
2788 + "first_author_ok": true,
2789 + "year_ok": true,
2790 + "status": "OK"
2791 + },
2792 + {
2793 + "key": "ohara2011fragmentation",
2794 + "doi": "10.1016/j.jfineco.2011.02.006",
2795 + "bib_title": "Is market fragmentation harming market quality?",
2796 + "bib_year": "2011",
2797 + "bib_journal": "Journal of Financial Economics",
2798 + "cr_title": "Is market fragmentation harming market quality?",
2799 + "cr_journal": "Journal of Financial Economics",
2800 + "cr_year_issued": 2011,
2801 + "cr_year_print": 2011,
2802 + "cr_authors": [
2803 + "O'Hara",
2804 + "Ye"
2805 + ],
2806 + "cr_volume": "100",
2807 + "cr_issue": "3",
2808 + "cr_page": "459-474",
2809 + "cr_doi": "10.1016/j.jfineco.2011.02.006",
2810 + "bib_volume": "100",
2811 + "bib_number": "3",
2812 + "bib_pages": "459--474",
2813 + "title_sim": 1.0,
2814 + "first_author_ok": true,
2815 + "year_ok": true,
2816 + "status": "OK"
2817 + },
2818 + {
2819 + "key": "bessembinder2003trade",
2820 + "doi": "10.2307/4126742",
2821 + "bib_title": "Trade Execution Costs and Market Quality after Decimalization",
2822 + "bib_year": "2003",
2823 + "bib_journal": "Journal of Financial and Quantitative Analysis",
2824 + "cr_title": "Trade Execution Costs and Market Quality after Decimalization",
2825 + "cr_journal": "The Journal of Financial and Quantitative Analysis",
2826 + "cr_year_issued": 2003,
2827 + "cr_year_print": 2003,
2828 + "cr_authors": [
2829 + "Bessembinder"
2830 + ],
2831 + "cr_volume": "38",
2832 + "cr_issue": "4",
2833 + "cr_page": "747",
2834 + "cr_doi": "10.2307/4126742",
2835 + "bib_volume": "38",
2836 + "bib_number": "4",
2837 + "bib_pages": "747--777",
2838 + "title_sim": 1.0,
2839 + "first_author_ok": true,
2840 + "year_ok": true,
2841 + "status": "OK"
2842 + },
2843 + {
2844 + "key": "chordia2008liquidity",
2845 + "doi": "10.1016/j.jfineco.2007.03.005",
2846 + "bib_title": "Liquidity and market efficiency",
2847 + "bib_year": "2008",
2848 + "bib_journal": "Journal of Financial Economics",
2849 + "cr_title": "Liquidity and market efficiency\u2606",
2850 + "cr_journal": "Journal of Financial Economics",
2851 + "cr_year_issued": 2008,
2852 + "cr_year_print": 2008,
2853 + "cr_authors": [
2854 + "CHORDIA",
2855 + "ROLL",
2856 + "SUBRAHMANYAM"
2857 + ],
2858 + "cr_volume": "87",
2859 + "cr_issue": "2",
2860 + "cr_page": "249-268",
2861 + "cr_doi": "10.1016/j.jfineco.2007.03.005",
2862 + "bib_volume": "87",
2863 + "bib_number": "2",
2864 + "bib_pages": "249--268",
2865 + "title_sim": 0.984,
2866 + "first_author_ok": true,
2867 + "year_ok": true,
2868 + "status": "OK"
2869 + },
2870 + {
2871 + "key": "chordia2011recent",
2872 + "doi": "10.1016/j.jfineco.2011.03.008",
2873 + "bib_title": "Recent trends in trading activity and market quality",
2874 + "bib_year": "2011",
2875 + "bib_journal": "Journal of Financial Economics",
2876 + "cr_title": "Recent trends in trading activity and market quality",
2877 + "cr_journal": "Journal of Financial Economics",
2878 + "cr_year_issued": 2011,
2879 + "cr_year_print": 2011,
2880 + "cr_authors": [
2881 + "Chordia",
2882 + "Roll",
2883 + "Subrahmanyam"
2884 + ],
2885 + "cr_volume": "101",
2886 + "cr_issue": "2",
2887 + "cr_page": "243-263",
2888 + "cr_doi": "10.1016/j.jfineco.2011.03.008",
2889 + "bib_volume": "101",
2890 + "bib_number": "2",
2891 + "bib_pages": "243--263",
2892 + "title_sim": 1.0,
2893 + "first_author_ok": true,
2894 + "year_ok": true,
2895 + "status": "OK"
2896 + }
2897 +]
\ No newline at end of file
added outputs/diagnostics/bibliography/crossref_new_candidates.json +1504 −0
@@ -0,0 +1,1504 @@
1 +{
2 + "toth2009epps": [
3 + {
4 + "doi": "10.1080/14697680802595668",
5 + "title": "The Epps effect revisited",
6 + "journal": "Quantitative Finance",
7 + "year": 2009,
8 + "print_year": 2009,
9 + "volume": "9",
10 + "issue": "7",
11 + "page": "793-802",
12 + "authors": [
13 + [
14 + "Bence",
15 + "T\u00f3th"
16 + ],
17 + [
18 + "J\u00e1nos",
19 + "Kert\u00e9sz"
20 + ]
21 + ]
22 + },
23 + {
24 + "doi": "10.1002/9780470455289",
25 + "title": "Quantitative Finance",
26 + "journal": "",
27 + "year": 2009,
28 + "print_year": 2009,
29 + "volume": null,
30 + "issue": null,
31 + "page": null,
32 + "authors": [
33 + [
34 + "T. W.",
35 + "Epps"
36 + ]
37 + ]
38 + },
39 + {
40 + "doi": "10.1080/14697688.2019.1659992",
41 + "title": "A closed-form formula characterization of the Epps effect",
42 + "journal": "Quantitative Finance",
43 + "year": 2019,
44 + "print_year": 2020,
45 + "volume": "20",
46 + "issue": "2",
47 + "page": "243-254",
48 + "authors": [
49 + [
50 + "Giuseppe",
51 + "Buccheri"
52 + ],
53 + [
54 + "Giulia",
55 + "Livieri"
56 + ],
57 + [
58 + "Davide",
59 + "Pirino"
60 + ],
61 + [
62 + "Alessandro",
63 + "Pollastri"
64 + ]
65 + ]
66 + }
67 + ],
68 + "reno2003closer": [
69 + {
70 + "doi": "10.1142/s0219024903001839",
71 + "title": "A CLOSER LOOK AT THE EPPS EFFECT",
72 + "journal": "International Journal of Theoretical and Applied Finance",
73 + "year": 2003,
74 + "print_year": 2003,
75 + "volume": "06",
76 + "issue": "01",
77 + "page": "87-102",
78 + "authors": [
79 + [
80 + "ROBERTO",
81 + "REN\u00d2"
82 + ]
83 + ]
84 + },
85 + {
86 + "doi": "10.2139/ssrn.314723",
87 + "title": "A Closer Look at the Epps Effect",
88 + "journal": "",
89 + "year": 2002,
90 + "print_year": null,
91 + "volume": null,
92 + "issue": null,
93 + "page": null,
94 + "authors": [
95 + [
96 + "Roberto",
97 + "Ren\u00f2"
98 + ]
99 + ]
100 + },
101 + {
102 + "doi": "10.2139/ssrn.917374",
103 + "title": "A Closer Look at Business Education - Finance",
104 + "journal": "",
105 + "year": 2006,
106 + "print_year": null,
107 + "volume": null,
108 + "issue": null,
109 + "page": null,
110 + "authors": [
111 + [
112 + "Rich",
113 + "Leimsider"
114 + ]
115 + ]
116 + }
117 + ],
118 + "hayashi2005covariance": [
119 + {
120 + "doi": "10.3150/bj/1116340299",
121 + "title": "On covariance estimation of non-synchronously observed diffusion processes",
122 + "journal": "Bernoulli",
123 + "year": 2005,
124 + "print_year": 2005,
125 + "volume": "11",
126 + "issue": "2",
127 + "page": "359-379",
128 + "authors": [
129 + [
130 + "Takaki",
131 + "Hayashi"
132 + ],
133 + [
134 + "Nakahiro",
135 + "Yoshida"
136 + ]
137 + ]
138 + },
139 + {
140 + "doi": "10.3150/07-bej1040",
141 + "title": "Moment estimation for ergodic diffusion processes",
142 + "journal": "Bernoulli",
143 + "year": 2007,
144 + "print_year": 2007,
145 + "volume": "13",
146 + "issue": "4",
147 + "page": "933-951",
148 + "authors": [
149 + [
150 + "Yury A.",
151 + "Kutoyants"
152 + ],
153 + [
154 + "Nakahiro",
155 + "Yoshida"
156 + ]
157 + ]
158 + },
159 + {
160 + "doi": "10.1007/s10463-007-0138-0",
161 + "title": "Asymptotic normality of a covariance estimator for nonsynchronously observed diffusion processes",
162 + "journal": "Annals of the Institute of Statistical Mathematics",
163 + "year": 2007,
164 + "print_year": 2008,
165 + "volume": "60",
166 + "issue": "2",
167 + "page": "367-406",
168 + "authors": [
169 + [
170 + "Takaki",
171 + "Hayashi"
172 + ],
173 + [
174 + "Nakahiro",
175 + "Yoshida"
176 + ]
177 + ]
178 + }
179 + ],
180 + "zhang2011estimating": [
181 + {
182 + "doi": "10.1016/j.jeconom.2010.03.012",
183 + "title": "Estimating covariation: Epps effect, microstructure noise",
184 + "journal": "Journal of Econometrics",
185 + "year": 2011,
186 + "print_year": 2011,
187 + "volume": "160",
188 + "issue": "1",
189 + "page": "33-47",
190 + "authors": [
191 + [
192 + "Lan",
193 + "Zhang"
194 + ]
195 + ]
196 + },
197 + {
198 + "doi": "10.2139/ssrn.885438",
199 + "title": "Estimating Covariation: Epps Effect, Microstructure Noise",
200 + "journal": "SSRN Electronic Journal",
201 + "year": 2006,
202 + "print_year": null,
203 + "volume": null,
204 + "issue": null,
205 + "page": null,
206 + "authors": [
207 + [
208 + "Lan",
209 + "Zhang"
210 + ]
211 + ]
212 + },
213 + {
214 + "doi": "10.1093/jjfinec/nbq031",
215 + "title": "Estimating Covariance via Fourier Method in the Presence of Asynchronous Trading and Microstructure Noise",
216 + "journal": "Journal of Financial Econometrics",
217 + "year": 2011,
218 + "print_year": 2011,
219 + "volume": "9",
220 + "issue": "2",
221 + "page": "367-408",
222 + "authors": [
223 + [
224 + "M. E.",
225 + "Mancino"
226 + ],
227 + [
228 + "S.",
229 + "Sanfelici"
230 + ]
231 + ]
232 + }
233 + ],
234 + "jacod2009testing": [
235 + {
236 + "doi": "10.1214/08-aos624",
237 + "title": "Testing for common arrivals of jumps for discretely observed multidimensional processes",
238 + "journal": "The Annals of Statistics",
239 + "year": 2009,
240 + "print_year": 2009,
241 + "volume": "37",
242 + "issue": "4",
243 + "page": "1792-1838",
244 + "authors": [
245 + [
246 + "Jean",
247 + "Jacod"
248 + ],
249 + [
250 + "Viktor",
251 + "Todorov"
252 + ]
253 + ]
254 + },
255 + {
256 + "doi": "10.1214/07-aos568",
257 + "title": "Testing for jumps in a discretely observed process",
258 + "journal": "The Annals of Statistics",
259 + "year": 2009,
260 + "print_year": 2009,
261 + "volume": "37",
262 + "issue": "1",
263 + "page": "184-222",
264 + "authors": [
265 + [
266 + "Yacine",
267 + "A\u00eft-Sahalia"
268 + ],
269 + [
270 + "Jean",
271 + "Jacod"
272 + ]
273 + ]
274 + },
275 + {
276 + "doi": "10.2139/ssrn.1069872",
277 + "title": "Testing for Jumps in a Discretely Observed Process",
278 + "journal": "",
279 + "year": 2007,
280 + "print_year": null,
281 + "volume": null,
282 + "issue": null,
283 + "page": null,
284 + "authors": [
285 + [
286 + "Yacine",
287 + "Ait-Sahalia"
288 + ],
289 + [
290 + "Jean",
291 + "Jacod"
292 + ]
293 + ]
294 + }
295 + ],
296 + "gilder2014cojumps": [
297 + {
298 + "doi": "10.1016/j.jbankfin.2013.04.025",
299 + "title": "Cojumps in stock prices: Empirical evidence",
300 + "journal": "Journal of Banking &amp; Finance",
301 + "year": 2014,
302 + "print_year": 2014,
303 + "volume": "40",
304 + "issue": null,
305 + "page": "443-459",
306 + "authors": [
307 + [
308 + "Dudley",
309 + "Gilder"
310 + ],
311 + [
312 + "Mark B.",
313 + "Shackleton"
314 + ],
315 + [
316 + "Stephen J.",
317 + "Taylor"
318 + ]
319 + ]
320 + },
321 + {
322 + "doi": "10.2139/ssrn.2028443",
323 + "title": "Cojumps in Stock Prices: Empirical Evidence",
324 + "journal": "",
325 + "year": 2012,
326 + "print_year": null,
327 + "volume": null,
328 + "issue": null,
329 + "page": null,
330 + "authors": [
331 + [
332 + "Dudley",
333 + "Gilder"
334 + ],
335 + [
336 + "Mark B.",
337 + "Shackleton"
338 + ],
339 + [
340 + "Stephen J.",
341 + "Taylor"
342 + ]
343 + ]
344 + },
345 + {
346 + "doi": "10.2139/ssrn.1343779",
347 + "title": "An Empirical Investigation of Intraday Jumps and Cojumps in US Equities",
348 + "journal": "",
349 + "year": 2009,
350 + "print_year": null,
351 + "volume": null,
352 + "issue": null,
353 + "page": null,
354 + "authors": [
355 + [
356 + "Dudley",
357 + "Gilder"
358 + ]
359 + ]
360 + }
361 + ],
362 + "bollerslev2018volume": [
363 + {
364 + "doi": "10.1093/restud/rdy003",
365 + "title": "Volume, Volatility, and Public News Announcements",
366 + "journal": "The Review of Economic Studies",
367 + "year": 2018,
368 + "print_year": 2018,
369 + "volume": "85",
370 + "issue": "4",
371 + "page": "2005-2041",
372 + "authors": [
373 + [
374 + "Tim",
375 + "Bollerslev"
376 + ],
377 + [
378 + "Jia",
379 + "Li"
380 + ],
381 + [
382 + "Yuan",
383 + "Xue"
384 + ]
385 + ]
386 + },
387 + {
388 + "doi": "10.2307/2298012",
389 + "title": "Intra-Day and Inter-Market Volatility in Foreign Exchange Rates",
390 + "journal": "The Review of Economic Studies",
391 + "year": 1991,
392 + "print_year": 1991,
393 + "volume": "58",
394 + "issue": "3",
395 + "page": "565",
396 + "authors": [
397 + [
398 + "Richard T.",
399 + "Baillie"
400 + ],
401 + [
402 + "Tim",
403 + "Bollerslev"
404 + ]
405 + ]
406 + },
407 + {
408 + "doi": "10.3386/w5783",
409 + "title": "DM-Dollar Volatility: Intraday Activity Patterns, Macroeconomic Announcements, and Longer Run Dependencies",
410 + "journal": "",
411 + "year": 1996,
412 + "print_year": null,
413 + "volume": null,
414 + "issue": null,
415 + "page": null,
416 + "authors": [
417 + [
418 + "Torben",
419 + "Andersen"
420 + ],
421 + [
422 + "Tim",
423 + "Bollerslev"
424 + ]
425 + ]
426 + }
427 + ],
428 + "kyle2016invariance": [
429 + {
430 + "doi": "10.3982/ecta10486",
431 + "title": "Market Microstructure Invariance: Empirical Hypotheses",
432 + "journal": "Econometrica",
433 + "year": 2016,
434 + "print_year": null,
435 + "volume": "84",
436 + "issue": "4",
437 + "page": "1345-1404",
438 + "authors": [
439 + [
440 + "Albert S.",
441 + "Kyle"
442 + ],
443 + [
444 + "Anna A.",
445 + "Obizhaeva"
446 + ]
447 + ]
448 + },
449 + {
450 + "doi": "10.2139/ssrn.2722524",
451 + "title": "Market Microstructure Invariance: Empirical Hypotheses",
452 + "journal": "SSRN Electronic Journal",
453 + "year": 2016,
454 + "print_year": null,
455 + "volume": null,
456 + "issue": null,
457 + "page": null,
458 + "authors": [
459 + [
460 + "Albert S.",
461 + "Kyle"
462 + ],
463 + [
464 + "Anna A.",
465 + "Obizhaeva"
466 + ]
467 + ]
468 + },
469 + {
470 + "doi": "10.2139/ssrn.1687965",
471 + "title": "Market Microstructure Invariance: Theory and Empirical Tests",
472 + "journal": "",
473 + "year": 2010,
474 + "print_year": null,
475 + "volume": null,
476 + "issue": null,
477 + "page": null,
478 + "authors": [
479 + [
480 + "Albert (Pete) S.",
481 + "Kyle"
482 + ],
483 + [
484 + "Anna A.",
485 + "Obizhaeva"
486 + ]
487 + ]
488 + }
489 + ],
490 + "boudt2011robust": [
491 + {
492 + "doi": "10.1016/j.jempfin.2010.11.005",
493 + "title": "Robust estimation of intraweek periodicity in volatility and jump detection",
494 + "journal": "Journal of Empirical Finance",
495 + "year": 2011,
496 + "print_year": 2011,
497 + "volume": "18",
498 + "issue": "2",
499 + "page": "353-367",
500 + "authors": [
501 + [
502 + "Kris",
503 + "Boudt"
504 + ],
505 + [
506 + "Christophe",
507 + "Croux"
508 + ],
509 + [
510 + "S\u00e9bastien",
511 + "Laurent"
512 + ]
513 + ]
514 + },
515 + {
516 + "doi": "10.2139/ssrn.1297371",
517 + "title": "Robust Estimation of Intraweek Periodicity in Volatility and Jump Detection",
518 + "journal": "",
519 + "year": 2008,
520 + "print_year": null,
521 + "volume": null,
522 + "issue": null,
523 + "page": null,
524 + "authors": [
525 + [
526 + "Kris",
527 + "Boudt"
528 + ],
529 + [
530 + "Christophe",
531 + "Croux"
532 + ],
533 + [
534 + "S\u00e9bastien",
535 + "Laurent"
536 + ]
537 + ]
538 + },
539 + {
540 + "doi": "10.2139/ssrn.1740369",
541 + "title": "Jump Robust Two Time Scale Covariance Estimation and Realized Volatility Budgets",
542 + "journal": "",
543 + "year": 2011,
544 + "print_year": null,
545 + "volume": null,
546 + "issue": null,
547 + "page": null,
548 + "authors": [
549 + [
550 + "Kris",
551 + "Boudt"
552 + ],
553 + [
554 + "Jin",
555 + "Zhang"
556 + ]
557 + ]
558 + }
559 + ],
560 + "christensen2022drift": [
561 + {
562 + "doi": "10.1016/j.jeconom.2020.11.004",
563 + "title": "The drift burst hypothesis",
564 + "journal": "Journal of Econometrics",
565 + "year": 2022,
566 + "print_year": 2022,
567 + "volume": "227",
568 + "issue": "2",
569 + "page": "461-497",
570 + "authors": [
571 + [
572 + "Kim",
573 + "Christensen"
574 + ],
575 + [
576 + "Roel",
577 + "Oomen"
578 + ],
579 + [
580 + "Roberto",
581 + "Ren\u00f2"
582 + ]
583 + ]
584 + },
585 + {
586 + "doi": "10.2139/ssrn.2842535",
587 + "title": "The Drift Burst Hypothesis",
588 + "journal": "SSRN Electronic Journal",
589 + "year": 2016,
590 + "print_year": null,
591 + "volume": null,
592 + "issue": null,
593 + "page": null,
594 + "authors": [
595 + [
596 + "Kim",
597 + "Christensen"
598 + ],
599 + [
600 + "Roel C. A.",
601 + "Oomen"
602 + ],
603 + [
604 + "Roberto",
605 + "Renn"
606 + ]
607 + ]
608 + },
609 + {
610 + "doi": "10.1016/j.jeconom.2010.04.008",
611 + "title": "Realised quantile-based estimation of the integrated variance",
612 + "journal": "Journal of Econometrics",
613 + "year": 2010,
614 + "print_year": 2010,
615 + "volume": "159",
616 + "issue": "1",
617 + "page": "74-98",
618 + "authors": [
619 + [
620 + "Kim",
621 + "Christensen"
622 + ],
623 + [
624 + "Roel",
625 + "Oomen"
626 + ],
627 + [
628 + "Mark",
629 + "Podolskij"
630 + ]
631 + ]
632 + }
633 + ],
634 + "calcagnile2018collective": [
635 + {
636 + "doi": "10.1080/14697688.2017.1403141",
637 + "title": "Collective synchronization and high frequency systemic instabilities in financial markets",
638 + "journal": "Quantitative Finance",
639 + "year": 2017,
640 + "print_year": 2018,
641 + "volume": "18",
642 + "issue": "2",
643 + "page": "237-247",
644 + "authors": [
645 + [
646 + "Lucio Maria",
647 + "Calcagnile"
648 + ],
649 + [
650 + "Giacomo",
651 + "Bormetti"
652 + ],
653 + [
654 + "Michele",
655 + "Treccani"
656 + ],
657 + [
658 + "Stefano",
659 + "Marmi"
660 + ],
661 + [
662 + "Fabrizio",
663 + "Lillo"
664 + ]
665 + ]
666 + },
667 + {
668 + "doi": "10.1080/14697688.2014.996586",
669 + "title": "Modelling systemic price cojumps with Hawkes factor models",
670 + "journal": "Quantitative Finance",
671 + "year": 2015,
672 + "print_year": 2015,
673 + "volume": "15",
674 + "issue": "7",
675 + "page": "1137-1156",
676 + "authors": [
677 + [
678 + "Giacomo",
679 + "Bormetti"
680 + ],
681 + [
682 + "Lucio Maria",
683 + "Calcagnile"
684 + ],
685 + [
686 + "Michele",
687 + "Treccani"
688 + ],
689 + [
690 + "Fulvio",
691 + "Corsi"
692 + ],
693 + [
694 + "Stefano",
695 + "Marmi"
696 + ],
697 + [
698 + "Fabrizio",
699 + "Lillo"
700 + ]
701 + ]
702 + },
703 + {
704 + "doi": "10.2139/ssrn.2209139",
705 + "title": "Modelling Systemic Price Cojumps with Hawkes Factor Models",
706 + "journal": "",
707 + "year": 2013,
708 + "print_year": null,
709 + "volume": null,
710 + "issue": null,
711 + "page": null,
712 + "authors": [
713 + [
714 + "Giacomo",
715 + "Bormetti"
716 + ],
717 + [
718 + "Lucio Maria",
719 + "Calcagnile"
720 + ],
721 + [
722 + "Michele",
723 + "Treccani"
724 + ],
725 + [
726 + "Fulvio",
727 + "Corsi"
728 + ],
729 + [
730 + "Stefano",
731 + "Marmi"
732 + ],
733 + [
734 + "Fabrizio",
735 + "Lillo"
736 + ]
737 + ]
738 + }
739 + ],
740 + "bormetti2015modelling": [
741 + {
742 + "doi": "10.2139/ssrn.2209139",
743 + "title": "Modelling Systemic Price Cojumps with Hawkes Factor Models",
744 + "journal": "",
745 + "year": 2013,
746 + "print_year": null,
747 + "volume": null,
748 + "issue": null,
749 + "page": null,
750 + "authors": [
751 + [
752 + "Giacomo",
753 + "Bormetti"
754 + ],
755 + [
756 + "Lucio Maria",
757 + "Calcagnile"
758 + ],
759 + [
760 + "Michele",
761 + "Treccani"
762 + ],
763 + [
764 + "Fulvio",
765 + "Corsi"
766 + ],
767 + [
768 + "Stefano",
769 + "Marmi"
770 + ],
771 + [
772 + "Fabrizio",
773 + "Lillo"
774 + ]
775 + ]
776 + },
777 + {
778 + "doi": "10.1080/14697688.2014.996586",
779 + "title": "Modelling systemic price cojumps with Hawkes factor models",
780 + "journal": "Quantitative Finance",
781 + "year": 2015,
782 + "print_year": 2015,
783 + "volume": "15",
784 + "issue": "7",
785 + "page": "1137-1156",
786 + "authors": [
787 + [
788 + "Giacomo",
789 + "Bormetti"
790 + ],
791 + [
792 + "Lucio Maria",
793 + "Calcagnile"
794 + ],
795 + [
796 + "Michele",
797 + "Treccani"
798 + ],
799 + [
800 + "Fulvio",
801 + "Corsi"
802 + ],
803 + [
804 + "Stefano",
805 + "Marmi"
806 + ],
807 + [
808 + "Fabrizio",
809 + "Lillo"
810 + ]
811 + ]
812 + },
813 + {
814 + "doi": "10.1080/14697688.2017.1403141",
815 + "title": "Collective synchronization and high frequency systemic instabilities in financial markets",
816 + "journal": "Quantitative Finance",
817 + "year": 2017,
818 + "print_year": 2018,
819 + "volume": "18",
820 + "issue": "2",
821 + "page": "237-247",
822 + "authors": [
823 + [
824 + "Lucio Maria",
825 + "Calcagnile"
826 + ],
827 + [
828 + "Giacomo",
829 + "Bormetti"
830 + ],
831 + [
832 + "Michele",
833 + "Treccani"
834 + ],
835 + [
836 + "Stefano",
837 + "Marmi"
838 + ],
839 + [
840 + "Fabrizio",
841 + "Lillo"
842 + ]
843 + ]
844 + }
845 + ],
846 + "lo1990nonsynchronous": [
847 + {
848 + "doi": "10.1016/0304-4076(90)90098-e",
849 + "title": "An econometric analysis of nonsynchronous trading",
850 + "journal": "Journal of Econometrics",
851 + "year": 1990,
852 + "print_year": 1990,
853 + "volume": "45",
854 + "issue": "1-2",
855 + "page": "181-211",
856 + "authors": [
857 + [
858 + "Andrew W.",
859 + "Lo"
860 + ],
861 + [
862 + "A.",
863 + "Craig MacKinlay"
864 + ]
865 + ]
866 + },
867 + {
868 + "doi": "10.3386/w2960",
869 + "title": "An Econometric Analysis of Nonsynchronous Trading",
870 + "journal": "",
871 + "year": 1989,
872 + "print_year": null,
873 + "volume": null,
874 + "issue": null,
875 + "page": null,
876 + "authors": [
877 + [
878 + "Andrew",
879 + "Lo"
880 + ],
881 + [
882 + "A. Craig",
883 + "MacKinlay"
884 + ]
885 + ]
886 + },
887 + {
888 + "doi": "10.1515/9781400829095.85",
889 + "title": "4. An Econometric Analysis of Nonsynchronous Trading",
890 + "journal": "A Non-Random Walk Down Wall Street",
891 + "year": 2011,
892 + "print_year": 2011,
893 + "volume": null,
894 + "issue": null,
895 + "page": "85-114",
896 + "authors": []
897 + }
898 + ],
899 + "bollerslev2013jump": [
900 + {
901 + "doi": "10.1016/j.jeconom.2012.08.014",
902 + "title": "Jump tails, extreme dependencies, and the distribution of stock returns",
903 + "journal": "Journal of Econometrics",
904 + "year": 2013,
905 + "print_year": 2013,
906 + "volume": "172",
907 + "issue": "2",
908 + "page": "307-324",
909 + "authors": [
910 + [
911 + "Tim",
912 + "Bollerslev"
913 + ],
914 + [
915 + "Viktor",
916 + "Todorov"
917 + ],
918 + [
919 + "Sophia Zhengzi",
920 + "Li"
921 + ]
922 + ]
923 + },
924 + {
925 + "doi": "10.1016/j.jeconom.2014.05.007",
926 + "title": "Time-varying jump tails",
927 + "journal": "Journal of Econometrics",
928 + "year": 2014,
929 + "print_year": 2014,
930 + "volume": "183",
931 + "issue": "2",
932 + "page": "168-180",
933 + "authors": [
934 + [
935 + "Tim",
936 + "Bollerslev"
937 + ],
938 + [
939 + "Viktor",
940 + "Todorov"
941 + ]
942 + ]
943 + },
944 + {
945 + "doi": "10.2139/ssrn.1589753",
946 + "title": "Estimation of Jump Tails",
947 + "journal": "",
948 + "year": 2010,
949 + "print_year": null,
950 + "volume": null,
951 + "issue": null,
952 + "page": null,
953 + "authors": [
954 + [
955 + "Tim",
956 + "Bollerslev"
957 + ],
958 + [
959 + "Viktor",
960 + "Todorov"
961 + ]
962 + ]
963 + }
964 + ],
965 + "barndorffnielsen2006econometrics": [
966 + {
967 + "doi": "10.1093/jjfinec/nbi022",
968 + "title": "Econometrics of Testing for Jumps in Financial Economics Using Bipower Variation",
969 + "journal": "Journal of Financial Econometrics",
970 + "year": 2005,
971 + "print_year": null,
972 + "volume": "4",
973 + "issue": "1",
974 + "page": "1-30",
975 + "authors": [
976 + [
977 + "O. E.",
978 + "Barndorff-Nielsen"
979 + ]
980 + ]
981 + },
982 + {
983 + "doi": "10.1093/jjfinec/nbh001",
984 + "title": "Power and Bipower Variation with Stochastic Volatility and Jumps",
985 + "journal": "Journal of Financial Econometrics",
986 + "year": 2004,
987 + "print_year": 2004,
988 + "volume": "2",
989 + "issue": "1",
990 + "page": "1-37",
991 + "authors": [
992 + [
993 + "O. E.",
994 + "Barndorff-Nielsen"
995 + ]
996 + ]
997 + },
998 + {
999 + "doi": "10.2139/ssrn.751984",
1000 + "title": "Variation, Jumps, Market Frictions and High Frequency Data in Financial Econometrics",
1001 + "journal": "",
1002 + "year": 2005,
1003 + "print_year": null,
1004 + "volume": null,
1005 + "issue": null,
1006 + "page": null,
1007 + "authors": [
1008 + [
1009 + "Ole E.",
1010 + "Barndorff-Nielsen"
1011 + ],
1012 + [
1013 + "Neil",
1014 + "Shephard"
1015 + ]
1016 + ]
1017 + }
1018 + ],
1019 + "johnson2013abrupt": [
1020 + {
1021 + "doi": "10.1038/srep02627",
1022 + "title": "Abrupt rise of new machine ecology beyond human response time",
1023 + "journal": "Scientific Reports",
1024 + "year": 2013,
1025 + "print_year": null,
1026 + "volume": "3",
1027 + "issue": "1",
1028 + "page": null,
1029 + "authors": [
1030 + [
1031 + "Neil",
1032 + "Johnson"
1033 + ],
1034 + [
1035 + "Guannan",
1036 + "Zhao"
1037 + ],
1038 + [
1039 + "Eric",
1040 + "Hunsader"
1041 + ],
1042 + [
1043 + "Hong",
1044 + "Qi"
1045 + ],
1046 + [
1047 + "Nicholas",
1048 + "Johnson"
1049 + ],
1050 + [
1051 + "Jing",
1052 + "Meng"
1053 + ],
1054 + [
1055 + "Brian",
1056 + "Tivnan"
1057 + ]
1058 + ]
1059 + },
1060 + {
1061 + "doi": "10.2139/ssrn.2003874",
1062 + "title": "Financial Black Swans Driven by Ultrafast Machine Ecology",
1063 + "journal": "",
1064 + "year": 2012,
1065 + "print_year": null,
1066 + "volume": null,
1067 + "issue": null,
1068 + "page": null,
1069 + "authors": [
1070 + [
1071 + "Neil",
1072 + "Johnson"
1073 + ],
1074 + [
1075 + "Guannan",
1076 + "Zhao"
1077 + ],
1078 + [
1079 + "Eric",
1080 + "Hunsader"
1081 + ],
1082 + [
1083 + "Jing",
1084 + "Meng"
1085 + ],
1086 + [
1087 + "Amith",
1088 + "Ravindar"
1089 + ],
1090 + [
1091 + "Spencer",
1092 + "Carran"
1093 + ],
1094 + [
1095 + "Brian",
1096 + "Tivnan"
1097 + ]
1098 + ]
1099 + },
1100 + {
1101 + "doi": "10.1093/acprof:oso/9780199790692.003.0010",
1102 + "title": "The Rise of Human Rights During the Johnson Years",
1103 + "journal": "Beyond the Cold War",
1104 + "year": 2014,
1105 + "print_year": 2014,
1106 + "volume": null,
1107 + "issue": null,
1108 + "page": "237-260",
1109 + "authors": [
1110 + [
1111 + "Sarah B.",
1112 + "Snyder"
1113 + ]
1114 + ]
1115 + }
1116 + ],
1117 + "dungey2018testing": [
1118 + {
1119 + "doi": "10.1016/j.jeconom.2017.09.002",
1120 + "title": "Testing for mutually exciting jumps and financial flights in high frequency data",
1121 + "journal": "Journal of Econometrics",
1122 + "year": 2018,
1123 + "print_year": 2018,
1124 + "volume": "202",
1125 + "issue": "1",
1126 + "page": "18-44",
1127 + "authors": [
1128 + [
1129 + "Mardi",
1130 + "Dungey"
1131 + ],
1132 + [
1133 + "Deniz",
1134 + "Erdemlioglu"
1135 + ],
1136 + [
1137 + "Marius",
1138 + "Matei"
1139 + ],
1140 + [
1141 + "Xiye",
1142 + "Yang"
1143 + ]
1144 + ]
1145 + },
1146 + {
1147 + "doi": "10.2139/ssrn.2787304",
1148 + "title": "Testing for Mutually Exciting Jumps and Financial Flights in High Frequency Data",
1149 + "journal": "",
1150 + "year": 2016,
1151 + "print_year": null,
1152 + "volume": null,
1153 + "issue": null,
1154 + "page": null,
1155 + "authors": [
1156 + [
1157 + "Mardi H.",
1158 + "Dungey"
1159 + ],
1160 + [
1161 + "Deniz",
1162 + "Erdemlioglu"
1163 + ],
1164 + [
1165 + "Marius",
1166 + "Matei"
1167 + ],
1168 + [
1169 + "Xiye",
1170 + "Yang"
1171 + ]
1172 + ]
1173 + },
1174 + {
1175 + "doi": "10.2307/j.ctt6wq07x.15",
1176 + "title": "Testing for Jumps",
1177 + "journal": "High-Frequency Financial Econometrics",
1178 + "year": 2014,
1179 + "print_year": 2014,
1180 + "volume": null,
1181 + "issue": null,
1182 + "page": "329-392",
1183 + "authors": []
1184 + }
1185 + ],
1186 + "hansen2006realized": [
1187 + {
1188 + "doi": "10.1198/073500106000000071",
1189 + "title": "Realized Variance and Market Microstructure Noise",
1190 + "journal": "Journal of Business &amp; Economic Statistics",
1191 + "year": 2006,
1192 + "print_year": 2006,
1193 + "volume": "24",
1194 + "issue": "2",
1195 + "page": "127-161",
1196 + "authors": [
1197 + [
1198 + "Peter R",
1199 + "Hansen"
1200 + ],
1201 + [
1202 + "Asger",
1203 + "Lunde"
1204 + ]
1205 + ]
1206 + },
1207 + {
1208 + "doi": "10.2139/ssrn.873397",
1209 + "title": "Rejoinder (to Comments on Realized Variance and Market Microstructure Noise)",
1210 + "journal": "",
1211 + "year": 2006,
1212 + "print_year": null,
1213 + "volume": null,
1214 + "issue": null,
1215 + "page": null,
1216 + "authors": [
1217 + [
1218 + "Peter Reinhard",
1219 + "Hansen"
1220 + ],
1221 + [
1222 + "Asger",
1223 + "Lunde"
1224 + ]
1225 + ]
1226 + },
1227 + {
1228 + "doi": "10.2139/ssrn.506542",
1229 + "title": "Realized Variance and Market Microstructure Noise",
1230 + "journal": "",
1231 + "year": 2004,
1232 + "print_year": null,
1233 + "volume": null,
1234 + "issue": null,
1235 + "page": null,
1236 + "authors": [
1237 + [
1238 + "Peter Reinhard",
1239 + "Hansen"
1240 + ],
1241 + [
1242 + "Asger",
1243 + "Lunde"
1244 + ]
1245 + ]
1246 + }
1247 + ],
1248 + "scholes1977estimating": [
1249 + {
1250 + "doi": "10.1016/0304-405x(77)90041-1",
1251 + "title": "Estimating betas from nonsynchronous data",
1252 + "journal": "Journal of Financial Economics",
1253 + "year": 1977,
1254 + "print_year": 1977,
1255 + "volume": "5",
1256 + "issue": "3",
1257 + "page": "309-327",
1258 + "authors": [
1259 + [
1260 + "Myron",
1261 + "Scholes"
1262 + ],
1263 + [
1264 + "Joseph",
1265 + "Williams"
1266 + ]
1267 + ]
1268 + },
1269 + {
1270 + "doi": "10.1080/09603100500359773",
1271 + "title": "A different approach to estimating betas of securities subject to thin trading and serial correlation",
1272 + "journal": "Applied Financial Economics",
1273 + "year": 2005,
1274 + "print_year": 2005,
1275 + "volume": "15",
1276 + "issue": "16",
1277 + "page": "1145-1152",
1278 + "authors": [
1279 + [
1280 + "Peijie",
1281 + "Wang"
1282 + ],
1283 + [
1284 + "Trefor",
1285 + "Jones"
1286 + ]
1287 + ]
1288 + },
1289 + {
1290 + "doi": "10.2139/ssrn.2393057",
1291 + "title": "From Risk-Based Betas to Value-at-Risk: Estimating Monetary Risk with Market Data",
1292 + "journal": "",
1293 + "year": 2014,
1294 + "print_year": null,
1295 + "volume": null,
1296 + "issue": null,
1297 + "page": null,
1298 + "authors": [
1299 + [
1300 + "Alessio",
1301 + "Rombolotti Axelrod"
1302 + ]
1303 + ]
1304 + }
1305 + ],
1306 + "bessembinder2016liquidity": [
1307 + {
1308 + "doi": "10.2139/ssrn.2026802",
1309 + "title": "Liquidity, Resiliency and Market Quality Around Predictable Trades: Theory and Evidence",
1310 + "journal": "",
1311 + "year": 2012,
1312 + "print_year": null,
1313 + "volume": null,
1314 + "issue": null,
1315 + "page": null,
1316 + "authors": [
1317 + [
1318 + "Hendrik (Hank)",
1319 + "Bessembinder"
1320 + ],
1321 + [
1322 + "Allen",
1323 + "Carrion"
1324 + ],
1325 + [
1326 + "Laura A.",
1327 + "Tuttle"
1328 + ],
1329 + [
1330 + "Kumar",
1331 + "Venkataraman"
1332 + ]
1333 + ]
1334 + },
1335 + {
1336 + "doi": "10.1016/j.jfineco.2016.02.011",
1337 + "title": "Liquidity, resiliency and market quality around predictable trades: Theory and evidence",
1338 + "journal": "Journal of Financial Economics",
1339 + "year": 2016,
1340 + "print_year": 2016,
1341 + "volume": "121",
1342 + "issue": "1",
1343 + "page": "142-166",
1344 + "authors": [
1345 + [
1346 + "Hendrik",
1347 + "Bessembinder"
1348 + ],
1349 + [
1350 + "Allen",
1351 + "Carrion"
1352 + ],
1353 + [
1354 + "Laura",
1355 + "Tuttle"
1356 + ],
1357 + [
1358 + "Kumar",
1359 + "Venkataraman"
1360 + ]
1361 + ]
1362 + },
1363 + {
1364 + "doi": "10.1016/j.jfineco.2005.10.002",
1365 + "title": "Market transparency, liquidity externalities, and institutional trading costs in corporate bonds\u2606",
1366 + "journal": "Journal of Financial Economics",
1367 + "year": 2006,
1368 + "print_year": 2006,
1369 + "volume": "82",
1370 + "issue": "2",
1371 + "page": "251-288",
1372 + "authors": [
1373 + [
1374 + "H",
1375 + "BESSEMBINDER"
1376 + ],
1377 + [
1378 + "W",
1379 + "MAXWELL"
1380 + ],
1381 + [
1382 + "K",
1383 + "VENKATARAMAN"
1384 + ]
1385 + ]
1386 + }
1387 + ],
1388 + "menkveld2017speed": [
1389 + {
1390 + "doi": "10.1093/rfs/hhx006",
1391 + "title": "Need for Speed? Exchange Latency and Liquidity",
1392 + "journal": "The Review of Financial Studies",
1393 + "year": 2017,
1394 + "print_year": 2017,
1395 + "volume": "30",
1396 + "issue": "4",
1397 + "page": "1188-1228",
1398 + "authors": [
1399 + [
1400 + "Albert J.",
1401 + "Menkveld"
1402 + ],
1403 + [
1404 + "Marius A.",
1405 + "Zoican"
1406 + ]
1407 + ]
1408 + },
1409 + {
1410 + "doi": "10.2139/ssrn.2442690",
1411 + "title": "Need for Speed? Exchange Latency and Liquidity",
1412 + "journal": "",
1413 + "year": 2014,
1414 + "print_year": null,
1415 + "volume": null,
1416 + "issue": null,
1417 + "page": null,
1418 + "authors": [
1419 + [
1420 + "Albert J.",
1421 + "Menkveld"
1422 + ],
1423 + [
1424 + "Marius",
1425 + "Zoican"
1426 + ]
1427 + ]
1428 + },
1429 + {
1430 + "doi": "10.2139/ssrn.2304337",
1431 + "title": "Need for Speed? Low Latency Trading and Adverse Selection",
1432 + "journal": "SSRN Electronic Journal",
1433 + "year": 2013,
1434 + "print_year": null,
1435 + "volume": null,
1436 + "issue": null,
1437 + "page": null,
1438 + "authors": [
1439 + [
1440 + "Albert J.",
1441 + "Menkveld"
1442 + ],
1443 + [
1444 + "Marius Andrei",
1445 + "Zoican"
1446 + ]
1447 + ]
1448 + }
1449 + ],
1450 + "harris1991stock": [
1451 + {
1452 + "doi": "10.1093/rfs/4.3.389",
1453 + "title": "Stock Price Clustering and Discreteness",
1454 + "journal": "Review of Financial Studies",
1455 + "year": 1991,
1456 + "print_year": 1991,
1457 + "volume": "4",
1458 + "issue": "3",
1459 + "page": "389-415",
1460 + "authors": [
1461 + [
1462 + "Lawrence",
1463 + "Harris"
1464 + ]
1465 + ]
1466 + },
1467 + {
1468 + "doi": "10.1108/jfrc-07-2012-0025",
1469 + "title": "Stock price discreteness and clustering: decimals and ordered probit model",
1470 + "journal": "Journal of Financial Regulation and Compliance",
1471 + "year": 2014,
1472 + "print_year": 2014,
1473 + "volume": "22",
1474 + "issue": "1",
1475 + "page": "49-60",
1476 + "authors": [
1477 + [
1478 + "Haksoon",
1479 + "Kim"
1480 + ]
1481 + ]
1482 + },
1483 + {
1484 + "doi": "10.1111/1540-6288.00015",
1485 + "title": "Does Price Discreteness Affect the Increase in Return Volatility Following Stock Splits?",
1486 + "journal": "Financial Review",
1487 + "year": 2002,
1488 + "print_year": 2002,
1489 + "volume": "37",
1490 + "issue": "2",
1491 + "page": "281-293",
1492 + "authors": [
1493 + [
1494 + "Dan W.",
1495 + "French"
1496 + ],
1497 + [
1498 + "Taylor W.",
1499 + "Foster, III"
1500 + ]
1501 + ]
1502 + }
1503 + ]
1504 +}
\ No newline at end of file
added outputs/diagnostics/revision_checks.json +323 −0
@@ -0,0 +1,323 @@
1 +{
2 + "params": {
3 + "bootstrap": 500,
4 + "seed": 20260713
5 + },
6 + "pre_event": {
7 + "breadth": {
8 + "real": 0.1044142855242956,
9 + "placebo": 0.10038894990969427,
10 + "diff": 0.004025335614601333,
11 + "se": 0.00011777498900422227,
12 + "n": 1827891
13 + },
14 + "intensity": {
15 + "real": 1.322621532897214,
16 + "placebo": 1.2488441907734187,
17 + "diff": 0.07377734212379528,
18 + "se": 0.011512188572763942,
19 + "n": 1827891
20 + },
21 + "breadth_by_decade": [
22 + {
23 + "dec": "1992\u20131999",
24 + "diff": 0.006942451470193442,
25 + "se": 0.0004457996794364938,
26 + "n": 149458
27 + },
28 + {
29 + "dec": "2000\u20132009",
30 + "diff": 0.0034576259578378565,
31 + "se": 0.00020804296105244357,
32 + "n": 635269
33 + },
34 + {
35 + "dec": "2010\u20132019",
36 + "diff": 0.004423961736415502,
37 + "se": 0.00022466851734241028,
38 + "n": 602787
39 + },
40 + {
41 + "dec": "2020\u20132026",
42 + "diff": 0.003586281025321686,
43 + "se": 0.00022094232486162134,
44 + "n": 440377
45 + }
46 + ]
47 + },
48 + "base_rates": {
49 + "placebo_breadth_5": 0.1005985025174963,
50 + "placebo_breadth_30": 0.10279938779926524,
51 + "nelig_5_mean": 427.389714684811,
52 + "nelig_5_by_decade": {
53 + "1992\u20131999": 236.9354154154154,
54 + "2000\u20132009": 436.0374996079415,
55 + "2010\u20132019": 459.73702756182576,
56 + "2020\u20132026": 435.2988952717632
57 + },
58 + "placebo_nbh_5": 0.10040448383793098,
59 + "placebo_non_5": 0.10061434323466838,
60 + "event_days": 7282,
61 + "match_rate": 0.9523690886030078,
62 + "match_rate_by_decade": {
63 + "1992\u20131999": 0.7144348681935466,
64 + "2000\u20132009": 0.9552945345136264,
65 + "2010\u20132019": 0.9835023225123639,
66 + "2020\u20132026": 0.9854602398458134
67 + },
68 + "placebo_cum_15": 0.5769076280210454,
69 + "events_per_stock_week": 2.5465311727547375
70 + },
71 + "network_components": [
72 + {
73 + "scope": "all",
74 + "real_nbh_5": 0.11025821988115234,
75 + "real_nbh_5_se": 0.0005984599392453769,
76 + "real_non_5": 0.10430004605173746,
77 + "real_non_5_se": 0.0004188202069912482,
78 + "real_nd_5": 0.00613062092947073,
79 + "real_nd_5_se": 0.00033786443723446776,
80 + "placebo_nbh_5": 0.10040448383793098,
81 + "placebo_nbh_5_se": 0.0005211929831721808,
82 + "placebo_non_5": 0.10061434323466838,
83 + "placebo_non_5_se": 0.00038888763591128726,
84 + "placebo_nd_5": -0.0002709459636011156,
85 + "placebo_nd_5_se": 0.00030992502825580477,
86 + "dd": 0.006401566893071845,
87 + "dd_se": 0.00018218537330683685,
88 + "n": 1791444
89 + },
90 + {
91 + "scope": "1992\u20131999",
92 + "real_nbh_5": 0.11914134078294751,
93 + "real_nbh_5_se": 0.001334751839569883,
94 + "real_non_5": 0.10062255827628108,
95 + "real_non_5_se": 0.0011229483473088522,
96 + "real_nd_5": 0.022834145978753984,
97 + "real_nd_5_se": 0.0007698530826802734,
98 + "placebo_nbh_5": 0.1120921476269281,
99 + "placebo_nbh_5_se": 0.0014294873320712174,
100 + "placebo_non_5": 0.09321229985194236,
101 + "placebo_non_5_se": 0.001092345807558833,
102 + "placebo_nd_5": 0.019999582707871204,
103 + "placebo_nd_5_se": 0.0007894983382933114,
104 + "dd": 0.0028345632708827798,
105 + "dd_se": 0.0006882618270008896,
106 + "n": 109456
107 + },
108 + {
109 + "scope": "2000\u20132009",
110 + "real_nbh_5": 0.10993442455493217,
111 + "real_nbh_5_se": 0.0009345041240376674,
112 + "real_non_5": 0.11097799391339332,
113 + "real_non_5_se": 0.0006947193303634253,
114 + "real_nd_5": -0.001068972834372065,
115 + "real_nd_5_se": 0.00042547893765374005,
116 + "placebo_nbh_5": 0.10360921613277599,
117 + "placebo_nbh_5_se": 0.0009319426654686399,
118 + "placebo_non_5": 0.10764273183353922,
119 + "placebo_non_5_se": 0.0006886690773890117,
120 + "placebo_nd_5": -0.004064248792205989,
121 + "placebo_nd_5_se": 0.0004124342741015663,
122 + "dd": 0.0029952759578339232,
123 + "dd_se": 0.0002220026820671454,
124 + "n": 637083
125 + },
126 + {
127 + "scope": "2010\u20132019",
128 + "real_nbh_5": 0.10596971241292055,
129 + "real_nbh_5_se": 0.0011468428574011884,
130 + "real_non_5": 0.10226736158398393,
131 + "real_non_5_se": 0.0007298931599418086,
132 + "real_nd_5": 0.003698233379651755,
133 + "real_nd_5_se": 0.0006192751660271143,
134 + "placebo_nbh_5": 0.09410943324816824,
135 + "placebo_nbh_5_se": 0.0010189133876417188,
136 + "placebo_non_5": 0.09846960091996103,
137 + "placebo_non_5_se": 0.0007534628868046854,
138 + "placebo_nd_5": -0.0043656039171047385,
139 + "placebo_nd_5_se": 0.0005197821406488786,
140 + "dd": 0.008063837296756492,
141 + "dd_se": 0.0003730359714515336,
142 + "n": 603846
143 + },
144 + {
145 + "scope": "2020\u20132026",
146 + "real_nbh_5": 0.11439272043420276,
147 + "real_nbh_5_se": 0.001077402412144463,
148 + "real_non_5": 0.09868070575959409,
149 + "real_non_5_se": 0.0008412092900813695,
150 + "real_nd_5": 0.015714883462292648,
151 + "real_nd_5_se": 0.0007243447608618305,
152 + "placebo_nbh_5": 0.10228049051967733,
153 + "placebo_nbh_5_se": 0.0009578483446484562,
154 + "placebo_non_5": 0.09552040484276911,
155 + "placebo_non_5_se": 0.0008451119260404312,
156 + "placebo_nd_5": 0.0067547194567862305,
157 + "placebo_nd_5_se": 0.0005692863416051286,
158 + "dd": 0.008960164005506416,
159 + "dd_se": 0.0004311744987005946,
160 + "n": 441059
161 + }
162 + ],
163 + "coarrival": {
164 + "share_by_class": {
165 + "1": 0.21657488416638626,
166 + "2": 0.2707357267966074,
167 + "3\u20135": 0.43572582593465,
168 + "6+": 0.07696356310235636
169 + },
170 + "mean_events_per_event_minute": 2.054302562472097,
171 + "by_class": [
172 + {
173 + "cls": "1",
174 + "share": 0.21657488416638626,
175 + "db5": 0.00017444129006761585,
176 + "db5_se": 0.00016708437180743002,
177 + "n": 370526,
178 + "db30": 0.0003687338601283238,
179 + "db30_se": 0.00018905398700369108,
180 + "nd5": 0.005101435019021069,
181 + "nd5_se": 0.00025467998104708114,
182 + "nd15": 0.004232825254034365,
183 + "nd15_se": 0.00025286472578320415,
184 + "nd30": 0.0038045125734262773,
185 + "nd30_se": 0.00026715329671247356
186 + },
187 + {
188 + "cls": "2",
189 + "share": 0.2707357267966074,
190 + "db5": -0.0002970198795526921,
191 + "db5_se": 0.00017981197308079895,
192 + "n": 456948,
193 + "db30": -0.00022914048562588832,
194 + "db30_se": 0.00019846868194938304,
195 + "nd5": 0.006416565099413776,
196 + "nd5_se": 0.0002666876646652032,
197 + "nd15": 0.005073931484807639,
198 + "nd15_se": 0.00027602175698956206,
199 + "nd30": 0.004592961771667786,
200 + "nd30_se": 0.00027334886014155284
201 + },
202 + {
203 + "cls": "3\u20135",
204 + "share": 0.43572582593465,
205 + "db5": -0.00030146086313703714,
206 + "db5_se": 0.000179221423261845,
207 + "n": 727221,
208 + "db30": -0.0008345373807661645,
209 + "db30_se": 0.0002129368314714125,
210 + "nd5": 0.006686571645159107,
211 + "nd5_se": 0.0002993102864863163,
212 + "nd15": 0.0061606309399916456,
213 + "nd15_se": 0.00028580141376364127,
214 + "nd30": 0.005348586321906338,
215 + "nd30_se": 0.00029065211456614517
216 + },
217 + {
218 + "cls": "6+",
219 + "share": 0.07696356310235636,
220 + "db5": 0.0009119802649880374,
221 + "db5_se": 0.0005194030051313687,
222 + "n": 132600,
223 + "db30": 0.00013889973755474736,
224 + "db30_se": 0.0005678918069153508,
225 + "nd5": 0.007648667727279081,
226 + "nd5_se": 0.0007398715889683407,
227 + "nd15": 0.006528915485999489,
228 + "nd15_se": 0.0006962744363414618,
229 + "nd30": 0.00652203935337772,
230 + "nd30_se": 0.0007414276962825133
231 + }
232 + ]
233 + },
234 + "stratified": [
235 + {
236 + "h": 5,
237 + "post_diff": 0.003962518659143363,
238 + "post_se": 0.00012759941036939346,
239 + "dd": -0.0001036189203295987,
240 + "dd_se": 8.656607730897141e-05,
241 + "strat": 0.0019483063646004454,
242 + "strat_se": 8.785967457659237e-05,
243 + "n": 1806236
244 + },
245 + {
246 + "h": 15,
247 + "post_diff": 0.003946249426570195,
248 + "post_se": 0.00013558403976982316,
249 + "dd": -0.00014561085499940484,
250 + "dd_se": 9.98330528109048e-05,
251 + "strat": 0.001959351896473462,
252 + "strat_se": 9.827938024577038e-05,
253 + "n": 1757825
254 + },
255 + {
256 + "h": 30,
257 + "post_diff": 0.0037925191817651666,
258 + "post_se": 0.00014211271923074995,
259 + "dd": -0.00031036958489829214,
260 + "dd_se": 0.00010740223661279141,
261 + "strat": 0.0018414136761319396,
262 + "strat_se": 0.00010402011404494201,
263 + "n": 1687295
264 + }
265 + ],
266 + "cluster": {
267 + "db5": {
268 + "est": -0.0001036189203295987,
269 + "se_day": 8.994608987514974e-05,
270 + "se_ticker": 9.553204320319426e-05
271 + },
272 + "db30": {
273 + "est": -0.00031036958489829214,
274 + "se_day": 0.00010642681928722456,
275 + "se_ticker": 9.328144186397918e-05
276 + },
277 + "nd5": {
278 + "est": 0.006401566893071845,
279 + "se_day": 0.0001888274544791473,
280 + "se_ticker": 0.0004716259992487461
281 + }
282 + },
283 + "signed": [
284 + {
285 + "h": 1,
286 + "diff": 0.01513321286435076,
287 + "se": 0.010490338678985668,
288 + "n": 1854305
289 + },
290 + {
291 + "h": 5,
292 + "diff": 0.0009566661584060281,
293 + "se": 0.009195181940681063,
294 + "n": 1832710
295 + },
296 + {
297 + "h": 15,
298 + "diff": 0.012547558324088218,
299 + "se": 0.008114078638091559,
300 + "n": 1784278
301 + },
302 + {
303 + "h": 30,
304 + "diff": -0.006570153994793571,
305 + "se": 0.008922785846408422,
306 + "n": 1713759
307 + }
308 + ],
309 + "network_by_sign": [
310 + {
311 + "sign": "negative",
312 + "dd": 0.006554680577960512,
313 + "se": 0.00021941988291517274,
314 + "n": 886380
315 + },
316 + {
317 + "sign": "positive",
318 + "dd": 0.006251665115495484,
319 + "se": 0.00023286430887126946,
320 + "n": 905064
321 + }
322 + ]
323 +}
\ No newline at end of file
modified outputs/release/author_statement.pdf +0 −0

Binary file not shown.

modified outputs/release/cover_letter.pdf +0 −0

Binary file not shown.

modified outputs/release/cover_letter.tex +10 −8
@@ -13,16 +13,18 @@ The paper measures how idiosyncratic intraday shocks travel across the U.S.
13 13 equity cross-section using 7.7 billion one-minute bars spanning thirty-four
14 14 years, including delisted securities. Its three findings are, we believe, of
15 15 direct interest to the journal's readership: (i) isolated six-sigma
16 −one-minute shocks show no market-wide propagation relative to 1.8 million
17 −matched within-stock placebos, in any decade; (ii) propagation is instead
18 −concentrated in the shocked stock's lagged-correlation neighborhood, and this
19 −network channel roughly tripled between the floor-trading and algorithmic
20 −eras; and (iii) tail synchronization migrated to market-wide co-jump minutes,
21 −whose severity spikes in 2008 and 2020 while ordinary extreme minutes became
22 −an order of magnitude rarer.
16 +one-minute shocks raise market-wide tail activity by at most a few percent of
17 +its base rate relative to 1.8 million matched within-stock placebos, in any
18 +decade; (ii) propagation is instead concentrated in the shocked stock's
19 +lagged-correlation neighborhood, and this network channel roughly tripled
20 +between the floor-trading and algorithmic eras; and (iii) common information
21 +is impounded within the minute, ordinary extreme minutes became an order of
22 +magnitude rarer after decimalization, and severe co-jump minutes are roughly
23 +ten times more frequent in 2008 and 2020 than in calm years. The manuscript
24 +reports the design diagnostics that bound each of these statements.
23 25
24 26 The analysis is fully reproducible: the data snapshot is pinned by checksum,
25 −every number in the manuscript is generated programmatically from
27 +every headline number in the manuscript is generated programmatically from
26 28 machine-readable results, and replication code accompanies the submission.
27 29 The manuscript is not under consideration elsewhere, and there are no
28 30 competing interests to declare.
modified outputs/release/highlights.pdf +0 −0

Binary file not shown.

modified outputs/release/manuscript_anonymous.pdf +0 −0

Binary file not shown.

modified outputs/release/manuscript_irfa.pdf +0 −0

Binary file not shown.

modified outputs/release/online_appendix.pdf +0 −0

Binary file not shown.

modified outputs/release/submission_checklist.md +19 −13
@@ -1,31 +1,37 @@
1 −# IRFA Submission Checklist — 2026-07-13
1 +# IRFA Submission Checklist — regenerated 2026-10-17 (manuscript v1.1)
2 2
3 3 ## Manuscript
4 −- [x] `manuscript_irfa.pdf` — elsarticle (preprint, 12pt, authoryear), 41 pages,
5 − compiles clean, 0 unresolved references, elsarticle-harv bibliography
4 +- [x] `manuscript_irfa.pdf` — elsarticle (preprint, 12pt, authoryear), 50 pages,
5 + compiles clean, 0 unresolved references, elsarticle-harv bibliography (v1.1, 2026-10-17)
6 6 - [x] Title, author + affiliation + email, abstract (<200 words), keywords, JEL
7 7 - [x] Declaration of competing interest (in manuscript)
8 8 - [x] Data availability statement (in manuscript; source: Finnhub.io aggregation,
9 − pinned public snapshot for reproducibility)
9 + pinned public snapshot, licence unspecified → minute data not redistributed)
10 10 - [x] `highlights.pdf` — 5 bullets, ≤85 characters each
11 11 - [x] `cover_letter.pdf`
12 12 - [x] `manuscript_anonymous.pdf` — author identity/CRediT/acknowledgments stripped
13 13 - [x] `title_page.pdf` — full author details (Professeur, Dépt. sciences administratives,
14 14 UQO, Pavillon Alexandre-Taché, J9A 1L8) + declarations + CRediT statement
15 −- [x] `online_appendix.pdf` (skeleton; extended audit + full robustness to be
16 − finalized before submission)
15 +- [x] `online_appendix.pdf` (extended audit, pilot audit, descriptive figures,
16 + year-by-year evolution, list of analyses not executed)
17 17
18 18 ## Science
19 −- [x] 110 DOI-verified references (docs/LITERATURE_MAP.md)
19 +- [x] 124 DOI-verified references (Crossref, 2026-10-17; docs/LITERATURE_MAP.md)
20 20 - [x] All numbers macro-generated from machine-readable registries (no hand-typed
21 21 results); regenerate with scripts/15_build_results_macros.py
22 −- [x] Robustness battery (Table VIII): 11 designs incl. placebo-network
23 − falsification, month-cluster inference, post-decimalization
22 +- [x] Robustness battery (Table 6): 11 designs incl. placebo-network
23 + falsification, month-cluster inference, post-decimalization; design
24 + diagnostics (Table 3) and network components (Table 5) added 2026-10-17
24 25 - [x] Figures: vector PDF, CVD-safe palette + line styles, source data CSVs
25 26
26 27 ## Remaining before actual submission (author actions)
27 −- [x] Author details complete (title page + frontmatter); add acknowledgments if any
28 −- [ ] Finalize online appendix content (extended tables)
28 +- [x] Author details complete (title page + frontmatter)
29 +- [ ] Acknowledgments and funding statement (currently "All errors are the author's own.")
30 +- [ ] Repeat the novelty scan (2018–2026) and update §2.11 search date
29 31 - [ ] Check current IRFA Guide for Authors for any format updates (record date
30 − in docs/DECISION_LOG.md)
31 −- [ ] Clarify licensing note for the data aggregation (DATA_AUDIT §1)
32 + in docs/DECISION_LOG.md); consider moving Tables 3/5 or descriptive figures
33 + to the online appendix if a page limit applies (manuscript is 50 pp.)
34 +- [ ] Licensing note: dataset card has no licence; manuscript states non-redistribution
35 + (DATA_AUDIT §1) — confirm with the data provider if required by the journal
36 +- [ ] Strongly recommended before submission: neighbour sets on coarser-sampled
37 + correlations (REVIEW-REPORT §8, item 1) to defend the secular rise of the network channel
modified outputs/release/title_page.pdf +0 −0

Binary file not shown.

modified outputs/release/title_page.tex +1 −1
@@ -21,7 +21,7 @@ Pavillon Alexandre-Tach\'e\\
21 21 283, boulevard Alexandre-Tach\'e\\
22 22 Gatineau (Qu\'ebec), Canada, J9A 1L8\\[2.5em]
23 23
24 −\today
24 +October 2026
25 25 \end{center}
26 26
27 27 \vfill
modified outputs/tables/tabP1_pilot_summary.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/02_pilot_figures_tables.py — do not edit numbers by hand.
2 −\begin{table}[t]
2 +\begin{table}[htbp]
3 3 \centering
4 4 \begin{adjustbox}{max width=\textwidth}
5 5 \begin{threeparttable}
added outputs/tables/table10_network_components.csv +6 −0
@@ -0,0 +1,6 @@
1 +scope,real_nbh_5,real_nbh_5_se,real_non_5,real_non_5_se,real_nd_5,real_nd_5_se,placebo_nbh_5,placebo_nbh_5_se,placebo_non_5,placebo_non_5_se,placebo_nd_5,placebo_nd_5_se,dd,dd_se,n
2 +all,0.11025821988115234,0.0005984599392453769,0.10430004605173746,0.0004188202069912482,0.00613062092947073,0.00033786443723446776,0.10040448383793098,0.0005211929831721808,0.10061434323466838,0.00038888763591128726,-0.0002709459636011156,0.00030992502825580477,0.006401566893071845,0.00018218537330683685,1791444
3 +1992–1999,0.11914134078294751,0.001334751839569883,0.10062255827628108,0.0011229483473088522,0.022834145978753984,0.0007698530826802734,0.1120921476269281,0.0014294873320712174,0.09321229985194236,0.001092345807558833,0.019999582707871204,0.0007894983382933114,0.0028345632708827798,0.0006882618270008896,109456
4 +2000–2009,0.10993442455493217,0.0009345041240376674,0.11097799391339332,0.0006947193303634253,-0.001068972834372065,0.00042547893765374005,0.10360921613277599,0.0009319426654686399,0.10764273183353922,0.0006886690773890117,-0.004064248792205989,0.0004124342741015663,0.0029952759578339232,0.0002220026820671454,637083
5 +2010–2019,0.10596971241292055,0.0011468428574011884,0.10226736158398393,0.0007298931599418086,0.003698233379651755,0.0006192751660271143,0.09410943324816824,0.0010189133876417188,0.09846960091996103,0.0007534628868046854,-0.0043656039171047385,0.0005197821406488786,0.008063837296756492,0.0003730359714515336,603846
6 +2020–2026,0.11439272043420276,0.001077402412144463,0.09868070575959409,0.0008412092900813695,0.015714883462292648,0.0007243447608618305,0.10228049051967733,0.0009578483446484562,0.09552040484276911,0.0008451119260404312,0.0067547194567862305,0.0005692863416051286,0.008960164005506416,0.0004311744987005946,441059
added outputs/tables/table10_network_components.tex +23 −0
@@ -0,0 +1,23 @@
1 +% Auto-generated by scripts/17_revision_checks.py — do not edit numbers.
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}\begin{threeparttable}
3 +\caption{Components of the network differential at $h=5$, by decade}
4 +\label{tab:network_components}
5 +\small\begin{tabular}{lrrrrrrr}
6 +\toprule
7 + & \multicolumn{3}{c}{Real events} & \multicolumn{3}{c}{Placebo events} & \\
8 +\cmidrule(lr){2-4}\cmidrule(lr){5-7}
9 +Sample & Neighbors & Others & Diff. & Neighbors & Others & Diff. & Real $-$ placebo \\ \midrule
10 +Full sample & 1,103 & 1,043 & 61.3 (3.4) & 1,004 & 1,006 & -2.7 (3.1) & 64.0 (1.8) \\
11 +1992–1999 & 1,191 & 1,006 & 228.3 (7.7) & 1,121 & 932 & 200.0 (7.9) & 28.3 (6.9) \\
12 +2000–2009 & 1,099 & 1,110 & -10.7 (4.3) & 1,036 & 1,076 & -40.6 (4.1) & 30.0 (2.2) \\
13 +2010–2019 & 1,060 & 1,023 & 37.0 (6.2) & 941 & 985 & -43.7 (5.2) & 80.6 (3.7) \\
14 +2020–2026 & 1,144 & 987 & 157.1 (7.2) & 1,023 & 955 & 67.5 (5.7) & 89.6 (4.3) \\
15 +\bottomrule\end{tabular}
16 +\begin{tablenotes}[flushleft]\footnotesize
17 +\item Notes: Share of receivers with $|z|>2$ five minutes after the event minute, among
18 +the source's twenty lagged-correlation neighbors and among all other receivers, for real
19 +events and for their matched placebos ($\times10^{-4}$). `Diff.' is neighbors minus
20 +others within an arm; the last column is the network differential of
21 +Table~\ref{tab:network_mechanism} (real minus placebo). Day-block bootstrap SEs (500
22 +replications) in parentheses.
23 +\end{tablenotes}\end{threeparttable}\end{adjustbox}\end{table}
\ No newline at end of file
modified outputs/tables/table1_sample_construction.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Sample construction by decade}\label{tab:sample_construction}
5 5 \small\begin{tabular}{lrrrr}
modified outputs/tables/table2_summary_by_decade.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Summary statistics by decade, top-1000 universe}
5 5 \label{tab:summary_by_decade}
modified outputs/tables/table4_baseline_propagation.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/12_stage_f_baseline.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Baseline abnormal propagation after idiosyncratic shocks}
5 5 \label{tab:baseline_propagation}
modified outputs/tables/table7_network.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/14_stage_f2_results.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Network mechanism: neighbor minus non-neighbor abnormal response}
5 5 \label{tab:network_mechanism}
modified outputs/tables/table8_robustness.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/16_robustness.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Robustness of the market-wide null and the network channel}
5 5 \label{tab:robustness}
added outputs/tables/table9_design_checks.csv +13 −0
@@ -0,0 +1,13 @@
1 +panel,row,real,placebo,diff,se,share,db5,db5_se,db30,db30_se,nd5,nd5_se,post,post_se,dd,dd_se,strat,strat_se,est,se_day,se_ticker
2 +A,Breadth,0.1044142855242956,0.10038894990969427,0.004025335614601333,0.00011777498900422227,,,,,,,,,,,,,,,,
3 +A,Intensity,1.322621532897214,1.2488441907734187,0.07377734212379528,0.011512188572763942,,,,,,,,,,,,,,,,
4 +B,1,,,,,0.21657488416638626,0.00017444129006761585,0.00016708437180743002,0.0003687338601283238,0.00018905398700369108,0.005101435019021069,0.00025467998104708114,,,,,,,,,
5 +B,2,,,,,0.2707357267966074,-0.0002970198795526921,0.00017981197308079895,-0.00022914048562588832,0.00019846868194938304,0.006416565099413776,0.0002666876646652032,,,,,,,,,
6 +B,3–5,,,,,0.43572582593465,-0.00030146086313703714,0.000179221423261845,-0.0008345373807661645,0.0002129368314714125,0.006686571645159107,0.0002993102864863163,,,,,,,,,
7 +B,6+,,,,,0.07696356310235636,0.0009119802649880374,0.0005194030051313687,0.00013889973755474736,0.0005678918069153508,0.007648667727279081,0.0007398715889683407,,,,,,,,,
8 +C,5,,,,,,,,,,,,0.003962518659143363,0.00012759941036939346,-0.0001036189203295987,0.00008656607730897141,0.0019483063646004454,0.00008785967457659237,,,
9 +C,15,,,,,,,,,,,,0.003946249426570195,0.00013558403976982316,-0.00014561085499940484,0.0000998330528109048,0.001959351896473462,0.00009827938024577038,,,
10 +C,30,,,,,,,,,,,,0.0037925191817651666,0.00014211271923074995,-0.00031036958489829214,0.00010740223661279141,0.0018414136761319396,0.00010402011404494201,,,
11 +D,db5,,,,,,,,,,,,,,,,,,-0.0001036189203295987,0.00008994608987514974,0.00009553204320319426
12 +D,db30,,,,,,,,,,,,,,,,,,-0.00031036958489829214,0.00010642681928722456,0.00009328144186397918
13 +D,nd5,,,,,,,,,,,,,,,,,,0.006401566893071845,0.0001888274544791473,0.0004716259992487461
added outputs/tables/table9_design_checks.tex +47 −0
@@ -0,0 +1,47 @@
1 +% Auto-generated by scripts/17_revision_checks.py — do not edit numbers.
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}\begin{threeparttable}
3 +\caption{Design diagnostics: pre-event levels, co-arrivals, mean reversion, and clustering}
4 +\label{tab:design_checks}
5 +\small\begin{tabular}{lrrr}
6 +\toprule
7 +\multicolumn{4}{l}{\emph{Panel A: Pre-event levels at $t-5$ ($\times10^{-4}$)}} \\
8 + & Real & Placebo & Real $-$ placebo \\ \midrule
9 +Breadth & 1,044 & 1,004 & 40.25 (1.18) \\
10 +Intensity & 13,226 & 12,488 & 737.77 (115.12) \\
11 +\addlinespace
12 +\multicolumn{4}{l}{\emph{Panel B: By number of source events in the same minute}} \\
13 +Events in minute (share of events) & $\Delta\Delta$Breadth$_{5}$ & $\Delta\Delta$Breadth$_{30}$ & Network diff$_{5}$ \\ \midrule
14 +1 (21.7\%) & 1.74 (1.67) & 3.69 (1.89) & 51.01 (2.55) \\
15 +2 (27.1\%) & -2.97 (1.80) & -2.29 (1.98) & 64.17 (2.67) \\
16 +3–5 (43.6\%) & -3.01 (1.79) & -8.35 (2.13) & 66.87 (2.99) \\
17 +6+ (7.7\%) & 9.12 (5.19) & 1.39 (5.68) & 76.49 (7.40) \\
18 +\addlinespace
19 +\multicolumn{4}{l}{\emph{Panel C: Market-wide breadth, alternative estimands ($\times10^{-4}$)}} \\
20 +Horizon & Post level, real $-$ placebo & $\Delta\Delta$ (baseline) & Pre-level stratified \\ \midrule
21 +$h=5$ & 39.63 (1.28) & -1.04 (0.87) & 19.48 (0.88) \\
22 +$h=15$ & 39.46 (1.36) & -1.46 (1.00) & 19.59 (0.98) \\
23 +$h=30$ & 37.93 (1.42) & -3.10 (1.07) & 18.41 (1.04) \\
24 +\addlinespace
25 +\multicolumn{4}{l}{\emph{Panel D: Bootstrap standard errors by clustering unit ($\times10^{-4}$)}} \\
26 +Statistic & Estimate & Day clusters & Ticker clusters \\ \midrule
27 +$\Delta\Delta$Breadth$_{5}$ & -1.04 & (0.90) & (0.96) \\
28 +$\Delta\Delta$Breadth$_{30}$ & -3.10 & (1.06) & (0.93) \\
29 +Network diff$_{5}$ & 64.02 & (1.89) & (4.72) \\
30 +\addlinespace
31 +\multicolumn{4}{l}{\emph{Panel E: Signed response, real $-$ placebo ($\times10^{-4}$)}} \\
32 +Horizon & $h=1$ & $h=5$ & $h=30$ \\ \midrule
33 +sign$(e_{i,t})\times$ mean $z_{j,t+h}$ & 151.33 (104.90) & 9.57 (91.95) & -65.70 (89.23) \\
34 +\bottomrule\end{tabular}
35 +\begin{tablenotes}[flushleft]\footnotesize
36 +\item Notes: All statistics from the baseline event files (isolated $|z|\ge6$ source
37 +shocks, lagged top-500 universe, 1992--2026; matched within-stock placebos).
38 +Day-block bootstrap standard errors (500 replications) in parentheses unless stated.
39 +Panel A: levels of the outcomes five minutes before the event minute. Panel B: events
40 +classified by the number of source events detected in the same minute (the isolation
41 +filters cap the share of simultaneous $|z|>4$ names at 2\%, not the number of $|z|\ge6$
42 +events); each placebo inherits the class of its real event. Panel C: `post level' is
43 +the real-minus-placebo difference in breadth at $t+h$ without pre-differencing; `pre-level
44 +stratified' reweights the same difference within deciles of the pre-event breadth level.
45 +Panel D: standard errors from resampling trading days versus resampling source tickers.
46 +Panel E: cross-sectional mean standardized return of receivers signed by the source shock.
47 +\end{tablenotes}\end{threeparttable}\end{adjustbox}\end{table}
\ No newline at end of file
modified paper/bibliography/references.bib +192 −50
@@ -411,16 +411,6 @@
411 411 pages = {184--222},
412 412 doi = {10.1214/07-AOS568}
413 413 }
414 −@article{gu2020empirical,
415 − author = {Gu, Shihao and Kelly, Bryan and Xiu, Dacheng},
416 − title = {Empirical Asset Pricing via Machine Learning},
417 − journal = {Review of Financial Studies},
418 − year = {2020},
419 − volume = {33},
420 − number = {5},
421 − pages = {2223--2273},
422 − doi = {10.1093/rfs/hhaa009}
423 −}
424 414 @article{easley2012flow,
425 415 author = {Easley, David and L{\'o}pez de Prado, Marcos M. and O'Hara, Maureen},
426 416 title = {Flow Toxicity and Liquidity in a High-Frequency World},
@@ -450,16 +440,6 @@
450 440 pages = {291--298},
451 441 doi = {10.1080/01621459.1979.10482508}
452 442 }
453 −@article{jorda2005estimation,
454 − author = {Jord{\`a}, {\`O}scar},
455 − title = {Estimation and Inference of Impulse Responses by Local Projections},
456 − journal = {American Economic Review},
457 − year = {2005},
458 − volume = {95},
459 − number = {1},
460 − pages = {161--182},
461 − doi = {10.1257/0002828053828518}
462 −}
463 443 % ---- Stream: contagion theory, fire sales, ownership networks ----
464 444 % Verified via OpenAlex 2026-07-13.
465 445 @article{kyle2001contagion,
@@ -845,16 +825,6 @@
845 825 pages = {406--423},
846 826 doi = {10.1016/j.jfineco.2007.04.005}
847 827 }
848 −@article{hasbrouck1991measuring,
849 − author = {Hasbrouck, Joel},
850 − title = {Measuring the Information Content of Stock Trades},
851 − journal = {Journal of Finance},
852 − year = {1991},
853 − volume = {46},
854 − number = {1},
855 − pages = {179--207},
856 − doi = {10.1111/j.1540-6261.1991.tb03749.x}
857 −}
858 828 @article{boudt2014intraday,
859 829 author = {Boudt, Kris and Petitjean, Mika{\"e}l},
860 830 title = {Intraday Liquidity Dynamics and News Releases around Price Jumps: Evidence from the {DJIA} Stocks},
@@ -906,26 +876,6 @@
906 876 pages = {287--303},
907 877 doi = {10.1080/01621459.2017.1401542}
908 878 }
909 −@article{sirignano2019universal,
910 − author = {Sirignano, Justin and Cont, Rama},
911 − title = {Universal features of price formation in financial markets: perspectives from deep learning},
912 − journal = {Quantitative Finance},
913 − year = {2019},
914 − volume = {19},
915 − number = {9},
916 − pages = {1449--1459},
917 − doi = {10.1080/14697688.2019.1622295}
918 −}
919 −@article{easley2021microstructure,
920 − author = {Easley, David and L{\'o}pez de Prado, Marcos and O'Hara, Maureen and Zhang, Zhibai},
921 − title = {Microstructure in the Machine Age},
922 − journal = {Review of Financial Studies},
923 − year = {2021},
924 − volume = {34},
925 − number = {7},
926 − pages = {3316--3363},
927 − doi = {10.1093/rfs/hhaa078}
928 −}
929 879 @article{baltussen2019indexing,
930 880 author = {Baltussen, Guido and van Bekkum, Sjoerd and Da, Zhi},
931 881 title = {Indexing and stock market serial dependence around the world},
@@ -1118,3 +1068,195 @@
1118 1068 pages = {243--263},
1119 1069 doi = {10.1016/j.jfineco.2011.03.008}
1120 1070 }
1071 +
1072 +% ---- Added during the editorial-board revision of 2026-10-17 ----
1073 +% Each entry verified against the Crossref record of its DOI on 2026-10-17
1074 +% (outputs/diagnostics/bibliography/crossref_new_candidates.json).
1075 +@article{toth2009epps,
1076 + author = {T{\'o}th, Bence and Kert{\'e}sz, J{\'a}nos},
1077 + title = {The {E}pps effect revisited},
1078 + journal = {Quantitative Finance},
1079 + year = {2009},
1080 + volume = {9},
1081 + number = {7},
1082 + pages = {793--802},
1083 + doi = {10.1080/14697680802595668}
1084 +}
1085 +@article{reno2003closer,
1086 + author = {Ren{\`o}, Roberto},
1087 + title = {A closer look at the {E}pps effect},
1088 + journal = {International Journal of Theoretical and Applied Finance},
1089 + year = {2003},
1090 + volume = {6},
1091 + number = {1},
1092 + pages = {87--102},
1093 + doi = {10.1142/S0219024903001839}
1094 +}
1095 +@article{hayashi2005covariance,
1096 + author = {Hayashi, Takaki and Yoshida, Nakahiro},
1097 + title = {On covariance estimation of non-synchronously observed diffusion processes},
1098 + journal = {Bernoulli},
1099 + year = {2005},
1100 + volume = {11},
1101 + number = {2},
1102 + pages = {359--379},
1103 + doi = {10.3150/bj/1116340299}
1104 +}
1105 +@article{zhang2011estimating,
1106 + author = {Zhang, Lan},
1107 + title = {Estimating covariation: {E}pps effect, microstructure noise},
1108 + journal = {Journal of Econometrics},
1109 + year = {2011},
1110 + volume = {160},
1111 + number = {1},
1112 + pages = {33--47},
1113 + doi = {10.1016/j.jeconom.2010.03.012}
1114 +}
1115 +@article{jacod2009testing,
1116 + author = {Jacod, Jean and Todorov, Viktor},
1117 + title = {Testing for common arrivals of jumps for discretely observed multidimensional processes},
1118 + journal = {Annals of Statistics},
1119 + year = {2009},
1120 + volume = {37},
1121 + number = {4},
1122 + pages = {1792--1838},
1123 + doi = {10.1214/08-AOS624}
1124 +}
1125 +@article{gilder2014cojumps,
1126 + author = {Gilder, Dudley and Shackleton, Mark B. and Taylor, Stephen J.},
1127 + title = {Cojumps in stock prices: Empirical evidence},
1128 + journal = {Journal of Banking \& Finance},
1129 + year = {2014},
1130 + volume = {40},
1131 + pages = {443--459},
1132 + doi = {10.1016/j.jbankfin.2013.04.025}
1133 +}
1134 +@article{bollerslev2018volume,
1135 + author = {Bollerslev, Tim and Li, Jia and Xue, Yuan},
1136 + title = {Volume, Volatility, and Public News Announcements},
1137 + journal = {Review of Economic Studies},
1138 + year = {2018},
1139 + volume = {85},
1140 + number = {4},
1141 + pages = {2005--2041},
1142 + doi = {10.1093/restud/rdy003}
1143 +}
1144 +@article{boudt2011robust,
1145 + author = {Boudt, Kris and Croux, Christophe and Laurent, S{\'e}bastien},
1146 + title = {Robust estimation of intraweek periodicity in volatility and jump detection},
1147 + journal = {Journal of Empirical Finance},
1148 + year = {2011},
1149 + volume = {18},
1150 + number = {2},
1151 + pages = {353--367},
1152 + doi = {10.1016/j.jempfin.2010.11.005}
1153 +}
1154 +@article{christensen2022drift,
1155 + author = {Christensen, Kim and Oomen, Roel and Ren{\`o}, Roberto},
1156 + title = {The drift burst hypothesis},
1157 + journal = {Journal of Econometrics},
1158 + year = {2022},
1159 + volume = {227},
1160 + number = {2},
1161 + pages = {461--497},
1162 + doi = {10.1016/j.jeconom.2020.11.004}
1163 +}
1164 +@article{calcagnile2018collective,
1165 + author = {Calcagnile, Lucio Maria and Bormetti, Giacomo and Treccani, Michele and Marmi, Stefano and Lillo, Fabrizio},
1166 + title = {Collective synchronization and high frequency systemic instabilities in financial markets},
1167 + journal = {Quantitative Finance},
1168 + year = {2018},
1169 + volume = {18},
1170 + number = {2},
1171 + pages = {237--247},
1172 + doi = {10.1080/14697688.2017.1403141}
1173 +}
1174 +@article{bormetti2015modelling,
1175 + author = {Bormetti, Giacomo and Calcagnile, Lucio Maria and Treccani, Michele and Corsi, Fulvio and Marmi, Stefano and Lillo, Fabrizio},
1176 + title = {Modelling systemic price cojumps with {H}awkes factor models},
1177 + journal = {Quantitative Finance},
1178 + year = {2015},
1179 + volume = {15},
1180 + number = {7},
1181 + pages = {1137--1156},
1182 + doi = {10.1080/14697688.2014.996586}
1183 +}
1184 +@article{lo1990nonsynchronous,
1185 + author = {Lo, Andrew W. and MacKinlay, A. Craig},
1186 + title = {An econometric analysis of nonsynchronous trading},
1187 + journal = {Journal of Econometrics},
1188 + year = {1990},
1189 + volume = {45},
1190 + number = {1--2},
1191 + pages = {181--211},
1192 + doi = {10.1016/0304-4076(90)90098-E}
1193 +}
1194 +@article{bollerslev2013jump,
1195 + author = {Bollerslev, Tim and Todorov, Viktor and Li, Sophia Zhengzi},
1196 + title = {Jump tails, extreme dependencies, and the distribution of stock returns},
1197 + journal = {Journal of Econometrics},
1198 + year = {2013},
1199 + volume = {172},
1200 + number = {2},
1201 + pages = {307--324},
1202 + doi = {10.1016/j.jeconom.2012.08.014}
1203 +}
1204 +@article{johnson2013abrupt,
1205 + author = {Johnson, Neil and Zhao, Guannan and Hunsader, Eric and Qi, Hong and Johnson, Nicholas and Meng, Jing and Tivnan, Brian},
1206 + title = {Abrupt rise of new machine ecology beyond human response time},
1207 + journal = {Scientific Reports},
1208 + year = {2013},
1209 + volume = {3},
1210 + pages = {2627},
1211 + doi = {10.1038/srep02627}
1212 +}
1213 +@article{dungey2018testing,
1214 + author = {Dungey, Mardi and Erdemlioglu, Deniz and Matei, Marius and Yang, Xiye},
1215 + title = {Testing for mutually exciting jumps and financial flights in high frequency data},
1216 + journal = {Journal of Econometrics},
1217 + year = {2018},
1218 + volume = {202},
1219 + number = {1},
1220 + pages = {18--44},
1221 + doi = {10.1016/j.jeconom.2017.09.002}
1222 +}
1223 +@article{hansen2006realized,
1224 + author = {Hansen, Peter R. and Lunde, Asger},
1225 + title = {Realized Variance and Market Microstructure Noise},
1226 + journal = {Journal of Business \& Economic Statistics},
1227 + year = {2006},
1228 + volume = {24},
1229 + number = {2},
1230 + pages = {127--161},
1231 + doi = {10.1198/073500106000000071}
1232 +}
1233 +@article{harris1991stock,
1234 + author = {Harris, Lawrence},
1235 + title = {Stock Price Clustering and Discreteness},
1236 + journal = {Review of Financial Studies},
1237 + year = {1991},
1238 + volume = {4},
1239 + number = {3},
1240 + pages = {389--415},
1241 + doi = {10.1093/rfs/4.3.389}
1242 +}
1243 +@article{bessembinder2016liquidity,
1244 + author = {Bessembinder, Hendrik and Carrion, Allen and Tuttle, Laura and Venkataraman, Kumar},
1245 + title = {Liquidity, resiliency and market quality around predictable trades: Theory and evidence},
1246 + journal = {Journal of Financial Economics},
1247 + year = {2016},
1248 + volume = {121},
1249 + number = {1},
1250 + pages = {142--166},
1251 + doi = {10.1016/j.jfineco.2016.02.011}
1252 +}
1253 +@article{scholes1977estimating,
1254 + author = {Scholes, Myron and Williams, Joseph},
1255 + title = {Estimating betas from nonsynchronous data},
1256 + journal = {Journal of Financial Economics},
1257 + year = {1977},
1258 + volume = {5},
1259 + number = {3},
1260 + pages = {309--327},
1261 + doi = {10.1016/0304-405X(77)90041-1}
1262 +}
modified paper/highlights.pdf +0 −0

Binary file not shown.

modified paper/highlights.tex +4 −4
@@ -4,9 +4,9 @@
4 4 \section*{Highlights}
5 5 \begin{itemize}
6 6 \item 7.7 billion one-minute bars, full U.S. cross-section incl.\ delistings, 1992--2026
7 −\item Isolated $6\sigma$ shocks show zero market-wide propagation vs matched placebos
8 −\item Propagation is network-concentrated and tripled from the 1990s to the 2020s
9 −\item Epps effect vanished: one-minute cross-correlations rose from 0.00 to 0.30
10 −\item Extreme minutes fell $\sim$56-fold, but synchronized co-jumps spike in crises
7 +\item Isolated $6\sigma$ shocks raise market-wide tail activity by at most 2\% of base rate
8 +\item Propagation is concentrated among 20 correlation neighbours and tripled since the 1990s
9 +\item Epps effect receded: one-minute cross-correlations rose from 0.00 (1996) to 0.30 (2015)
10 +\item Extreme minutes fell $\sim$100-fold; severe co-jump minutes are 10$\times$ more frequent in crises
11 11 \end{itemize}
12 12 \end{document}
modified paper/main.pdf +0 −0

Binary file not shown.

modified paper/main.tex +22 −18
@@ -103,26 +103,30 @@
103 103 \newcommand{\WPnumber}{15}
104 104 \newcommand{\WPtitle}{Faster, Wider, More Fragile? The Evolution of Intraday Shock Propagation in U.S. Equity Markets, 1992--2026}
105 105 \newcommand{\WPsubtitle}{}
106 −\newcommand{\WPdate}{July 2026}
107 −\newcommand{\WPversion}{1.0}
106 +\newcommand{\WPdate}{October 2026}
107 +\newcommand{\WPversion}{1.1}
108 108 \newcommand{\WPabstract}{%
109 109 We study how intraday shocks travel across U.S.\ equities using
110 110 \DataRowsBn{} billion one-minute bars covering the full listed and delisted
111 −cross-section from 1992 to 2026. Three facts emerge. First, markets absorb
112 −idiosyncratic shocks: across \NEventsRealMM{} million isolated $\pm6\sigma$
113 −one-minute events, abnormal market-wide propagation---measured against matched
114 −within-stock placebos---is economically zero at horizons of 1 to 120 minutes,
115 −in every decade. Second, transmission is concentrated and has intensified:
116 −the shocked stock's twenty closest lagged-correlation neighbors respond
117 −strongly (+\NetDiffFive$\times10^{-4}$ abnormal jump probability at five
118 −minutes), and this network channel tripled between the 1990s and the 2020s
119 −while market-wide spillover stayed at zero. Third, tail synchronization
120 −migrated to the market-wide minute: ordinary extreme minute returns became
121 −roughly \ExtremeDropFactor{} times less frequent than in the 1990s, the Epps
122 −curve flattened (one-minute pairwise correlation \EppsNinetySixOne{} in 1996
123 −versus \EppsFifteenOne{} in 2015), and severe co-jump minutes spike by orders
124 −of magnitude in 2008 and 2020. Modern markets absorb the idiosyncratic and
125 −transmit the systemic.
111 +cross-section from 1992 to 2026. Three findings emerge. First, isolated
112 +idiosyncratic shocks barely spread: across \NEventsRealMM{} million
113 +$\pm6\sigma$ one-minute events, the abnormal share of other securities with
114 +an extreme standardized return, measured against matched within-stock
115 +placebos, lies between zero and about \StratShareOfBase\% of its base rate at
116 +horizons of 1 to 120 minutes, in every decade. Second, transmission is
117 +concentrated and has intensified: the shocked stock's twenty closest
118 +lagged-correlation neighbors respond (+\NetDiffFive$\times10^{-4}$ in the
119 +probability of an extreme move at five minutes), and this network channel
120 +roughly tripled between the 1990s and the 2020s while market-wide spillover
121 +stayed bounded near zero---a rise that partly reflects sharper identification
122 +of the network as the Epps effect receded. Third, common information is
123 +impounded far faster (one-minute pairwise correlation \EppsNinetySixOne{} in
124 +1996 versus \EppsFifteenOne{} in 2015), ordinary extreme minutes became an
125 +order of magnitude rarer after decimalization, and severe co-jump
126 +minutes---more than a tenth of the universe moving together---are roughly
127 +ten times more frequent in 2008 and 2020 than in calm years. Modern markets
128 +absorb the idiosyncratic, transmit it narrowly along statistical linkages,
129 +and synchronize on the systemic.
126 130 }
127 131 \newcommand{\WPkeywords}{intraday shock propagation, market fragility, high-frequency data, contagion, co-jumps, Epps effect, market microstructure}
128 132 \newcommand{\WPjel}{G10, G12, G14}
@@ -160,7 +164,7 @@ transmit the systemic.
160 164 \input{sections/09_conclusion}
161 165
162 166 % --- References ---
163 −\newpage
167 +\clearpage
164 168 \bibliography{bibliography/references}
165 169
166 170 \end{document}
modified paper/main_irfa.pdf +0 −0

Binary file not shown.

modified paper/main_irfa.tex +25 −18
@@ -37,21 +37,25 @@ Propagation in U.S. Equity Markets, 1992--2026}
37 37 \begin{abstract}
38 38 We study how intraday shocks travel across U.S.\ equities using
39 39 \DataRowsBn{} billion one-minute bars covering the full listed and delisted
40 −cross-section from 1992 to 2026. Three facts emerge. First, markets absorb
41 −idiosyncratic shocks: across \NEventsRealMM{} million isolated $\pm6\sigma$
42 −one-minute events, abnormal market-wide propagation---measured against matched
43 −within-stock placebos---is economically zero at horizons of 1 to 120 minutes,
44 −in every decade. Second, transmission is concentrated and has intensified:
45 −the shocked stock's twenty closest lagged-correlation neighbors respond
46 −strongly (+\NetDiffFive$\times10^{-4}$ abnormal jump probability at five
47 −minutes), and this network channel tripled between the 1990s and the 2020s
48 −while market-wide spillover stayed at zero. Third, tail synchronization
49 −migrated to the market-wide minute: ordinary extreme minute returns became
50 −roughly \ExtremeDropFactor{} times less frequent than in the 1990s, the Epps
51 −curve flattened (one-minute pairwise correlation \EppsNinetySixOne{} in 1996
52 −versus \EppsFifteenOne{} in 2015), and severe co-jump minutes spike by orders
53 −of magnitude in 2008 and 2020. Modern markets absorb the idiosyncratic and
54 −transmit the systemic.
40 +cross-section from 1992 to 2026. Three findings emerge. First, isolated
41 +idiosyncratic shocks barely spread: across \NEventsRealMM{} million
42 +$\pm6\sigma$ one-minute events, the abnormal share of other securities with
43 +an extreme standardized return, measured against matched within-stock
44 +placebos, lies between zero and about \StratShareOfBase\% of its base rate at
45 +horizons of 1 to 120 minutes, in every decade. Second, transmission is
46 +concentrated and has intensified: the shocked stock's twenty closest
47 +lagged-correlation neighbors respond (+\NetDiffFive$\times10^{-4}$ in the
48 +probability of an extreme move at five minutes), and this network channel
49 +roughly tripled between the 1990s and the 2020s while market-wide spillover
50 +stayed bounded near zero---a rise that partly reflects sharper identification
51 +of the network as the Epps effect receded. Third, common information is
52 +impounded far faster (one-minute pairwise correlation \EppsNinetySixOne{} in
53 +1996 versus \EppsFifteenOne{} in 2015), ordinary extreme minutes became an
54 +order of magnitude rarer after decimalization, and severe co-jump
55 +minutes---more than a tenth of the universe moving together---are roughly
56 +ten times more frequent in 2008 and 2020 than in calm years. Modern markets
57 +absorb the idiosyncratic, transmit it narrowly along statistical linkages,
58 +and synchronize on the systemic.
55 59 \end{abstract}
56 60
57 61 \begin{keyword}
@@ -71,6 +75,7 @@ contagion \sep co-jumps \sep Epps effect \sep market microstructure
71 75 \input{sections/07_mechanisms}
72 76 \input{sections/08_robustness}
73 77 \input{sections/09_conclusion}
78 +\clearpage
74 79
75 80 \section*{CRediT authorship contribution statement}
76 81 \textbf{Simon-Pierre Boucher:} Conceptualization, Methodology, Software,
@@ -84,12 +89,14 @@ paper.
84 89
85 90 \section*{Data availability}
86 91 The underlying market data were originally sourced from Finnhub.io
87 −(\url{https://finnhub.io}). The aggregated monthly archive used in this study is publicly accessible;
88 −the exact pinned snapshot, full replication code, manifests, checksums, and
92 +(\url{https://finnhub.io}). The aggregated monthly archive used in this study
93 +is publicly accessible but carries no explicit licence; the minute data are
94 +therefore not redistributed. The exact pinned snapshot identifier and
95 +checksums, the full replication code, the event-level estimates, and all
89 96 machine-readable results accompany the paper.
90 97
91 98 \section*{Acknowledgments}
92 −[To be completed.]
99 +All errors are the author's own.
93 100
94 101 \bibliographystyle{elsarticle-harv}
95 102 \bibliography{bibliography/references}
modified paper/main_irfa_anon.pdf +0 −0

Binary file not shown.

modified paper/main_irfa_anon.tex +24 −17
@@ -28,21 +28,25 @@ Propagation in U.S. Equity Markets, 1992--2026}
28 28 \begin{abstract}
29 29 We study how intraday shocks travel across U.S.\ equities using
30 30 \DataRowsBn{} billion one-minute bars covering the full listed and delisted
31 −cross-section from 1992 to 2026. Three facts emerge. First, markets absorb
32 −idiosyncratic shocks: across \NEventsRealMM{} million isolated $\pm6\sigma$
33 −one-minute events, abnormal market-wide propagation---measured against matched
34 −within-stock placebos---is economically zero at horizons of 1 to 120 minutes,
35 −in every decade. Second, transmission is concentrated and has intensified:
36 −the shocked stock's twenty closest lagged-correlation neighbors respond
37 −strongly (+\NetDiffFive$\times10^{-4}$ abnormal jump probability at five
38 −minutes), and this network channel tripled between the 1990s and the 2020s
39 −while market-wide spillover stayed at zero. Third, tail synchronization
40 −migrated to the market-wide minute: ordinary extreme minute returns became
41 −roughly \ExtremeDropFactor{} times less frequent than in the 1990s, the Epps
42 −curve flattened (one-minute pairwise correlation \EppsNinetySixOne{} in 1996
43 −versus \EppsFifteenOne{} in 2015), and severe co-jump minutes spike by orders
44 −of magnitude in 2008 and 2020. Modern markets absorb the idiosyncratic and
45 −transmit the systemic.
31 +cross-section from 1992 to 2026. Three findings emerge. First, isolated
32 +idiosyncratic shocks barely spread: across \NEventsRealMM{} million
33 +$\pm6\sigma$ one-minute events, the abnormal share of other securities with
34 +an extreme standardized return, measured against matched within-stock
35 +placebos, lies between zero and about \StratShareOfBase\% of its base rate at
36 +horizons of 1 to 120 minutes, in every decade. Second, transmission is
37 +concentrated and has intensified: the shocked stock's twenty closest
38 +lagged-correlation neighbors respond (+\NetDiffFive$\times10^{-4}$ in the
39 +probability of an extreme move at five minutes), and this network channel
40 +roughly tripled between the 1990s and the 2020s while market-wide spillover
41 +stayed bounded near zero---a rise that partly reflects sharper identification
42 +of the network as the Epps effect receded. Third, common information is
43 +impounded far faster (one-minute pairwise correlation \EppsNinetySixOne{} in
44 +1996 versus \EppsFifteenOne{} in 2015), ordinary extreme minutes became an
45 +order of magnitude rarer after decimalization, and severe co-jump
46 +minutes---more than a tenth of the universe moving together---are roughly
47 +ten times more frequent in 2008 and 2020 than in calm years. Modern markets
48 +absorb the idiosyncratic, transmit it narrowly along statistical linkages,
49 +and synchronize on the systemic.
46 50 \end{abstract}
47 51
48 52 \begin{keyword}
@@ -62,6 +66,7 @@ contagion \sep co-jumps \sep Epps effect \sep market microstructure
62 66 \input{sections/07_mechanisms}
63 67 \input{sections/08_robustness}
64 68 \input{sections/09_conclusion}
69 +\clearpage
65 70
66 71 \section*{Declaration of competing interest}
67 72 The author declares no known competing financial interests or personal
@@ -70,8 +75,10 @@ paper.
70 75
71 76 \section*{Data availability}
72 77 The underlying market data were originally sourced from Finnhub.io
73 −(\url{https://finnhub.io}). The aggregated monthly archive used in this study is publicly accessible;
74 −the exact pinned snapshot, full replication code, manifests, checksums, and
78 +(\url{https://finnhub.io}). The aggregated monthly archive used in this study
79 +is publicly accessible but carries no explicit licence; the minute data are
80 +therefore not redistributed. The exact pinned snapshot identifier and
81 +checksums, the full replication code, the event-level estimates, and all
75 82 machine-readable results accompany the paper.
76 83
77 84 \bibliographystyle{elsarticle-harv}
modified paper/results_macros.tex +49 −0
@@ -47,3 +47,52 @@
47 47 \newcommand{\CoJumpSevCovid}{6.4}
48 48 \newcommand{\SevereCovid}{83.7}
49 49 \newcommand{\SeverePreCovid}{9.53}
50 +\newcommand{\SevereGFC}{91.1}
51 +\newcommand{\SevereEarlyNineties}{128}
52 +\newcommand{\CoJumpSevEarlyNineties}{7.1}
53 +\newcommand{\CoJumpPerThousandRecent}{131}
54 +\newcommand{\ZeroRetNinetySix}{52}
55 +\newcommand{\ZeroRetTwentyFour}{13}
56 +\newcommand{\ExtremeGtOneTwoThousandTwo}{50}
57 +\newcommand{\ExtremeGtOneNinetyTwo}{673}
58 +\newcommand{\PreBreadthReal}{1,044}
59 +\newcommand{\PreBreadthPlacebo}{1,004}
60 +\newcommand{\PreBreadthDiff}{40.3}
61 +\newcommand{\PreBreadthDiffSE}{1.2}
62 +\newcommand{\PreIntensityDiff}{0.074}
63 +\newcommand{\PreIntensityDiffSE}{0.012}
64 +\newcommand{\BaseBreadthShare}{0.10}
65 +\newcommand{\NEligMean}{427}
66 +\newcommand{\NEligNineties}{237}
67 +\newcommand{\EventsPerStockWeek}{2.5}
68 +\newcommand{\EventDays}{7,282}
69 +\newcommand{\NeighborBaseShare}{0.10}
70 +\newcommand{\MatchRate}{95.2}
71 +\newcommand{\MatchRateNineties}{71}
72 +\newcommand{\MatchRateMinPost}{96}
73 +\newcommand{\CumBaseFifteen}{58}
74 +\newcommand{\MDEFive}{2.5}
75 +\newcommand{\MDEEventsPer}{9}
76 +\newcommand{\ShareSingleton}{22}
77 +\newcommand{\MeanCoArrival}{2.1}
78 +\newcommand{\DDSingletonFive}{1.7}
79 +\newcommand{\DDSingletonFiveSE}{1.7}
80 +\newcommand{\DDSingletonThirty}{3.7}
81 +\newcommand{\DDSingletonThirtySE}{1.9}
82 +\newcommand{\DDMultiThirty}{-8.3}
83 +\newcommand{\DDMultiThirtySE}{2.1}
84 +\newcommand{\NetDiffSingleton}{51}
85 +\newcommand{\NetDiffSingletonSE}{2.5}
86 +\newcommand{\PostDiffFive}{40}
87 +\newcommand{\StratFive}{19}
88 +\newcommand{\StratFiveSE}{0.9}
89 +\newcommand{\StratThirty}{18}
90 +\newcommand{\StratThirtySE}{1.0}
91 +\newcommand{\StratShareOfBase}{2}
92 +\newcommand{\NetDiffFiveTickerSE}{4.7}
93 +\newcommand{\NetDiffNeg}{66}
94 +\newcommand{\NetDiffPos}{63}
95 +\newcommand{\NetRealNdNineties}{228}
96 +\newcommand{\NetPlaceboNdNineties}{200}
97 +\newcommand{\NetRealNdTens}{37}
98 +\newcommand{\NetPlaceboNdTens}{-44}
modified paper/sections/01_introduction.tex +112 −92
@@ -1,117 +1,137 @@
1 1 \section{Introduction}
2 2 \label{sec:introduction}
3 3
4 −When one security suffers a violent price move, does the shock stay local or
5 −does it spread? The answer shapes almost everything that happens inside a
6 −trading day: how fast prices aggregate information, how much liquidity
7 −providers must charge for standing in the market, how execution algorithms
8 −should schedule parent orders, and whether an idiosyncratic event can grow
9 −into a market-wide dislocation. Regulators confront the same question from the
10 −other side: episodes such as May 6, 2010 showed that modern equity markets can
11 −move from normal functioning to synchronized collapse within minutes, faster
12 −than any human intervention. Yet the empirical record on intraday shock
13 −transmission is built almost entirely from short samples, small panels, or
14 −single episodes. Whether the propagation mechanism itself has changed over the
15 −three decades in which trading moved from exchange floors to fragmented,
16 −algorithmic, ETF-intermediated electronic markets is, to our knowledge,
17 −undocumented.
4 +When one security suffers a violent price move within a trading minute, does
5 +the shock stay local or does it spread? The answer shapes how fast prices
6 +aggregate information, how much liquidity providers must charge for standing
7 +in the market, how execution algorithms should schedule parent orders, and
8 +whether an idiosyncratic event can grow into a market-wide dislocation.
9 +Regulators confront the same question from the other side: May~6, 2010 showed
10 +that modern equity markets can move from normal functioning to synchronized
11 +collapse within minutes, faster than any human intervention
12 +\citep{kirilenko2017flash,menkveld2019flash}. Yet the empirical record on
13 +intraday shock transmission is built almost entirely from short samples,
14 +small panels, or single episodes. Whether the propagation mechanism itself
15 +changed over the three decades in which trading moved from exchange floors to
16 +fragmented, algorithmic, ETF-intermediated electronic markets is, to our
17 +knowledge, undocumented.
18 18
19 −The tension motivating this paper is that theory predicts change in both
20 −directions. Faster, cheaper, more competitive markets should absorb ordinary
21 −idiosyncratic shocks more efficiently: liquidity provision reacts within
22 −seconds, arbitrage capital is abundant in normal times, and information is
23 −incorporated nearly continuously. But the same institutions that supply this
24 −efficiency---high-speed market making with tight inventory limits, cross-asset
25 −arbitrage, index and basket intermediation---withdraw or transmit pressure
26 −precisely when many securities move together. Modern markets may therefore
27 −absorb ordinary shocks more efficiently while transmitting extreme shocks more
28 −rapidly and discontinuously.
19 +Theory predicts change in both directions. Faster, cheaper, more competitive
20 +markets should absorb ordinary idiosyncratic shocks more efficiently:
21 +liquidity provision reacts within seconds, arbitrage capital is abundant in
22 +normal times, and information is incorporated nearly continuously
23 +\citep{hendershott2011algorithmic,menkveld2013high,brogaard2014high}. But the
24 +same institutions that supply this efficiency---high-speed market making with
25 +tight inventory limits, cross-asset arbitrage, index and basket
26 +intermediation---withdraw or transmit pressure precisely when many securities
27 +move together \citep{brogaard2018high,cespa2014illiquidity,bendavid2018etfs}.
28 +Modern markets may therefore absorb ordinary shocks more efficiently while
29 +synchronizing more tightly on common ones. The title asks whether U.S.\
30 +equity markets became faster, wider, and more fragile at the intraday horizon.
31 +The answer this paper documents is: faster, yes; wider, no---transmission
32 +became sharper rather than broader; more fragile, only in a specific and
33 +bounded sense that we make precise.
29 34
30 −We ask how the speed, breadth, persistence, and asymmetry of intraday shock
31 −propagation across U.S.\ equities evolved over the past three decades, and
32 −under which market conditions an initially idiosyncratic shock becomes
33 −sectoral or market-wide.
34 −
35 −Our laboratory is a panel of \DataRows{} one-minute OHLCV bars, originally
36 −sourced from Finnhub.io, covering
37 −\DataTickers{} tickers---including delisted securities---from January 1992
35 +Our laboratory is a panel of \DataRows{} one-minute bars covering
36 +\DataTickers{} tickers, including delisted securities, from January 1992
38 37 through March 2026. The data are audited exhaustively before use: observed
39 38 trading dates match the official exchange calendar exactly over thirty-four
40 39 years, timestamp conventions are verified on both sides of every
41 40 daylight-saving transition, and integrity violations are negligible
42 41 (\DataOHLCViol{} impossible OHLC bars in \DataRowsBn{} billion). Because the
43 −source provides bars only for minutes with trades, and identifies securities
44 −by ticker alone, the empirical design works with liquidity-screened,
45 −lagged-rank universes, ticker spells rather than tickers, and detected
46 −corporate-action windows; Section~\ref{sec:data} documents every screen.
42 +source provides bars only for minutes with trades, identifies securities by
43 +ticker alone, and does not document its venue scope, the empirical design
44 +works with liquidity-screened, lagged-rank universes, ticker spells rather
45 +than tickers, and detected corporate-action windows; Section~\ref{sec:data}
46 +documents every screen and every limitation we could not remove.
47 47
48 −Three findings organize the paper. First, \emph{markets absorb isolated
49 −idiosyncratic shocks---and we can measure how completely}. We detect
50 −\NEventsRealMM{} million one-minute events in which a single top-500 stock
51 −moves by at least six robust standard deviations of its own residual return
52 −while the rest of the market is calm. Relative to matched within-stock
53 −placebos, the abnormal share of other securities responding within 5 to 120
54 −minutes is economically zero (point estimates of $-1$ to $-3\times10^{-4}$
55 −with day-block bootstrap standard errors of about $1\times10^{-4}$), in every
56 −decade, for positive and negative shocks alike. Idiosyncratic events die where
57 −they land.
48 +Three findings organize the paper. First, \emph{isolated idiosyncratic shocks
49 +do not raise market-wide tail activity by more than a few percent of its base
50 +rate}. We detect \NEventsRealMM{} million one-minute events in which a
51 +top-500 stock moves by at least six robust standard deviations of its own
52 +residual return while the market factor is calm and fewer than 2\% of other
53 +names move simultaneously. Relative to matched within-stock placebos, the
54 +difference-in-differences estimate of the share of other securities with a
55 +standardized residual beyond two in absolute value is between $-1$ and
56 +$-3\times10^{-4}$ at horizons of 1 to 120 minutes, against a base rate of
57 +about \BaseBreadthShare{}, with day-block bootstrap standard errors near
58 +$1\times10^{-4}$. We also show, however, that real events occur in minutes
59 +of elevated ambient cross-sectional activity (pre-event breadth is
60 +\PreBreadthDiff$\times10^{-4}$ higher than for placebos), so that the
61 +difference-in-differences and a pre-level-stratified comparison bracket the
62 +market-wide response between zero and roughly \StratShareOfBase\% of the base
63 +rate. Both bounds are economically small, and the pattern holds in every
64 +decade and for positive and negative shocks alike.
58 65
59 −Second, \emph{the transmission of common information has become nearly
60 −instantaneous}. In 1996 the average pairwise correlation of one-minute returns
61 −among the hundred largest stocks was \EppsNinetySixOne{}---economically zero,
62 −the classical Epps effect---while thirty-minute returns already comoved at
63 −\EppsNinetySixThirty. By 2015 the one-minute correlation was
64 −\EppsFifteenOne{} against \EppsFifteenThirty{} at thirty minutes: the Epps
65 −curve has flattened almost completely. Information that once took half an hour
66 −to spread across the cross-section now spreads within the minute.
66 +Second, \emph{what transmission exists is concentrated and, within the limits
67 +of our measurement, has intensified}. The shocked stock's twenty closest
68 +lagged-correlation neighbors are \NetDiffFive$\times10^{-4}$ more likely than
69 +other receivers to show a tail exceedance five minutes later, net of the same
70 +contrast for placebos---a response of about six percent of the neighbor base
71 +rate, estimated with a standard error of \NetDiffFiveSE$\times10^{-4}$
72 +(\NetDiffFiveTickerSE$\times10^{-4}$ with ticker clustering). The
73 +differential is \NetDiffFiveNineties$\times10^{-4}$ in the 1990s,
74 +\NetDiffFiveOughts{} in the 2000s, \NetDiffFiveTens{} in the 2010s, and
75 +\NetDiffFiveTwenties{} in the 2020s. We present this increase with an
76 +explicit caveat: neighbor sets are identified from one-minute residual
77 +correlations in the preceding month, and the paper's own Epps evidence shows
78 +that those correlations were close to zero in the 1990s, so part of the
79 +secular rise reflects sharper identification of the network rather than
80 +stronger transmission along it.
67 81
68 −Third, \emph{tail risk has migrated from the security to the minute}. Ordinary
69 −extreme moves became rare: the frequency of one-minute returns exceeding 1\%
70 −fell from roughly \ExtremeGtOneNineties{} per 10,000 in the 1990s to
71 −\ExtremeGtOneTens{} per 10,000 in the 2010s---a factor of about
72 −\ExtremeDropFactor. But synchronized extremes did not: minutes in which more
73 −than 2\% of the universe jumps beyond $4\sigma$ simultaneously---systemic
74 −co-jump minutes---spike by orders of magnitude in 2008 and 2020, and the
82 +Third, \emph{common information is impounded across the cross-section far
83 +faster than it was, and tail risk within the post-decimalization era is
84 +concentrated in market-wide minutes}. In 1996 the mean pairwise correlation
85 +of one-minute returns among the hundred largest stocks was
86 +\EppsNinetySixOne{} against \EppsNinetySixThirty{} at thirty minutes---the
87 +classical Epps effect---whereas in 2015 it was \EppsFifteenOne{} against
88 +\EppsFifteenThirty{}. The frequency of one-minute returns exceeding 1\% fell
89 +from roughly \ExtremeGtOneNineties{} per 10,000 in the 1990s to
90 +\ExtremeGtOneTens{} per 10,000 in the 2010s; part of that decline is the
91 +mechanical disappearance of eighth and sixteenth ticks, and the decline
92 +measured from 2002 onward is roughly tenfold. Minutes in which more than 10\%
93 +of the universe moves beyond four robust standard deviations
94 +simultaneously---severe co-jump minutes---are roughly an order of magnitude
95 +more frequent in 2008 and 2020 than in surrounding calm years, and the
75 96 frequency of $|r|>5\%$ minute returns rose from \ExtremeGtFivePreCovid{} per
76 −10,000 in 2017--2019 to \ExtremeGtFiveCovid{} in 2020, a factor of
77 −\ExtremeCovidFactor. The joint pattern---absorption of the idiosyncratic,
78 −synchronization of the systemic---is the modern form of market fragility.
97 +10,000 in 2017--2019 to \ExtremeGtFiveCovid{} in 2020. We do not claim that
98 +such synchronization is secularly more frequent than in the 1990s: under the
99 +same standardization the wide-tick 1990s produce severe-minute counts that
100 +exceed 2008, which we attribute to discreteness rather than fragility and
101 +which precludes a cross-regime comparison.
79 102
80 103 Identification receives explicit attention throughout. The central threat is
81 −that securities respond to common information rather than to each other. Our
104 +that securities respond to common information rather than to each other. The
82 105 shock definition residualizes against a leave-one-out market factor with
83 106 strictly lagged loadings and scales; isolation filters require a calm market
84 −and fewer than 2\% simultaneous extremes at the event minute; every event is
85 −compared to a placebo in the same stock at the same minute of day on a calm
86 −day; and the pre-event window shows no abnormal response, ruling out
87 −anticipation or mechanical selection. What the isolation filters remove---the
88 −simultaneous-extreme minutes---we do not discard but study separately as the
89 −systemic co-jump sample. Common-information contamination is thus measured,
90 −not assumed away.
91 −
92 −Mechanically, we ask \emph{where} propagation should hide if it exists:
93 −responses concentrated among the shocked stock's closest lagged-correlation
94 −neighbors, cumulative rather than point-in-time responses, and state
95 −dependence across calm and stressed markets and across shock magnitudes.
96 −Section~\ref{sec:mechanisms} reports these probes.
107 +factor and fewer than 2\% simultaneous extremes at the event minute; every
108 +event is compared to a placebo in the same stock at the same minute of day on
109 +a calm day; and the pre-event level of every outcome is reported rather than
110 +assumed equal across arms. The minutes that the isolation filters remove are
111 +not discarded but studied separately as the systemic co-jump sample. What the
112 +design identifies is the predictive association between an isolated
113 +idiosyncratic extreme and subsequent cross-sectional tail activity; it does
114 +not separate fundamental information transmission from trading-channel
115 +transmission, and we phrase every claim accordingly.
97 116
98 117 The paper contributes to four literatures. To the contagion literature, which
99 118 measures transmission at daily or lower frequency for small panels, we
100 119 contribute event-level evidence from the full U.S.\ cross-section at the
101 120 minute scale over thirty-four years. To the market-microstructure and
102 121 high-frequency-trading debate, which contrasts normal-times liquidity gains
103 −with episodic fragility, we contribute the missing time series: propagation
104 −and synchronization measured uniformly across the entire structural
105 −transformation of U.S.\ trading. To the co-jump econometrics literature, we
106 −contribute a long-sample account of how systemic co-jump frequency and
107 −severity evolved. And to the measurement literature, we document what OHLCV
108 −bars can and cannot identify about liquidity and transmission, with every
109 −limitation quantified.
122 +with episodic fragility, we contribute a uniform time series of propagation
123 +and synchronization measured across the entire structural transformation of
124 +U.S.\ trading. To the co-jump literature, we contribute a long-sample account
125 +of how the frequency and severity of synchronized extreme minutes evolved,
126 +with the comparability limits of such a series made explicit. And to the
127 +measurement literature, we document what OHLCV bars can and cannot identify
128 +about transmission, with every limitation quantified.
110 129
111 −Section~\ref{sec:data} describes and audits the data.
112 −Section~\ref{sec:measurement} defines shocks, outcomes, and networks.
113 −Section~\ref{sec:design} presents the empirical design.
114 −Section~\ref{sec:results} reports the descriptive transformation of the market
115 −and the baseline propagation results. Section~\ref{sec:mechanisms} examines
116 −networks, cumulative responses, and co-jumps. Section~\ref{sec:robustness}
117 −summarizes robustness, and Section~\ref{sec:conclusion} concludes.
130 +Section~\ref{sec:literature} positions the paper. Section~\ref{sec:data}
131 +describes and audits the data. Section~\ref{sec:measurement} defines shocks,
132 +outcomes, and networks. Section~\ref{sec:design} presents the empirical
133 +design and the inference. Section~\ref{sec:results} reports the descriptive
134 +transformation of the market and the baseline propagation results, including
135 +the design diagnostics. Section~\ref{sec:mechanisms} examines networks,
136 +cumulative responses, and co-jumps. Section~\ref{sec:robustness} summarizes
137 +robustness, and Section~\ref{sec:conclusion} concludes.
modified paper/sections/02_literature.tex +215 −237
@@ -1,291 +1,269 @@
1 1 \section{Related literature}
2 2 \label{sec:literature}
3 3
4 −% All 110 citations verified at DOI level via OpenAlex/Crossref on 2026-07-13;
5 −% see docs/LITERATURE_MAP.md. Novelty statements are qualified by the documented
6 −% search of the same date.
4 +% Citations verified at DOI level via OpenAlex/Crossref on 2026-07-13 and
5 +% re-verified against Crossref on 2026-10-17 (docs/CHANGELOG-REVISION.md).
7 6
8 −This paper measures how quickly, how broadly, how persistently, and how
9 −asymmetrically idiosyncratic intraday shocks travel across the U.S.\ equity
10 −cross-section, and how that transmission evolved between 1992 and 2026. The
11 −question sits at the intersection of several literatures: theories of
7 +The question sits at the intersection of several literatures: theories of
12 8 cross-asset shock transmission; empirical work on fire sales, flows, and
13 −ownership linkages; the measurement of comovement, commonality, and liquidity
14 −dynamics; lead--lag effects and information diffusion; network approaches to
15 −contagion and systemic risk; the transformation of U.S.\ market structure and
16 −the high-frequency trading debate; the econometrics of jumps, co-jumps, and
17 −scheduled announcements; and the measurement toolkit for high-frequency and
18 −OHLCV data. We take each in turn and close with the paper's positioning.
9 +ownership linkages; the measurement of comovement and liquidity commonality;
10 +lead--lag effects and the Epps effect; network approaches to contagion; the
11 +transformation of U.S.\ market structure and the high-frequency trading
12 +debate; and the econometrics of jumps, co-jumps, and scheduled announcements.
13 +We take each in turn and close with the paper's positioning.
19 14
20 15 \subsection{Theories of cross-asset shock transmission}
21 16
22 17 Theory identifies two broad families of channels through which a shock to one
23 18 security moves others. The first is informational. \citet{king1990transmission}
24 −model rational agents who infer information from \emph{other} markets' price
25 −changes, so that an idiosyncratic ``mistake'' in one market transmits broadly---
26 −a mechanism designed to explain why nearly all markets crashed together in
27 −October 1987. \citet{kodres2002rational} show that cross-market portfolio
28 −rebalancing of shared macro risk exposures transmits idiosyncratic shocks even
29 −without common fundamentals, with severity increasing in information asymmetry.
19 +model agents who infer information from other markets' price changes, so that
20 +an idiosyncratic ``mistake'' in one market transmits broadly, and
21 +\citet{kodres2002rational} show that cross-market rebalancing of shared macro
22 +exposures transmits idiosyncratic shocks even without common fundamentals.
30 23 The second family operates through balance sheets and funding.
31 −\citet{kyle2001contagion} generate contagion as a wealth effect: convergence
32 −traders who lose money in one market liquidate positions in others, raising
33 −volatility and correlations exactly when capital is scarce. \citet{yuan2005asymmetric}
34 −shows that borrowing constraints make crises, contagion, and asymmetric price
35 −movements likelier after intermediate losses. \citet{brunnermeier2009market}
36 −link market liquidity to funding liquidity in mutually reinforcing margin
37 −spirals that explain sudden, correlated liquidity dry-ups and flight to
38 −quality. Most directly related to our intraday setting,
39 −\citet{cespa2014illiquidity} show that when liquidity providers learn from
40 −\emph{other securities' prices}, illiquidity becomes contagious and the market
41 −admits liquidity-crash equilibria---a mechanism they connect explicitly to
42 −May~6, 2010. These theories jointly predict that transmission should be
43 −state-dependent (stronger when volatility is high and capital constrained),
44 −asymmetric (stronger for negative shocks), and increasing in the degree to
45 −which pricing and liquidity provision are cross-security activities. Our
46 −measures of propagation speed, breadth, asymmetry, and market-state dependence
47 −are designed to test exactly these comparative statics at the minute scale.
24 +\citet{kyle2001contagion} generate contagion as a wealth effect of convergence
25 +traders; \citet{yuan2005asymmetric} shows that borrowing constraints make
26 +contagion and asymmetric price movements likelier after losses; and
27 +\citet{brunnermeier2009market} link market liquidity to funding liquidity in
28 +margin spirals that explain sudden, correlated liquidity dry-ups. Most
29 +directly related to our setting, \citet{cespa2014illiquidity} show that when
30 +liquidity providers learn from other securities' prices, illiquidity becomes
31 +contagious and the market admits liquidity-crash equilibria, a mechanism they
32 +connect to May~6, 2010. These theories jointly predict that transmission
33 +should be state-dependent, asymmetric, and increasing in the degree to which
34 +pricing and liquidity provision are cross-security activities. Our measures of
35 +propagation breadth, network concentration, sign asymmetry, and market-state
36 +dependence are designed to test these comparative statics at the minute scale.
48 37
49 38 \subsection{Fire sales, flows, and ownership networks}
50 39
51 40 A complementary empirical literature documents non-fundamental transmission
52 −through investor portfolios. Mutual funds facing outflows engage in forced
53 −sales concentrated in their holdings, generating price pressure and reversals
54 −\citep{coval2007asset}; predictable flows generate predictable price pressure
55 −\citep{lou2012flow}; and uninformative trading imbalances in one stock spill
56 −over to correlated stocks through risk-averse liquidity providers
57 −\citep{andrade2008trading}. Ownership structure itself predicts fragility:
58 −stocks with concentrated or correlated-flow ownership have higher volatility
59 −and comovement \citep{greenwood2011stock}, and common active ownership
60 −forecasts cross-sectional correlation beyond fundamentals-based similarity
61 −\citep{anton2014connected}. At the institutional level, overlapping portfolios
62 −create fire-sale spillovers that aggregate into measurable systemic
63 −vulnerability \citep{greenwood2015vulnerable,duarte2021fire}. This work
64 −establishes that idiosyncratic demand shocks \emph{do} propagate, but its
65 −resolution is coarse---monthly or quarterly holdings and flows. Whether the
66 −same propagation operates within the hour, and whether its speed has changed as
67 −intermediation became algorithmic, is the margin our data address.
41 +through investor portfolios: forced sales by mutual funds facing outflows
42 +generate price pressure and reversals \citep{coval2007asset}, predictable
43 +flows generate predictable pressure \citep{lou2012flow}, and uninformative
44 +imbalances in one stock spill over to correlated stocks through risk-averse
45 +liquidity providers \citep{andrade2008trading}. Ownership structure predicts
46 +fragility \citep{greenwood2011stock}, common active ownership forecasts
47 +correlation beyond fundamentals \citep{anton2014connected}, and overlapping
48 +portfolios aggregate into measurable systemic vulnerability
49 +\citep{greenwood2015vulnerable,duarte2021fire}. This work establishes that
50 +idiosyncratic demand shocks do propagate, but at monthly or quarterly
51 +resolution. Whether the same propagation operates within the hour, and
52 +whether its speed changed as intermediation became algorithmic, is the margin
53 +our data address.
68 54
69 55 \subsection{Comovement, commonality, and liquidity dynamics}
70 56
71 −The commonality literature establishes that liquidity has a systematic
72 −component: spreads and depths co-move with market- and industry-wide liquidity
73 −\citep{chordia2000commonality}, common factors in order flow explain much of
74 −the commonality in returns \citep{hasbrouck2001common}, and commonality is
75 −supplied in part by common intermediaries---stocks sharing a NYSE specialist
76 −firm co-move in liquidity beyond the market factor
77 −\citep{coughenour2004common}. Time variation is systematic as well: liquidity
78 −falls after market declines, especially when funding is tight
79 −\citep{hameed2010stock}; specialist inventories and trading losses forecast
80 −spread widening \citep{comertonforde2010time}; and returns to liquidity
81 −provision spike with the VIX as liquidity ``evaporates'' in turmoil
82 −\citep{nagel2012evaporating}. Aggregate liquidity risk is priced
83 −\citep{pastor2003liquidity,acharya2005asset}, and commonality strengthens in
84 −downturns across forty markets \citep{karolyi2012understanding}. Relative to
85 −this literature, we do not measure liquidity \emph{levels}; we measure the
86 −\emph{transmission of extreme events} and use liquidity states---constructed
87 −from OHLCV proxies---as conditioning variables. At very short horizons,
88 −measured comovement is attenuated mechanically by asynchronous trading
89 −\citep{epps1979comovements}, a bias any minute-level design must engineer
90 −around; our breadth measures therefore condition on trading eligibility rather
91 −than assuming synchronous observation.
57 +Liquidity has a systematic component: spreads and depths co-move with market-
58 +and industry-wide liquidity \citep{chordia2000commonality}, common factors in
59 +order flow explain much of the commonality in returns
60 +\citep{hasbrouck2001common}, and commonality is supplied in part by common
61 +intermediaries \citep{coughenour2004common}. Liquidity falls after market
62 +declines, especially when funding is tight \citep{hameed2010stock}, specialist
63 +losses forecast spread widening \citep{comertonforde2010time}, and returns to
64 +liquidity provision spike as liquidity ``evaporates'' in turmoil
65 +\citep{nagel2012evaporating}; aggregate liquidity risk is priced
66 +\citep{pastor2003liquidity,acharya2005asset} and commonality strengthens in
67 +downturns across markets \citep{karolyi2012understanding}. Relative to this
68 +literature we do not measure liquidity levels: bars contain no quotes, and the
69 +only liquidity information we use is prior-month dollar volume as a ranking
70 +device, in the spirit of \citet{amihud2002illiquidity}, and the high--low
71 +range as a descriptive volatility proxy \citep{parkinson1980extreme,
72 +garman1980estimation}. Such OHLC-based proxies have a long pedigree
73 +\citep{roll1984simple,corwin2012simple,goyenko2009liquidity}, but they do not
74 +identify the quote-level mechanisms that the commonality literature studies.
92 75
93 −\subsection{Lead--lag effects and gradual information diffusion}
76 +\subsection{Lead--lag effects, asynchronous trading, and the Epps effect}
94 77
95 78 Cross-autocorrelation in returns is among the oldest facts in empirical asset
96 79 pricing: large stocks lead small stocks \citep{lo1990contrarian}, predominantly
97 −within industries, consistent with gradual information diffusion
98 −\citep{hou2007industry}. Economic links sharpen the pattern: returns of
99 −principal customers predict suppliers' returns \citep{cohen2008economic}, and
100 −cross-predictability follows supplier--customer industry links broadly,
101 −declining with analyst coverage and institutional ownership
102 −\citep{menzly2010market}. At high frequency, tick-data studies document that
103 −liquid assets lead illiquid ones with intraday-seasonal lead--lag structure
104 −\citep{huth2014high}, and \citet{chinco2019sparse} show that one-minute-ahead
105 −cross-stock predictability exists but is sparse, short-lived, and news-related.
106 −These studies measure \emph{unconditional} predictive structure. Our object is
107 −different: the conditional response of the cross-section to an identified
108 −extreme idiosyncratic event, its horizon profile, and the secular evolution of
109 −both. The lead--lag literature nonetheless supplies one of our falsification
110 −benchmarks: propagation estimated through lagged-return networks should
111 −reproduce known liquidity-ordered asymmetries if it is measuring genuine
112 −transmission.
80 +within industries \citep{hou2007industry}, along customer--supplier links
81 +\citep{cohen2008economic,menzly2010market}, and, at tick resolution, from
82 +liquid to illiquid assets with intraday-seasonal structure
83 +\citep{huth2014high}; \citet{chinco2019sparse} show that one-minute-ahead
84 +cross-stock predictability is sparse and short-lived. At the shortest
85 +horizons measured comovement is attenuated mechanically. \citet{epps1979comovements}
86 +first documented that correlations among stocks fall toward zero as the
87 +sampling interval shrinks. Part of this ``Epps effect'' is a pure artifact of
88 +nonsynchronous trading and price discreteness
89 +\citep{scholes1977estimating,lo1990nonsynchronous,harris1991stock}, which
90 +\citet{reno2003closer} shows can reproduce the decay by simulation and which
91 +estimators built on actual transaction times remove
92 +\citep{hayashi2005covariance,zhang2011estimating}; microstructure noise
93 +further contaminates high-frequency second moments
94 +\citep{hansen2006realized}. Part is genuine delay in the incorporation of
95 +common information. \citet{toth2009epps} document that the Epps curve among
96 +NYSE stocks flattened steadily between 1993 and 2003 and argue that denser
97 +trading explains only part of the change. Our Epps evidence extends this
98 +series to 2024 on previous-tick synchronized grids and must be read with
99 +both mechanisms in mind; we return to the point in
100 +Section~\ref{sec:results}. The same mechanics matter for our design, because
101 +neighbor sets are identified from one-minute residual correlations and are
102 +therefore identified less precisely in eras where the Epps effect was strong.
113 103
114 −\subsection{Networks: real linkages, financial linkages, and measured connectedness}
104 +\subsection{Networks, connectedness, and contagion measurement}
115 105
116 −Network theory shows when idiosyncratic shocks fail to diversify away:
117 −sufficiently asymmetric input--output structures let micro shocks aggregate
118 −\citep{acemoglu2012network}, and intermediate integration and diversification
119 −maximize cascade susceptibility in financial cross-holdings
120 −\citep{elliott2014financial}; \citet{glasserman2016contagion} survey the
121 −contagion-network literature. Real-economy links propagate firm-level shocks
122 −measurably---natural-disaster shocks to suppliers impose large value losses on
123 −customers \citep{barrot2016input}---and network positions carry asset-pricing
124 −implications \citep{herskovic2018networks}. On the measurement side,
125 −connectedness is estimated from variance decompositions
126 −\citep{diebold2012better,diebold2014network}, Granger-causality graphs among
127 −financial institutions \citep{billio2012econometric}, sparse VARs
128 −\citep{barigozzi2019nets}, and tail-risk networks \citep{hautsch2015financial}.
129 −The contagion-measurement tradition warns that unconditional correlation
130 −comparisons are biased in volatile regimes \citep{forbes2002no}; defining
131 −contagion as comovement beyond fundamentals, \citet{bekaert2014global} find
132 −``wake-up call'' transmission tied to weak fundamentals. Classic
133 −volatility-spillover evidence---meteor showers in FX
134 −\citep{engle1990meteor}, international open/close spillovers
135 −\citep{hamao1990correlations}, and cross-market volatility transmission
136 −\citep{king1990transmission}---anticipates our design at coarser resolution
137 −and for a handful of markets. Across all of this work the networks are
106 +Network theory shows when idiosyncratic shocks fail to diversify away
107 +\citep{acemoglu2012network,elliott2014financial}; \citet{glasserman2016contagion}
108 +survey the contagion-network literature; real-economy links propagate
109 +firm-level shocks measurably \citep{barrot2016input} and network positions
110 +carry asset-pricing implications \citep{herskovic2018networks}. On the
111 +measurement side, connectedness is estimated from variance decompositions
112 +\citep{diebold2012better,diebold2014network}, Granger-causality graphs
113 +\citep{billio2012econometric}, sparse VARs \citep{barigozzi2019nets}, and
114 +tail-risk networks \citep{hautsch2015financial}. The contagion-measurement
115 +tradition warns that unconditional correlation comparisons are biased in
116 +volatile regimes \citep{forbes2002no} and defines contagion as comovement
117 +beyond fundamentals \citep{bekaert2014global}; classic volatility-spillover
118 +evidence \citep{engle1990meteor,hamao1990correlations} anticipates our design
119 +at coarser resolution and for a handful of markets. These networks are
138 120 estimated at daily or lower frequency on small panels. We estimate pre-shock
139 −networks on minute data for thousands of securities and use them as lagged
140 −conditioning information---asking whether network position predicts
141 −transmitters, receivers, amplifiers, and absorbers---with permuted-network
142 −placebos in the spirit of the falsification tradition.
121 +correlation neighborhoods on minute data for the five hundred most active
122 +securities each month and use them as lagged conditioning information, with
123 +the same neighborhood contrast computed on placebo events as the benchmark.
143 124
144 125 \subsection{Market-structure evolution and the high-frequency trading debate}
145 126
146 −Our sample spans essentially the entire modern transformation of U.S.\ equity
147 −trading, and the literature documents each step. Decimalization in 2001 cut
148 −spreads, particularly for active stocks \citep{bessembinder2003trade};
149 −liquidity itself promotes price efficiency \citep{chordia2008liquidity};
150 −turnover exploded in the 2000s alongside algorithmic trading
151 −\citep{chordia2011recent}; the NYSE's 2006--07 Hybrid automation made prices
152 −more efficient while raising the cost of immediacy
153 −\citep{hendershott2011automation}; and fragmentation across venues did not
154 −measurably harm quality \citep{ohara2011fragmentation}. On algorithmic and
155 −high-frequency trading, the normal-times evidence is broadly positive:
156 −algorithmic trading narrows spreads and improves price discovery
157 −\citep{hendershott2011algorithmic,brogaard2014high}, modern HFT market makers
158 −halve spreads on entry \citep{menkveld2013high}, low-latency activity improves
159 −short-horizon quality \citep{hasbrouck2013low}, machines improve FX efficiency
160 −\citep{chaboud2014rise}, and colocation-based evidence replicates
161 −internationally, with higher short-term volatility as the caveat
162 −\citep{boehmer2021algorithmic}; \citet{menkveld2016economics} surveys the
163 −channel-by-channel evidence. Theory tempers the optimism: speed investment can
164 −be socially excessive \citep{biais2015equilibrium}, fast news trading
165 −concentrates volume and short-run price correlation \citep{foucault2016news},
166 −toxic arbitrage degrades liquidity \citep{foucault2017toxic}, the continuous
167 −limit order book hard-wires a latency arms race
168 −\citep{budish2015high}---measured directly in message data as races won in
169 −microseconds, a fifth of volume \citep{aquilina2022quantifying}---and speed can
170 −crowd out information acquisition \citep{weller2018does,baldauf2020high}, with
171 −HFTs eventually trading \emph{with} informed institutional orders
172 −\citep{vankervel2019high}. \citet{ohara2015high} argues the machine age changes
173 −what microstructure concepts mean. This literature motivates our central
174 −hypothesis---normal-times absorption improving while tail transmission
175 −accelerates---but it measures liquidity levels, price discovery, or single
176 −events. We contribute the missing time series: propagation speed and breadth of
177 −extreme shocks, measured uniformly across three decades of structural change,
178 −with flexible year effects rather than imposed trend breaks.
127 +Our sample spans the modern transformation of U.S.\ equity trading.
128 +Decimalization in 2001 cut spreads \citep{bessembinder2003trade}; liquidity
129 +promotes efficiency \citep{chordia2008liquidity}; turnover exploded in the
130 +2000s alongside algorithmic trading \citep{chordia2011recent}; NYSE automation
131 +made prices more efficient while raising the cost of immediacy
132 +\citep{hendershott2011automation}; and fragmentation did not measurably harm
133 +quality \citep{ohara2011fragmentation}. The normal-times evidence on
134 +algorithmic and high-frequency trading is broadly positive
135 +\citep{hendershott2011algorithmic,brogaard2014high,menkveld2013high,
136 +hasbrouck2013low,chaboud2014rise,boehmer2021algorithmic}, with
137 +\citet{menkveld2016economics} surveying the channel-by-channel evidence, and
138 +markets absorb even large predictable trades with limited lasting impact
139 +\citep{bessembinder2016liquidity}. Theory tempers the optimism: speed
140 +investment can be socially excessive \citep{biais2015equilibrium}, fast news
141 +trading concentrates volume and short-run correlation
142 +\citep{foucault2016news}, toxic arbitrage degrades liquidity
143 +\citep{foucault2017toxic}, the continuous limit order book hard-wires a
144 +latency arms race \citep{budish2015high,aquilina2022quantifying}, and speed
145 +can crowd out information acquisition \citep{weller2018does,baldauf2020high},
146 +with fast traders eventually trading with informed institutional orders
147 +\citep{vankervel2019high}; \citet{ohara2015high} argues the machine age
148 +changes what microstructure concepts mean. This literature motivates our
149 +central hypothesis but measures liquidity levels, price discovery, or single
150 +events. We contribute the missing time series: propagation breadth and
151 +network concentration of extreme shocks measured uniformly across three
152 +decades of structural change, with year-by-year estimates rather than
153 +imposed trend breaks.
179 154
180 155 \subsection{Fragility and flash events}
181 156
182 −Episodic evidence shows modern liquidity can vanish abruptly. On May 6, 2010,
183 −high-frequency traders did not trigger the crash but amplified selling once it
184 −began \citep{kirilenko2017flash}; cross-market arbitrage broke down, leaving
185 −the seller dependent on local liquidity \citep{menkveld2019flash}; and
186 −fragmentation shaped which ETFs were hit \citep{madhavan2012exchange}. More
187 −systematically, HFTs supply liquidity in single-stock extreme moves but their
188 −provision weakens when \emph{many} stocks move together
189 −\citep{brogaard2018high}---precisely the margin our breadth measure tracks.
190 −Proposed early-warning metrics remain contested: VPIN signaled deteriorating
191 −conditions before the Flash Crash \citep{easley2012flow}, but its power appears
192 −mechanically tied to volatility and trading intensity
157 +Episodic evidence shows that modern liquidity can vanish abruptly. On May~6,
158 +2010, high-frequency traders did not trigger the crash but amplified selling
159 +once it began \citep{kirilenko2017flash}; cross-market arbitrage broke down
160 +\citep{menkveld2019flash}; and fragmentation shaped which ETFs were hit
161 +\citep{madhavan2012exchange}. Sub-second ``fractures'' in individual stocks
162 +proliferated as machine ecologies moved beyond human response times
163 +\citep{johnson2013abrupt}. More systematically, high-frequency traders supply
164 +liquidity in single-stock extreme moves but their provision weakens when many
165 +stocks move together \citep{brogaard2018high}---precisely the margin our
166 +breadth measure tracks. Proposed early-warning metrics remain contested: VPIN
167 +signaled deteriorating conditions before the Flash Crash
168 +\citep{easley2012flow}, but its power appears mechanically tied to volatility
193 169 \citep{andersen2014vpin}. This debate cautions against fragility indicators
194 170 that repackage volatility; our propagation measures are therefore defined
195 −relative to lagged conditional distributions and validated against matched
196 −placebo times.
171 +relative to lagged conditional distributions and benchmarked against matched
172 +placebo minutes.
197 173
198 174 \subsection{ETFs, indexing, and passive intermediation}
199 175
200 176 A structural change specific to the later sample is the growth of index-linked
201 177 intermediation. ETF ownership raises non-fundamental volatility of underlying
202 178 stocks through the arbitrage channel \citep{bendavid2018etfs} and contributes
203 −to return comovement, especially in small and illiquid names
204 −\citep{da2018exchange}; index-product arbitrage flipped index return serial
205 −dependence from positive to negative across twenty markets
206 −\citep{baltussen2019indexing}. Effects on efficiency are contested: ETF
207 −ownership is associated with weaker firm-specific information environments
208 −\citep{israeli2017dark}, yet ETF activity speeds the incorporation of
209 −systematic earnings information where information environments are weak
210 −\citep{glosten2021etf}, and Russell-reconstitution evidence suggests price
211 −efficiency survives passive growth because the remaining informed investors
212 −adjust \citep{coles2022index}. For us, the ETF layer is both a candidate
213 −mechanism for the \emph{widening} of propagation (basket arbitrage converts
214 −idiosyncratic shocks into correlated flows) and a required robustness screen:
215 −no headline result may be driven by ETFs themselves.
179 +to return comovement \citep{da2018exchange}; index-product arbitrage flipped
180 +index return serial dependence from positive to negative across twenty markets
181 +\citep{baltussen2019indexing}. Effects on efficiency are contested
182 +\citep{israeli2017dark,glosten2021etf,coles2022index}. For us, the ETF layer
183 +is a candidate mechanism for the sharpening of propagation along correlation
184 +neighborhoods, since basket arbitrage converts idiosyncratic shocks into
185 +correlated flows. We cannot test it directly: the data carry no security-type
186 +labels, so ETFs are neither identified nor excluded, a limitation we state
187 +rather than paper over.
216 188
217 189 \subsection{Jumps, co-jumps, and scheduled announcements}
218 190
219 −Our shock measurement builds directly on high-frequency econometrics. Intraday
191 +Our shock measurement builds on high-frequency econometrics. Intraday
220 192 volatility periodicity must be filtered before anything else
221 −\citep{andersen1997intraday}; bipower variation separates jumps from diffusion
222 −\citep{barndorffnielsen2004power}; formal jump tests exist at the observation
223 −level \citep{lee2008jumps,aitsahalia2009testing}. Two cautions discipline our
224 −thresholds: at ultra-high frequency much of what coarser methods call jumps is
225 −friction or ``drift bursts'' \citep{christensen2014fact}, and multiple-testing
226 −corrections eliminate a large share of spurious detections, with true jumps
227 −rarer and news-related \citep{bajgrowicz2016jumps}. The co-jump literature is
228 −the nearest empirical neighbor to our propagation object: index-level jumps are
229 −systematic co-jumps linked to macro news while idiosyncratic jumps diversify
230 −away \citep{bollerslev2008risk}; simultaneous multivariate jumps are frequent,
231 −associated with liquidity dry-ups, and priced \citep{caporin2017systemic};
232 −jumps coincide with liquidity-demand shocks and news
233 −\citep{boudt2014intraday}; and simultaneous jumps across countries carry
234 −portfolio consequences \citep{das2004systemic}. Self- and cross-excitation of
235 −jumps has a canonical language in Hawkes processes
236 −\citep{hawkes1971spectra,bacry2015hawkes}, estimated on international indices
237 −by \citet{aitsahalia2015modeling}. Scheduled macro announcements provide the
238 −identified common shocks against which idiosyncratic propagation must be
239 −benchmarked: announcement surprises produce jump-like conditional-mean
240 −responses in FX and global markets \citep{andersen2003micro,andersen2007real},
241 −generate the cojumps documented by \citet{lahaye2011jumps}, and structure
242 −expected returns around announcement times
243 −\citep{savor2013how,lucca2015pre}. Accordingly, our baseline excludes
244 −announcement windows and treats announcement-window propagation as a separate,
245 −labeled object. Where the co-jump literature tests for \emph{simultaneity}, our
246 −measures trace the \emph{sequence}---breadth and intensity at one through 120
247 −minutes---and its secular change.
193 +\citep{andersen1997intraday}, and robust periodicity estimators are needed
194 +because the periodicity estimate is itself contaminated by the jumps one
195 +wishes to detect \citep{boudt2011robust}; our minute-of-day factor is a
196 +cross-sectional median for that reason. Bipower variation separates jumps from
197 +diffusion \citep{barndorffnielsen2004power}, formal jump tests exist at the
198 +observation level \citep{lee2008jumps,aitsahalia2009testing}, and jump tails
199 +carry extreme cross-sectional dependence \citep{bollerslev2013jump}. Two
200 +cautions discipline our thresholds: at ultra-high frequency much of what
201 +coarser methods call jumps is friction \citep{christensen2014fact} or a
202 +``drift burst'' extending over several minutes \citep{christensen2022drift},
203 +and multiple-testing corrections eliminate a large share of spurious
204 +detections \citep{bajgrowicz2016jumps}. Our events are threshold exceedances
205 +of standardized residuals, not test-based jump detections, and we use the
206 +term ``extreme move'' rather than ``jump'' wherever the distinction matters.
248 207
249 −\subsection{High-frequency measurement: factor structure, OHLCV proxies, and machine learning}
208 +The co-jump literature is the nearest empirical neighbor to our propagation
209 +object. Index-level jumps are systematic co-jumps linked to macro news while
210 +idiosyncratic jumps diversify away \citep{bollerslev2008risk}; formal tests
211 +for common jump arrivals exist \citep{jacod2009testing}; co-jumps among
212 +large U.S.\ stocks are frequent and concentrated around announcements
213 +\citep{gilder2014cojumps}; simultaneous multivariate jumps are associated with
214 +liquidity dry-ups and priced \citep{caporin2017systemic}; jumps coincide with
215 +liquidity-demand shocks and news \citep{boudt2014intraday}; and simultaneous
216 +jumps across countries carry portfolio consequences \citep{das2004systemic}.
217 +Closest to our co-jump series, \citet{bormetti2015modelling} model systemic
218 +price co-jumps among many stocks with Hawkes factor processes and
219 +\citet{calcagnile2018collective} document that the number of stocks jumping
220 +in the same minute is far larger than independence or a single factor would
221 +imply, with collective synchronization events increasing in frequency in the
222 +high-frequency era. Self- and cross-excitation of jumps has a canonical
223 +language in Hawkes processes \citep{hawkes1971spectra,bacry2015hawkes},
224 +estimated on international indices by \citet{aitsahalia2015modeling} and
225 +tested at high frequency by \citet{dungey2018testing}. Scheduled macro
226 +announcements provide the identified common shocks against which
227 +idiosyncratic propagation must be benchmarked: announcement surprises produce
228 +jump-like responses \citep{andersen2003micro,andersen2007real}, generate
229 +co-jumps \citep{lahaye2011jumps}, structure volume and volatility
230 +\citep{bollerslev2018volume}, and structure expected returns around
231 +announcement times \citep{savor2013how,lucca2015pre}. Our robustness battery
232 +accordingly excludes the main announcement windows. Where the co-jump
233 +literature tests for simultaneity, our propagation measures trace the
234 +sequence---breadth at one through 120 minutes---and our co-jump series
235 +traces the secular evolution of simultaneity itself.
250 236
251 −Separating idiosyncratic from systematic movement at the minute scale requires
252 −a factor structure: common factors exist in high-frequency returns and order
253 −flow \citep{hasbrouck2001common}, high-frequency PCA is consistent with
254 −low-frequency covariance structure \citep{aitsahalia2019principal}, latent
255 −continuous and jump factors are estimable and economically interpretable
256 −\citep{pelger2020understanding}, and betas decompose into continuous, jump, and
257 −overnight components with distinct pricing \citep{bollerslev2016roughing};
258 −signed semicovariances refine the decomposition \citep{bollerslev2020realized}.
259 −Jump-tail risk itself commands a large, time-varying premium
260 −\citep{bollerslev2011tails}, which is why our propagation outcomes separate
261 −tail responses from ordinary ones.
262 −Because our data are bars rather than quotes, liquidity enters only through
263 −proxies with a long pedigree: spread estimators from return autocovariance
264 −\citep{roll1984simple} and high--low ranges \citep{corwin2012simple},
265 −range-based volatility \citep{parkinson1980extreme,garman1980estimation},
266 −price-impact ratios \citep{amihud2002illiquidity}, and trade-based information
267 −measures \citep{hasbrouck1991measuring}; \citet{goyenko2009liquidity} validate
268 −such proxies against TAQ benchmarks. For estimation we use local projections
269 −\citep{jorda2005estimation} for horizon-specific responses, and the
270 −machine-learning extension follows the out-of-sample discipline of
271 −\citet{gu2020empirical}, informed by evidence that microstructure features
272 −carry universal predictive content \citep{sirignano2019universal} and remain
273 −informative in the machine age \citep{easley2021microstructure}.
237 +\subsection{Factor structure at the minute scale}
238 +
239 +Separating idiosyncratic from systematic movement requires a factor
240 +structure: common factors exist in high-frequency returns and order flow
241 +\citep{hasbrouck2001common}, high-frequency principal components are
242 +consistent with low-frequency covariance structure
243 +\citep{aitsahalia2019principal}, latent continuous and jump factors are
244 +estimable \citep{pelger2020understanding}, and betas decompose into
245 +continuous, jump, and overnight components with distinct pricing
246 +\citep{bollerslev2016roughing,bollerslev2020realized}; jump-tail risk commands
247 +a large, time-varying premium \citep{bollerslev2011tails}. We deliberately use
248 +a single, strictly lagged market loading per stock-month rather than these
249 +richer structures, because any loading estimated on the event month would
250 +overfit the very minutes it is used to classify; the cost, a coarser
251 +residual, is absorbed by the placebo contrast.
274 252
275 253 \subsection{Positioning and contribution}
276 254
277 −Each literature above contributes a piece: theory predicts state-dependent,
255 +Each literature contributes a piece: theory predicts state-dependent,
278 256 asymmetric transmission; the flows literature demonstrates non-fundamental
279 257 propagation at monthly resolution; the network literature measures
280 258 connectedness on small panels at daily frequency; the market-structure
281 259 literature documents the transformation of trading but measures liquidity
282 260 levels and single events; the co-jump literature tests simultaneity at high
283 261 frequency over short samples. To our knowledge---based on a documented search
284 −dated 2026-07-13 (replication package)---no prior study combines (i)
285 −minute-level resolution, (ii) the full U.S.\ equity cross-section including
286 −delisted securities, and (iii) a 1992--2026 sample spanning floor trading,
287 −decimalization, Reg~NMS fragmentation, and ETF-intermediated, algorithmic
288 −markets. The joint analysis of propagation speed, breadth, persistence,
289 −asymmetry, secular evolution, and network conditioning at that resolution and
290 −length---with placebo-timing, permuted-network, and announcement-exclusion
291 −falsifications---is the contribution.
262 +dated 2026-07-13 and repeated on 2026-10-17---no prior study combines
263 +minute-level resolution, the full U.S.\ equity cross-section including
264 +delisted securities, and a 1992--2026 sample spanning floor trading,
265 +decimalization, Reg~NMS fragmentation, and ETF-intermediated algorithmic
266 +markets. The joint analysis of market-wide and network-concentrated
267 +propagation, its secular evolution, and the evolution of synchronized extreme
268 +minutes at that resolution and length---with placebo-timing, placebo-network,
269 +announcement-exclusion, and tick-regime falsifications---is the contribution.
modified paper/sections/03_data.tex +81 −88
@@ -1,8 +1,8 @@
1 1 \section{Data}
2 2 \label{sec:data}
3 3
4 −% Every statement below is verified in docs/DATA_AUDIT.md (Stage A pilot audit,
5 −% 2026-07-13). Full-sample statements will be added after the Stage C audit.
4 +% Every statement below is verified in docs/DATA_AUDIT.md (Stage A pilot
5 +% audit and Stage C full-sample audit, 2026-07-13).
6 6
7 7 \subsection{Source and structure}
8 8
@@ -10,87 +10,67 @@ Our data are one-minute OHLCV bars for U.S.-listed securities originally
10 10 sourced from Finnhub.io (\url{https://finnhub.io}), a market data provider,
11 11 and aggregated from the provider's monthly archives into 411 monthly Parquet
12 12 files covering January 1992 through March 2026 (87.7~GB compressed) with no
13 −missing months. We access the aggregation through a public archive whose exact byte-level
14 −snapshot is pinned and checksum-verified in the replication package, fixing
15 −the data used throughout for reproducibility.
16 −Each row records the open, high, low, and close price (U.S.\ dollars, unadjusted
17 −for corporate actions), share volume, and ticker for one minute of trading in one
18 −security. Timestamps are UTC and denote the start of the bar minute; we verify the
19 −convention empirically by mapping bar times to the Eastern-time trading session in
13 +missing months. We access the aggregation through a public archive whose exact
14 +byte-level snapshot is pinned and checksum-verified in the replication
15 +package, which fixes the data used throughout. The archive carries no explicit
16 +licence; we use it for research with attribution and do not redistribute the
17 +minute data, only the aggregates, event-level estimates, and code. Each row
18 +records the open, high, low, and close price (U.S.\ dollars, unadjusted for
19 +corporate actions), share volume, and ticker for one minute of trading in one
20 +security. Timestamps are UTC and denote the start of the bar minute; we verify
21 +the convention empirically by mapping bar times to the Eastern-time session in
20 22 months on both sides of daylight-saving transitions.
21 23
22 24 A bar exists only for minutes with at least one reported trade: there are no
23 25 zero-volume rows in the source. Missing minutes are therefore informative about
24 −trading activity, and we treat them explicitly rather than filling a regular grid.
25 −
26 −\subsection{Pilot audit}
27 −
28 −Before downloading the archive we audited three pilot months (June 1992, June
29 −2009, December 2025; 58.8~million bars). Trading-day counts match the U.S.\
30 −exchange calendar in all three months, including the December~24, 2025 early
31 −close. No OHLC-consistency violations ($L_{i,t} > \min(O_{i,t},C_{i,t})$,
32 −$H_{i,t} < \max(O_{i,t},C_{i,t})$, or $L_{i,t}>H_{i,t}$), null values, negative
33 −volumes, or zero-volume bars occur in the pilot sample; 14 duplicate
34 −ticker--minute pairs (all in inactive over-the-counter names) appear in December
35 −2025. Liquid securities are densely covered throughout: the four most active
36 −tickers in December 2025 have complete 390-bar regular sessions, and large
37 −issuers already average 277--369 bars per day in June 1992.
38 −
39 −Coverage widens over time, from roughly 8,300 symbols per month in 1992 to about
40 −23,100 in late 2025; much of the recent expansion reflects over-the-counter
41 −instruments rather than listed common equity, and our sample filters hold the
42 −universe definition fixed using lagged dollar-volume ranks. Only 22--26\% of June
43 −1992 tickers appear in the later pilot months, indicating that delisted
44 −securities are retained in the historical files.
45 −
46 −Figure~\ref{fig:pilot_coverage} summarizes raw data volume and per-day symbol
47 −counts; Figure~\ref{fig:pilot_seasonality} documents the familiar intraday
48 −patterns in absolute returns, volume, and range. Table~\ref{tab:pilot_summary}
49 −reports pilot summary statistics for the 500 highest dollar-volume securities in
50 −each pilot month.
51 −
52 −\begin{figure}[t]
53 −\centering
54 −\includegraphics[width=\textwidth]{figP1_coverage.pdf}
55 −\caption{\textbf{Dataset coverage.} Panel (a): compressed size of each monthly
56 −source file, 1992--2026 (all 411 files of the pinned snapshot). Panel
57 −(b): number of symbols with at least one bar per trading day in the three pilot
58 −months.}
59 −\label{fig:pilot_coverage}
60 −\end{figure}
61 −
62 −\begin{figure}[t]
63 −\centering
64 −\includegraphics[width=\textwidth]{figP2_seasonality.pdf}
65 −\caption{\textbf{Intraday seasonality, pilot months.} Minute-of-day means during
66 −regular trading hours for the top 500 securities by within-month dollar volume:
67 −(a) absolute one-minute log return in basis points; (b) share of the stock-day's
68 −volume; (c) high--low log range in basis points. Descriptive full-month
69 −normalization; no corporate-action screens.}
70 −\label{fig:pilot_seasonality}
71 −\end{figure}
72 −
73 −\input{tables/tabP1_pilot_summary}
74 −
75 −\subsection{Full-sample audit and clean panel}
76 −
77 −The full archive contains \DataRows{} bars. Observed trading dates match the
26 +trading activity, and we treat them explicitly rather than filling a regular
27 +grid. Two properties of the source cannot be established from the data and are
28 +carried as limitations. The archive does not document whether bars aggregate
29 +consolidated-tape trades or a single venue; the presence of sparse low-volume
30 +prints after the 13:00 close on early-closing days suggests off-exchange
31 +trades are included, but volumes could not be reconciled against a
32 +consolidated benchmark for more than a few securities. And securities are
33 +identified by ticker alone, without a permanent identifier or security-type
34 +label, so that ETFs, ADRs, preferred shares, and warrants are present and can
35 +be screened only heuristically.
36 +
37 +\subsection{Full-sample audit}
38 +
39 +The archive contains \DataRows{} bars. Observed trading dates match the
78 40 official XNYS calendar exactly---zero missing sessions and zero dates outside
79 41 the calendar over thirty-four years---and the regular-hours share of bars is
80 −flat across calendar months, ruling out any daylight-saving mishandling.
42 +flat across calendar months, which rules out any daylight-saving mishandling
43 +(an offset error would displace roughly 15\% of session rows in March and
44 +November). Early-closing sessions are truncated at the true exchange close.
81 45 Integrity failures are negligible: \DataOHLCViol{} impossible OHLC bars
82 −(concentrated in 1993--1995) and \DataDupKeys{} duplicate ticker-minute keys
83 −(1.3 per million), which we merge under a logged deterministic rule. The panel
84 −spans \DataTickers{} distinct tickers forming \DataSpells{} continuous
85 −spells; \DataMultiSpellTickers{} tickers have multiple spells, so spells, not
86 −tickers, are the identity unit. Entries and exits occur continuously
87 −(delisted securities are present throughout), with one discrete
88 −source-coverage expansion in 2020 that our fixed-rule universes absorb
89 −(Figure~\ref{fig:coverage}). A conservative overnight-ratio screen detects
90 −\NSplitEvents{} corporate-action events (\NSplitFwd{} splits,
91 −\NSplitRev{} reverse splits); event days $\pm1$ trading day are excluded
92 −from all return-based analyses. Table~\ref{tab:sample_construction} tracks
93 −the sample from raw bars to the regular-hours analysis panel.
46 +(concentrated in 1993--1995, flagged and excluded) and \DataDupKeys{}
47 +duplicate ticker--minute keys (1.3 per million), which we merge under a logged
48 +deterministic rule. No bar has a non-positive price or a zero or negative
49 +volume. A pilot audit of three months (June 1992, June 2009, December 2025)
50 +conducted before the full download, summarized in the Online Appendix,
51 +established the same facts on 58.8~million bars and verified volume units
52 +against public daily volumes for benchmark securities.
53 +
54 +The panel spans \DataTickers{} distinct tickers forming \DataSpells{}
55 +continuous spells, a spell being a run of trading days without a gap longer
56 +than thirty calendar days; \DataMultiSpellTickers{} tickers have multiple
57 +spells, so spells, not tickers, are the identity unit of every statistic
58 +that follows a security through time. Entries and exits occur continuously,
59 +with exits running between 800 and 1,600 per year: delisted securities are
60 +present throughout, and only 22--26\% of the tickers active in June 1992
61 +appear in 2009 or 2025. The data are therefore not a backfilled survivor
62 +universe, although we cannot verify that every delisting is captured to its
63 +final trading day. There is one discrete source-coverage expansion in 2020
64 +(about 11,000 entries, largely over-the-counter names) that fixed-rule lagged
65 +universes absorb (Figure~\ref{fig:coverage}). A conservative overnight-ratio
66 +screen (canonical split ratios within $\pm1.5\%$, both-day dollar volume of
67 +at least \$50,000, prior close of at least \$1) detects \NSplitEvents{}
68 +corporate-action events (\NSplitFwd{} splits, \NSplitRev{} reverse splits);
69 +event days $\pm1$ trading day are excluded from all return-based analyses.
70 +Dividends and other distributions are not detected; their overnight effect
71 +never enters an intraday return because returns are computed within the
72 +trading day only. Table~\ref{tab:sample_construction} tracks the sample from
73 +raw bars to the regular-hours analysis panel.
94 74
95 75 \input{tables/table1_sample_construction}
96 76
@@ -104,17 +84,30 @@ top-1000 universe.}
104 84 \label{fig:coverage}
105 85 \end{figure}
106 86
107 −\subsection{Known limitations}
87 +\subsection{Universes}
108 88
109 −The archive contains no quotes, spreads, depth, or trade direction; all liquidity
110 −statements in this paper refer to OHLCV-based proxies. Securities are identified
111 −by ticker only, prices are unadjusted, and security types are unlabeled; the
112 −corporate-action detection, ticker-spell construction, and universe
113 −classification underlying the final sample are described in
114 −Section~\ref{sec:measurement} and the Internet Appendix. In the pilot months the
115 −standard deviation of unscreened minute returns is an order of magnitude larger
116 −than the 1st--99th percentile spread, reflecting split-like jumps and isolated
117 −bad prints that the cleaning stage must remove.
89 +Coverage widens from roughly 8,300 symbols per month in 1992 to about 23,100
90 +in late 2025, and much of the recent expansion reflects over-the-counter
91 +instruments rather than listed common equity. All analyses therefore operate
92 +on universes defined by strictly lagged rules: a security is eligible in
93 +month $m$ if in month $m{-}1$ it traded on at least ten days with a median
94 +close of at least \$5 and a median of at least 100 regular-hours bars per
95 +day, and the top-500, top-1000, and top-3000 universes rank eligible
96 +securities by month-$(m{-}1)$ regular-hours dollar volume. Membership never
97 +uses information from month $m$. A ticker-shape heuristic (fifth-letter
98 +suffixes and warrant, unit, and preferred suffixes) flags likely non-common
99 +instruments; it does not identify ETFs, which trade under ordinary tickers.
100 +
101 +\subsection{Known limitations}
118 102
119 −% [Full-sample statistics, sample-construction table (Table I), and survivorship
120 −% curves to be added after Stages B–D.]
103 +The archive contains no quotes, spreads, depth, or trade direction; all
104 +statements about liquidity in this paper rest on dollar volume and the
105 +high--low range. Security types are unlabeled, prices are unadjusted, venue
106 +scope is undocumented, and the licence is unspecified. Pre-2001 minute
107 +returns are discrete at eighths and sixteenths, and roughly half of the
108 +one-minute returns of top-1000 securities in the mid-1990s are exactly zero
109 +(\ZeroRetNinetySix\% in 1996 against \ZeroRetTwentyFour\% in 2024,
110 +Online Appendix); any second-moment or tail statistic compared across the
111 +decimalization boundary inherits this discreteness. We flag the tick-regime
112 +caveat wherever it binds and anchor every secular comparison on a
113 +within-era placebo contrast.
modified paper/sections/04_measurement.tex +148 −115
@@ -16,28 +16,32 @@ rates) are known before any economic claim rests on it.
16 16
17 17 The unit of observation is a one-minute bar during the regular session
18 18 (09:30--16:00 ET, truncated at the true exchange close on early-closing days).
19 −Let $P_{i,t}$ be the last trade price of security $i$ in minute $t$.
20 −One-minute log returns,
19 +Let $P_{i,t}$ be the last trade price of security $i$ in minute $t$. One-minute
20 +log returns,
21 21 \begin{equation}
22 22 r_{i,t} \;=\; \log P_{i,t} - \log P_{i,t-1},
23 +\label{eq:return}
23 24 \end{equation}
24 25 are computed within ticker and trading day only: the first observed minute of
25 26 each stock-day has no return, and overnight intervals are never bridged, so no
26 27 intraday statistic mixes overnight information. Because the source records a
27 28 bar only when at least one trade occurs, a missing minute means ``no trade,''
28 −not ``zero return''; returns are accordingly defined only across consecutively
29 −observed minutes and missingness is carried, never imputed. Bars flagged for
30 −OHLC inconsistency and all bars inside detected corporate-action windows
29 +not ``zero return''; returns are defined only across consecutively observed
30 +minutes and missingness is carried, never imputed. Bars flagged for OHLC
31 +inconsistency and all bars inside detected corporate-action windows
31 32 (Section~\ref{sec:data}) are excluded before any return is computed.
32 33
33 34 All shock-level analysis operates on the lagged top-500 dollar-volume
34 35 universe. This choice trades breadth for interpretability: top-500 names trade
35 −essentially every minute (median 382--390 bars per day from the mid-1990s
36 −onward), so the cross-sectional response measures defined below have stable,
36 +in nearly every minute from the mid-1990s onward (median 382--390 bars per
37 +day), so the cross-sectional response measures defined below have stable,
37 38 nearly complete denominators, and the asynchronicity that depresses
38 −high-frequency comovement among thinner names \citep{epps1979comovements}
39 −cannot masquerade as absent propagation. Results for top-1000 and top-3000
40 −universes appear in the robustness materials.
39 +high-frequency comovement among thinner names
40 +\citep{epps1979comovements,lo1990nonsynchronous} cannot masquerade as absent
41 +propagation. Even so, the number of receivers with an observable return at a
42 +given minute averages \NEligNineties{} in the 1990s and about
43 +\NEligMean{} over the full sample, and every share below is computed over
44 +observed receivers only.
41 45
42 46 \subsection{Residualization}
43 47 \label{sec:residualization}
@@ -49,85 +53,93 @@ contaminations.
49 53
50 54 \emph{Leave-one-out factor.} The market return at minute $t$, from the
51 55 perspective of security $i$, is the equal-weighted mean return of the universe
52 −\emph{excluding} $i$:
56 +excluding $i$,
53 57 \begin{equation}
54 58 r^{(-i)}_{m,t} \;=\; \frac{1}{N_t-1}\sum_{j\ne i} r_{j,t},
59 +\label{eq:loo}
55 60 \end{equation}
56 −where $N_t$ counts securities with an observable return at $t$. Without the
57 −leave-one-out correction, a large own return mechanically moves the factor and
58 −attenuates the measured shock; with 500 names the correction is small but we
59 −apply it exactly.
61 +where $N_t$ counts securities with an observable return at $t$ and $r_{m,t}$
62 +without the superscript denotes the full equal-weighted mean. Without the
63 +leave-one-out correction a large own return mechanically moves the factor and
64 +attenuates the measured shock; with several hundred names the correction is
65 +small but we apply it exactly.
60 66
61 67 \emph{Lagged loadings.} The loading $\beta_i$ is the OLS slope of $r_{i,t}$ on
62 −the equal-weighted market return, estimated on all minutes of month $m{-}1$ in
63 −which the security has at least 300 valid return observations and the market
64 −aggregate at least 50 contributing names; securities with insufficient history
68 +$r_{m,t}$ over all minutes of month $m{-}1$ in which the market aggregate has
69 +at least 50 contributing names, provided security $i$ has at least 300 valid
70 +return observations in that month; securities with insufficient history
65 71 receive $\beta_i=1$, and estimates are winsorized to $[-3,5]$. Residuals in
66 72 month $m$ are then
67 73 \begin{equation}
68 74 e_{i,t} \;=\; r_{i,t} - \beta_i\, r^{(-i)}_{m,t}.
69 75 \label{eq:residual}
70 76 \end{equation}
71 −A single lagged loading per month is deliberately coarse: it cannot overfit
72 −the event minutes it will later be used to classify, and any misspecification
73 −it induces is absorbed by the placebo differencing described in
74 −Section~\ref{sec:design}.
77 +A single lagged loading per stock-month is deliberately coarse: it cannot
78 +overfit the event minutes it will later be used to classify, and any
79 +misspecification it induces---including the downward bias in betas estimated
80 +from nonsynchronous minute returns \citep{scholes1977estimating}---is common
81 +to real and placebo minutes of the same stock-month and is absorbed by the
82 +placebo contrast of Section~\ref{sec:design}.
75 83
76 84 \emph{Robust, seasonally adjusted scale.} Minute returns are strongly
77 85 leptokurtic and their dispersion follows the familiar intraday U-shape
78 86 \citep{andersen1997intraday}; a sample standard deviation would be dominated
79 87 by the very tail events we wish to detect, and an unadjusted scale would
80 −over-detect shocks at the open. We therefore standardize by a
81 −median-absolute-deviation scale interacted with a minute-of-day factor, both
82 −from month $m{-}1$:
88 +over-detect shocks at the open. Let $\mathrm{MAD}_i$ be the median absolute
89 +deviation of $e_{i,\cdot}$ from its median over month $m{-}1$, and let
90 +$u_{i,t}=e_{i,t}/(1.4826\,\mathrm{MAD}_i)$. The minute-of-day factor is the
91 +cross-sectional median of $|u|$ at each minute of day $\tau$ in month $m{-}1$,
92 +normalized to unit mean over the session,
83 93 \begin{equation}
84 −z_{i,t} \;=\; \frac{e_{i,t}}{1.4826\,\mathrm{MAD}_i \times f_{\tau(t)}},
94 +f_{\tau} \;=\; \frac{\operatorname{med}_{(i,t):\,\tau(t)=\tau}\,|u_{i,t}|}
95 +{\frac{1}{390}\sum_{\tau'}\operatorname{med}_{(i,t):\,\tau(t)=\tau'}\,|u_{i,t}|},
85 96 \qquad
86 −f_{\tau} \;=\; \frac{\mathrm{med}_i\,|e_{i,\cdot}/(1.4826\,\mathrm{MAD}_i)|_{\tau}}
87 −{\overline{\mathrm{med}}},
97 +z_{i,t} \;=\; \frac{e_{i,t}}{1.4826\,\mathrm{MAD}_i \, f_{\tau(t)}}.
88 98 \label{eq:zscore}
89 99 \end{equation}
90 −where $\tau(t)$ is the minute of day, the numerator of $f_\tau$ is the
91 −cross-sectional median absolute standardized residual at minute-of-day $\tau$
92 −in month $m{-}1$, and $\overline{\mathrm{med}}$ normalizes the profile to unit
93 −mean. The factor $1.4826$ makes the MAD consistent for the standard deviation
94 −under normality; under the actual fat-tailed distribution, $z$ is a ranking
95 −device whose thresholds are calibrated empirically, not a Gaussian quantile.
96 −The same construction applied to the market return, with the prior month's
97 −market MAD, yields the standardized market factor $z_{m,t}$ used in the
98 −isolation filters.
100 +The cross-sectional median makes the periodicity estimate robust to the jumps
101 +it is later used to detect, in the spirit of \citet{boudt2011robust}. The
102 +constant $1.4826$ makes the MAD consistent for the standard deviation under
103 +normality; under the actual fat-tailed distribution, $z$ is a ranking device
104 +whose thresholds are calibrated empirically, not a Gaussian quantile. A
105 +security whose prior-month residuals are exactly zero in more than half of
106 +the minutes has $\mathrm{MAD}_i=0$ and is excluded from the month, which acts
107 +as a staleness screen that binds mainly in the pre-decimalization years. The
108 +same construction applied to $r_{m,t}$, with the prior month's market MAD,
109 +yields the standardized market factor $z_{m,t}$ used in the isolation
110 +filters.
99 111
100 112 \subsection{Source shocks and isolation}
101 113 \label{sec:shockdef}
102 114
103 −A \emph{source shock} is an event $(i,t)$ with
104 −\begin{equation}
105 −|z_{i,t}| \;\ge\; 6,
106 −\end{equation}
107 −with the sign recorded and analyzed separately. The threshold is deliberately
108 −extreme: it selects on the order of one event per stock-week
109 −(\NEventsReal{} events over 410 months), deep enough into the tail that
115 +A \emph{source shock} is an event $(i,t)$ with $|z_{i,t}|\ge 6$, with the sign
116 +recorded and analyzed separately. The threshold is deliberately extreme: it
117 +selects \NEventsReal{} events over 410 months, about \EventsPerStockWeek{}
118 +per stock-week in the top-500 universe, deep enough into the tail that
110 119 bid--ask bounce alone cannot generate qualifying events for top-500 names
111 120 after 2001, while leaving enough events for precise inference within every
112 121 year. Nested thresholds ($|z|\ge8$, $|z|\ge12$) are reported in the
113 122 robustness table.
114 123
115 −Because the question concerns \emph{idiosyncratic} origins, two isolation
116 −filters apply at the event minute:
117 −\begin{enumerate}
118 −\item \textbf{Calm market:} $|z_{m,t}| < 1$. The event does not coincide with
119 −a market-factor movement.
120 −\item \textbf{No simultaneous wave:} the share of eligible securities with
121 −$|z_{j,t}|>4$ at minute $t$ is at most $2\%$, and at least 50 securities are
122 −observable. The event is not part of a market-wide burst.
123 −\end{enumerate}
124 −Both filters are measured, not assumed: minutes excluded by filter (2) form
125 −the systemic co-jump sample of Section~\ref{sec:cojumpdef}, so the design
126 −partitions all extreme activity into an idiosyncratic branch and a systemic
127 −branch rather than discarding the inconvenient one. The filters are
128 −intentionally conservative about common information; the cost, acknowledged
129 −throughout, is that any propagation operating \emph{through} an immediate
130 −market-wide reaction is assigned to the systemic branch by construction.
124 +Because the question concerns idiosyncratic origins, two isolation filters
125 +apply at the event minute. The first requires a calm market factor,
126 +$|z_{m,t}|<1$, so that the event does not coincide with a market-wide
127 +movement. The second requires that no simultaneous wave be under way: the
128 +share of observable securities with $|z_{j,t}|>4$ at minute $t$ is at most
129 +2\%, and at least 50 securities are observable. Both filters are measured, not
130 +assumed: minutes excluded by the second filter form the systemic co-jump
131 +sample of Section~\ref{sec:cojumpdef}, so the design partitions all extreme
132 +activity into an idiosyncratic branch and a systemic branch rather than
133 +discarding the inconvenient one. The filters are intentionally conservative
134 +about common information; the cost, acknowledged throughout, is that any
135 +propagation operating through an immediate market-wide reaction is assigned
136 +to the systemic branch by construction. The filters do not, however, require
137 +that the source be the only $|z|\ge6$ event in its minute: with a 2\% cap on
138 +simultaneous $|z|>4$ names, up to about nine other securities may move
139 +sharply at $t$. In practice \ShareSingleton\% of events are alone in their
140 +minute and the mean number of source events per event minute is
141 +\MeanCoArrival{}; Section~\ref{sec:results} reports every estimate
142 +separately for singleton minutes.
131 143
132 144 \subsection{Propagation outcomes}
133 145 \label{sec:outcomes}
@@ -137,84 +149,105 @@ $h\in\{1,2,5,10,15,30,60,120\}$ minutes. Receivers are all $j\ne i$ in the
137 149 universe with an observable bar at the measurement minute; outcomes require at
138 150 least 50 eligible receivers and are right-censored---recorded as missing, not
139 151 zero---whenever $t+h$ crosses the session close, so end-of-day events never
140 −contribute mechanically attenuated responses. The outcomes are:
152 +contribute mechanically attenuated responses. With $N_{t+h}$ the number of
153 +observable receivers, the outcomes are
141 154 \begin{align}
142 −\mathrm{Breadth}_{i,t}(h) &= \frac{1}{N_{t,h}-1}\sum_{j\ne i}
155 +\mathrm{Breadth}_{i,t}(h) &= \frac{1}{N_{t+h}-1}\sum_{j\ne i}
143 156 \mathbf{1}\{|z_{j,t+h}|>2\},
144 157 \label{eq:breadth}\\
145 −\mathrm{Intensity}_{i,t}(h) &= \frac{1}{N_{t,h}-1}\sum_{j\ne i}|z_{j,t+h}|,
158 +\mathrm{Intensity}_{i,t}(h) &= \frac{1}{N_{t+h}-1}\sum_{j\ne i}|z_{j,t+h}|,
146 159 \label{eq:intensity}\\
147 −\mathrm{Signed}_{i,t}(h) &= \frac{\operatorname{sign}(e_{i,t})}{N_{t,h}-1}
160 +\mathrm{Signed}_{i,t}(h) &= \frac{\operatorname{sign}(e_{i,t})}{N_{t+h}-1}
148 161 \sum_{j\ne i} z_{j,t+h}.
149 162 \label{eq:signed}
150 163 \end{align}
151 −The receiver threshold $|z|>2$ is calibrated to the fat-tailed base rate: it
152 −fires for roughly 7\% of stock-minutes unconditionally, high enough that
153 −breadth is a well-populated share whose \emph{changes} are measured with
154 −precision, low enough to respond to genuine tail activity. Cumulative breadth
155 −replaces the point-in-time indicator with
156 −$\mathbf{1}\{\max_{s\in(t,t+h]}|z_{j,s}|>2\}$, so a receiver that jumps at
157 −$t+3$ and is quiet at $t+15$ still counts; its base rate is mechanically
158 −higher (about 53\% at $h=15$) and its pre-event benchmark uses the mirrored
159 −window $[t-5-h,\,t-5)$.
164 +Breadth counts receivers with a \emph{tail exceedance}, a standardized
165 +residual beyond two in absolute value; we avoid the word ``jump'' for this
166 +object, since the threshold is far below any formal jump-test cutoff. The
167 +threshold is calibrated to the fat-tailed base rate: in placebo windows about
168 +\BaseBreadthShare{} of receivers exceed it, high enough that breadth is a
169 +well-populated share whose changes are measured with precision, low enough
170 +to respond to genuine tail activity. Intensity captures diffuse
171 +sub-threshold responses, and the signed outcome captures directional
172 +comovement with the source. Cumulative breadth replaces the point-in-time
173 +indicator with $\mathbf{1}\{\max_{s\in(t,t+h]}|z_{j,s}|>2\}$, so a receiver
174 +that moves at $t+3$ and is quiet at $t+15$ still counts; its base rate is
175 +mechanically higher (about \CumBaseFifteen\% at $h=15$ in placebo windows) and its pre-event benchmark uses
176 +the mirrored window $[t-5-h,\,t-5)$.
160 177
161 178 Every outcome is also evaluated at $t-5$ (point-in-time) or on the mirrored
162 −pre-window (cumulative). The pre-event value serves two distinct purposes: it
163 −is the \emph{differencing baseline}---shocks arrive disproportionately in
164 −minutes of elevated ambient activity, and the post-minus-pre difference
165 −removes that level---and it is a \emph{falsification statistic}, since genuine
166 −propagation cannot precede its cause.
179 +pre-window (cumulative). The pre-event value is the differencing baseline:
180 +shocks arrive disproportionately in minutes of elevated ambient activity, and
181 +the post-minus-pre difference removes that level. It is also informative in
182 +its own right, and Section~\ref{sec:results} reports the pre-event level of
183 +each arm rather than asserting their equality.
167 184
168 185 \subsection{Lagged correlation networks}
169 186 \label{sec:networkdef}
170 187
171 −For each month $m$ and security $i$, the neighbor set
172 −$\mathcal{N}_i$ collects the twenty securities with the highest absolute
173 −pairwise correlation with $i$'s residuals over month $m{-}1$ (pairwise
174 −complete observations; neighbor sets require at least ten members present in
175 −month $m$ to be used). Splitting equation~\eqref{eq:breadth} into
176 −\begin{equation}
177 −\mathrm{Breadth}^{\mathcal{N}}_{i,t}(h) = \frac{1}{|\mathcal{N}_i|}
188 +For each month $m$ and security $i$, the neighbor set $\mathcal{N}_i$ collects
189 +the twenty securities with the highest absolute pairwise correlation with $i$'s
190 +residuals $e$ over month $m{-}1$ (pairwise complete observations; a neighbor
191 +set is used only if at least ten of its members are present in month $m$).
192 +Splitting equation~\eqref{eq:breadth} into
193 +\begin{align}
194 +\mathrm{Breadth}^{\mathcal{N}}_{i,t}(h) &= \frac{1}{|\mathcal{N}_i|}
178 195 \sum_{j\in\mathcal{N}_i}\mathbf{1}\{|z_{j,t+h}|>2\},
179 −\qquad
180 −\mathrm{Breadth}^{\overline{\mathcal{N}}}_{i,t}(h) = \text{same over }
181 −j\notin\mathcal{N}_i\cup\{i\},
182 −\end{equation}
183 −yields the \emph{network differential}
196 +\label{eq:nbreadth}\\
197 +\mathrm{Breadth}^{\overline{\mathcal{N}}}_{i,t}(h) &= \frac{1}{N_{t+h}-|\mathcal{N}_i|-1}
198 +\sum_{j\notin\mathcal{N}_i\cup\{i\}}\mathbf{1}\{|z_{j,t+h}|>2\},
199 +\label{eq:nbreadth_other}
200 +\end{align}
201 +yields the \emph{network contrast}
184 202 $\mathrm{Breadth}^{\mathcal{N}}-\mathrm{Breadth}^{\overline{\mathcal{N}}}$,
185 203 the paper's sharpest instrument: transmission concentrated among twenty
186 204 connected names moves it one-for-one, while market-wide breadth dilutes the
187 −same response by a factor of twenty-five. Because neighbor sets are lagged,
188 −the differential cannot reflect the event's own contemporaneous comovement;
189 −and because the identical differential is computed for placebo events, any
190 −unconditional tendency of correlated stocks to co-move nets out.
205 +same response by a factor of about twenty. Because neighbor sets are lagged,
206 +the contrast cannot reflect correlations estimated on the event itself; and
207 +because the identical contrast is computed for placebo events, any
208 +unconditional tendency of correlated stocks to be more or less active than
209 +the rest of the universe nets out in the real-minus-placebo
210 +\emph{network differential} of Section~\ref{sec:design}.
211 +
212 +Two properties of this construction must be kept in view. First, neighbors
213 +are identified from one-minute residual correlations, and the precision of
214 +that identification is not constant over time: when the Epps effect is
215 +strong, as it was in the 1990s (Section~\ref{sec:results}), one-minute
216 +correlations are dominated by asynchronous trading and discreteness, and the
217 +twenty ``closest'' names are a noisier selection of true economic peers than
218 +they are in the 2010s. Secular changes in the network differential therefore
219 +combine changes in transmission with changes in how well the network is
220 +measured. Second, the network contrast is computed at $t+h$ only; its
221 +pre-event counterpart at $t-5$ was not stored by the pipeline and is listed
222 +among the analyses left for future work.
191 223
192 224 \subsection{Systemic co-jump minutes}
193 225 \label{sec:cojumpdef}
194 226
195 −A minute is a \emph{systemic co-jump} if the share of eligible securities with
196 −$|z_{j,t}|>4$ exceeds 2\% (with at least 50 eligible names)---exactly the
197 −complement of isolation filter (2). For each such minute we record the
198 −affected shares at the $4\sigma$ and $6\sigma$ thresholds and the standardized
199 −market return. Two summary series are used: the frequency of \emph{severe}
200 −synchronization (affected share above 10\%) per 1,000 eligible minutes, and
201 −the mean affected share within co-jump minutes. We emphasize severity and the
202 −severe-minute count rather than the raw count at the 2\% threshold because
203 −the latter is not comparable across regimes: idiosyncratic tail activity was
204 −two orders of magnitude more frequent in the wide-tick 1990s, so a fixed
205 −low threshold is breached almost continuously early in the sample for reasons
206 −unrelated to synchronization.
227 +A minute is a \emph{systemic co-jump minute} if the share of observable
228 +securities with $|z_{j,t}|>4$ exceeds 2\% (with at least 50 observable
229 +names)---exactly the complement of the second isolation filter. For each such
230 +minute we record the affected shares at the $4\sigma$ and $6\sigma$
231 +thresholds and the standardized market return. Two summary series are used:
232 +the frequency of \emph{severe} synchronization (affected share above 10\%)
233 +per 1,000 observable minutes, and the mean affected share within co-jump
234 +minutes. We emphasize severity and the severe-minute count rather than the
235 +raw count at the 2\% threshold because the latter is not comparable across
236 +regimes: idiosyncratic tail activity was two orders of magnitude more
237 +frequent in the wide-tick 1990s, so a fixed low threshold is breached almost
238 +continuously early in the sample for reasons unrelated to synchronization.
239 +The same discreteness also inflates the severe-minute count before 2001, and
240 +we do not compare the 1990s with later decades on this series.
207 241
208 242 \subsection{What these measures can and cannot see}
209 243 \label{sec:measurement_limits}
210 244
211 245 Three mechanical properties discipline interpretation. First, breadth
212 −measures respond to \emph{tail} activity only; diffuse sub-threshold
213 −responses would appear in intensity, which is reported alongside breadth
214 −throughout. Second, all receiver measures condition on trading: a receiver
215 −without a bar at $t+h$ leaves the denominator rather than entering as a
216 −zero, so illiquidity cannot fabricate absorption. Third, the pre-2001 tick
217 −regime (eighths and sixteenths) inflates both the level of $|z|$ activity and
218 −its base rates; every cross-era comparison is therefore anchored on the
219 −within-era placebo contrast, and the post-decimalization subsample reproduces
220 −all headline results.
246 +responds to tail activity only; diffuse sub-threshold responses appear in
247 +intensity, which is reported alongside breadth. Second, all receiver measures
248 +condition on trading: a receiver without a bar at $t+h$ leaves the
249 +denominator rather than entering as a zero, so illiquidity cannot fabricate
250 +absorption. Third, the pre-2001 tick regime inflates both the level of $|z|$
251 +activity and its base rates; every cross-era comparison is therefore anchored
252 +on the within-era placebo contrast, and the post-decimalization subsample is
253 +reported for every headline estimate.
modified paper/sections/05_empirical_design.tex +80 −53
@@ -4,18 +4,20 @@
4 4 \subsection{The estimand}
5 5
6 6 For an outcome $Y$ from Section~\ref{sec:outcomes}, define the within-event
7 −change $\Delta Y_{i,t}(h) = Y_{i,t}(t{+}h) - Y_{i,t}(t{-}5)$. The estimand is
8 −the difference-in-differences
7 +change $\Delta Y_{i,t}(h) = Y_{i,t}(h) - Y_{i,t}(-5)$, where $Y_{i,t}(-5)$
8 +denotes the same outcome evaluated five minutes before the event. The
9 +baseline estimand is the difference-in-differences
9 10 \begin{equation}
10 11 \Delta\Delta Y(h) \;=\;
11 12 \mathbb{E}\big[\Delta Y(h)\,\big|\,\text{real event}\big]
12 13 \;-\;
13 −\mathbb{E}\big[\Delta Y(h)\,\big|\,\text{matched placebo}\big].
14 +\mathbb{E}\big[\Delta Y(h)\,\big|\,\text{matched placebo}\big],
14 15 \label{eq:dd}
15 16 \end{equation}
16 −In potential-outcomes language, the placebo arm estimates the counterfactual
17 −path of cross-sectional activity that the event window would have followed had
18 −the source stock not jumped, holding fixed the stock, the minute of day, the
17 +estimated as the difference between the two arms' sample means. In
18 +potential-outcomes language, the placebo arm estimates the path of
19 +cross-sectional activity that the event window would have followed had the
20 +source stock not moved, holding fixed the stock, the minute of day, the
19 21 month, and the calm-market state. Three biases that a naive event study would
20 22 absorb are differenced out explicitly. \emph{Volatility clustering}: shocks
21 23 arrive in already-active minutes, so post-event activity is elevated even
@@ -26,66 +28,91 @@ placebo, matched on minute of day, follows the same drift.
26 28 minutes followed by reversion toward normal activity; the placebo is drawn
27 29 under the identical filters and inherits the identical reversion.
28 30
31 +The inner difference has a cost that we quantify rather than assume away. If
32 +real events select minutes whose ambient activity is elevated for transient
33 +reasons, part of that elevation reverts over the following minutes, and
34 +$\Delta\Delta$ is biased downward by the reversion. We therefore report,
35 +alongside~\eqref{eq:dd}, the post-event level difference without
36 +pre-differencing and a stratified estimator that compares real and placebo
37 +post-event levels within deciles of the pre-event level,
38 +\begin{equation}
39 +\widehat{\theta}^{\,S}(h) \;=\; \sum_{b=1}^{10} w_b\,
40 +\Big(\bar Y^{\,\mathrm{real}}_{b}(h) - \bar Y^{\,\mathrm{placebo}}_{b}(h)\Big),
41 +\qquad w_b = \frac{n^{\mathrm{real}}_b}{\sum_{b'} n^{\mathrm{real}}_{b'}},
42 +\label{eq:strat}
43 +\end{equation}
44 +where $b$ indexes deciles of $Y_{i,t}(-5)$ over the pooled sample. Because the
45 +placebo arm has a lower latent activity level than the real arm within any
46 +stratum defined on a noisy pre-event measurement, regression to the mean
47 +pushes~\eqref{eq:strat} upward; the two estimators therefore bracket the
48 +market-wide response from below and from above.
49 +
29 50 \subsection{Placebo construction}
30 51
31 −For each real event $(i,t)$ the matched placebo is drawn by the following
32 −algorithm, with a fixed seed recorded in the replication package:
33 −\begin{enumerate}
34 −\item collect all minutes of the same month with the same minute of day
35 −$\tau(t)$, on a different trading day, at which (a) security $i$ has an
36 −observable return with $|z_{i,\cdot}|<2$, (b) the market-calm filter holds,
37 −and (c) the simultaneity filter holds;
38 −\item draw one candidate uniformly at random; compute every outcome of
39 −Section~\ref{sec:outcomes} at the placebo minute by the identical code path;
40 −\item label the placebo with the real event's magnitude and sign so that all
41 −conditional cuts (by $|z|$ bin, by sign, by state) compare real and placebo
42 −events within the same cell.
43 −\end{enumerate}
44 −Events with an empty candidate set (rare) contribute to the real arm only and
45 −are dropped from difference-in-differences estimation by the pairing on
46 −dates within cells. The match rate exceeds 95\% throughout.
52 +For each real event $(i,t)$ the matched placebo is drawn as follows, with a
53 +fixed seed recorded in the replication package. The candidate set collects all
54 +minutes of the same month with the same minute of day $\tau(t)$ on a different
55 +trading day at which security $i$ has an observable return with
56 +$|z_{i,\cdot}|<2$, the market-calm filter holds, and the simultaneity filter
57 +holds. One candidate is drawn uniformly at random, and every outcome of
58 +Section~\ref{sec:outcomes} is computed at the placebo minute by the identical
59 +code path. The placebo is labelled with the real event's magnitude and sign so
60 +that all conditional cuts---by $|z|$ bin, by sign, by number of co-arriving
61 +events---compare real and placebo events within the same cell. Events with an
62 +empty candidate set contribute to the real arm only; the match rate is
63 +\MatchRate\% over the full sample, \MatchRateNineties\% in the 1990s, when
64 +calm candidate minutes with an observable return are scarcer, and at least
65 +\MatchRateMinPost\% in each later decade. The placebo is not
66 +matched on the pre-event level of the outcome, which is why the pre-event
67 +levels of the two arms can differ and are reported.
47 68
48 69 \subsection{Inference}
49 70
50 71 Outcomes are correlated across events within a trading day---mechanically so,
51 −because events of the same minute share receivers. All inference therefore
52 −resamples \emph{days}: for $b=1,\dots,500$, draw event dates with replacement,
53 −recompute date-level sums and counts for both arms, and form the bootstrap
54 −distribution of the full statistic~\eqref{eq:dd}. Standard errors are
55 −bootstrap standard deviations; the month-cluster variant (resampling
56 −year-months) appears in the robustness table and roughly doubles the standard
57 −error of the network differential without changing any conclusion. With
58 −roughly 8,500 event days, the effective sample is large, and the design's
59 −minimum detectable effect for market-wide breadth at $h=5$ is approximately
60 −$2\times10^{-4}$ at the 5\% level with 80\% power---one additional responding
61 −security per twenty-five events. Null results below are powered nulls.
72 +because events of the same minute share receivers and \MeanCoArrival{} source
73 +events on average share each event minute---and within a source security
74 +across days. All inference therefore resamples clusters: for
75 +$b=1,\dots,500$, draw event dates with replacement, recompute date-level sums
76 +and counts for both arms, and form the bootstrap distribution of the full
77 +statistic. Standard errors are bootstrap standard deviations. Two
78 +alternatives are reported: a month-cluster variant (resampling year-months),
79 +which roughly doubles the standard error of the network differential without
80 +changing any conclusion, and a ticker-cluster variant (resampling source
81 +securities), which more than doubles it. With \EventDays{} event days and
82 +\NEventsRealMM{} million events, the minimum detectable effect for
83 +market-wide breadth at $h=5$ at the 5\% level with 80\% power is about
84 +\MDEFive$\times10^{-4}$, or one additional responding security per
85 +\MDEEventsPer{} events given \NEligMean{} observable receivers on average.
86 +Null results below are powered nulls in this sense.
62 87
63 88 \subsection{Identification: what is and is not claimed}
64 89
65 −The design identifies the \emph{association} between an isolated idiosyncratic
66 −jump and subsequent cross-sectional tail activity, purged of ambient levels,
67 −intraday seasonality, calm-market selection, and the stock's own base rates.
68 −It does not identify the structural channel behind the network differential:
90 +The design identifies the predictive association between an isolated
91 +idiosyncratic extreme and subsequent cross-sectional tail activity, purged of
92 +intraday seasonality, calm-market selection, and the stock's own base rates,
93 +and bounded with respect to ambient-activity selection as described above. It
94 +does not identify the structural channel behind the network differential:
69 95 lagged-correlation neighbors are typically industry peers, ETF co-members, and
70 96 factor-exposure twins, and the differential aggregates fundamental
71 −information transmission (investors rationally repricing linked firms) with
72 −trading-channel transmission (basket flows, hedging, liquidity withdrawal).
73 −Separating those requires timestamped news and order-flow data and is beyond
74 −OHLCV bars; the paper's claims are correspondingly phrased as predictive
97 +information transmission (investors repricing linked firms on the same news)
98 +with trading-channel transmission (basket flows, hedging, liquidity
99 +withdrawal). Separating those requires timestamped news and order-flow data
100 +and is beyond OHLCV bars; the paper's claims are phrased as predictive
75 101 propagation, not causal transmission. Likewise, the systemic co-jump series
76 102 conflates common news arrivals with endogenous synchronization; we document
77 103 its evolution and severity without attributing minutes to either source.
78 104
79 −\subsection{Falsification battery}
105 +\subsection{Falsification and diagnostic battery}
80 106
81 −Five falsifiable predictions are tested and reported. (1) \emph{No
82 −pre-event response}: real-event outcomes at $t-5$ equal placebo levels.
83 −(2) \emph{No placebo network structure}: the neighbor--non-neighbor
84 −differential computed on placebo events alone is zero
85 −($-2.7\pm2.8\times10^{-4}$). (3) \emph{No announcement artifact}: excluding
86 −the 10:00 and 14:00--14:30 ET windows that host most scheduled releases
87 −leaves all estimates unchanged. (4) \emph{No coverage artifact}: fixed-rule
88 −lagged universes make the 2020 source-coverage expansion irrelevant to any
89 −comparison. (5) \emph{No tick-regime artifact}: the post-decimalization
90 −subsample reproduces the market-wide null and the network differential. All
91 −five pass; Table~\ref{tab:robustness} quantifies each.
107 +Five checks are reported. (1) \emph{Pre-event levels}: the levels of every
108 +outcome at $t-5$ in the two arms, with their difference; a non-zero
109 +difference is not a failure of the design but a measured quantity that the
110 +estimators above handle in different ways. (2) \emph{Placebo network
111 +contrast}: the neighbor-minus-other contrast computed on placebo events
112 +alone, over the full sample and by decade. (3) \emph{Announcement windows}:
113 +excluding the 10:00 and 14:00--14:30 ET windows that host most scheduled
114 +releases. (4) \emph{Coverage}: fixed-rule lagged universes make the 2020
115 +source-coverage expansion irrelevant to any comparison. (5)
116 +\emph{Tick regime}: the post-decimalization subsample is reported for every
117 +headline estimate. Tables~\ref{tab:design_checks},
118 +\ref{tab:network_components}, and~\ref{tab:robustness} quantify each.
modified paper/sections/06_results.tex +128 −53
@@ -12,29 +12,52 @@ collapsed. Mean absolute one-minute returns on the top-1000 universe fell
12 12 decade by decade at every minute of the trading day
13 13 (Figure~\ref{fig:seasonality_decade}a), and the frequency of one-minute
14 14 returns exceeding 1\% fell from about \ExtremeGtOneNineties{} per 10,000 in
15 −the 1990s to \ExtremeGtOneTens{} per 10,000 in the 2010s---a factor of roughly
16 −\ExtremeDropFactor{} (Figure~\ref{fig:extreme_minutes}). Part of the 1990s
17 −level reflects bid--ask bounce under eighth and sixteenth ticks rather than
18 −true volatility; the decade-by-decade decline survives range-based measures
19 −and coarser sampling, and we flag the tick-regime caveat wherever it binds.
15 +the 1990s to \ExtremeGtOneTens{} per 10,000 in the 2010s, a factor of
16 +roughly \ExtremeDropFactor{} (Figure~\ref{fig:extreme_minutes}). A large
17 +part of the 1990s level is bid--ask bounce and price discreteness under
18 +eighth and sixteenth ticks rather than volatility
19 +\citep{roll1984simple,harris1991stock}: a single tick on a \$20 stock is a
20 +0.6\% return, and about half of all one-minute returns in the mid-1990s are
21 +exactly zero (Online Appendix). The decline is nonetheless not only a tick
22 +artifact. Measured from 2002, the first full year after decimalization, the
23 +frequency of $|r|>1\%$ minutes falls from \ExtremeGtOneTwoThousandTwo{} per
24 +10,000 to \ExtremeGtOneTens{} in the 2010s, roughly tenfold, and the
25 +cross-section of daily realized variance shifts down over the same years
26 +(Online Appendix).
20 27
21 28 Second, the collapse was not monotone in the tails. The frequency of
22 29 $|r|>5\%$ minutes rose from \ExtremeGtFivePreCovid{} per 10,000 in 2017--2019
23 −to \ExtremeGtFiveCovid{} in 2020---a factor of \ExtremeCovidFactor---and 2008
24 −shows the same signature (Figure~\ref{fig:extreme_minutes}). Extremes became
25 −rarer but, when the market is stressed, arrive in synchronized bursts; we
26 −return to this in Section~\ref{sec:mechanisms}.
27 −
28 −Third, cross-sectional information transmission accelerated dramatically. In
29 −1996 the mean pairwise correlation of one-minute returns among the hundred
30 −largest stocks was \EppsNinetySixOne{}, against \EppsNinetySixThirty{} at
31 −thirty minutes: comovement existed but took tens of minutes to materialize,
32 −the classical Epps effect \citep{epps1979comovements}. By 2015 the one-minute
33 −correlation was \EppsFifteenOne{} against \EppsFifteenThirty{} at thirty
34 −minutes, and 2024 shows the same near-flat profile
35 −(Figure~\ref{fig:epps}). The Epps effect has essentially disappeared from
36 −large-cap U.S. equities: common information is now impounded across the
37 −cross-section within a minute.
30 +to \ExtremeGtFiveCovid{} in 2020, a factor of \ExtremeCovidFactor{}, and 2008
31 +shows the same signature at a smaller scale (Figure~\ref{fig:extreme_minutes}).
32 +Part of the 2020--2021 level reflects the composition of the top-1000
33 +universe in those years, which dollar-volume ranking populates with
34 +leveraged exchange-traded products and distressed issuers that we cannot
35 +label; the qualitative pattern---extremes became rarer but, when the market
36 +is stressed, arrive in synchronized bursts---survives the exclusion of
37 +heuristically identified non-common tickers and is the subject of
38 +Section~\ref{sec:mechanisms}.
39 +
40 +Third, cross-sectional comovement at the one-minute horizon rose from nothing
41 +to most of its half-hour level. In 1996 the mean pairwise correlation of
42 +one-minute returns among the hundred largest stocks was \EppsNinetySixOne{},
43 +against \EppsNinetySixThirty{} at thirty minutes: comovement existed but was
44 +invisible at the minute, the classical Epps effect
45 +\citep{epps1979comovements}. By 2015 the one-minute correlation was
46 +\EppsFifteenOne{} against \EppsFifteenThirty{} at thirty minutes, and 2024
47 +shows the same near-flat profile at a lower level (\EppsTwentyFourOne{}
48 +against \EppsTwentyFourThirty{}; Figure~\ref{fig:epps}). Two mechanisms
49 +produce this flattening and our previous-tick synchronized grids do not
50 +separate them. One is mechanical: nonsynchronous trading and stale prices
51 +attenuate measured high-frequency correlation
52 +\citep{reno2003closer,hayashi2005covariance,zhang2011estimating}, and the
53 +share of stale one-minute prices fell from \ZeroRetNinetySix\% to
54 +\ZeroRetTwentyFour\% between 1996 and 2024. The other is economic: common
55 +information is now impounded across the cross-section within the minute.
56 +\citet{toth2009epps} reach the same conclusion for 1993--2003 and find that
57 +denser trading explains only part of the change. Whatever the mix, the
58 +operational consequence for this paper is the same: one-minute residual
59 +correlations identify economic neighbors far more sharply after 2005 than
60 +before, a point we carry into the interpretation of the network results.
38 61
39 62 \begin{figure}[t]
40 63 \centering
@@ -51,64 +74,116 @@ stock-day's volume; (c) high--low log range.}
51 74 \includegraphics[width=.72\textwidth]{figE5_extreme_minutes.pdf}
52 75 \caption{\textbf{Extreme one-minute returns.} Annual frequency (log scale) of
53 76 one-minute absolute log returns exceeding 1\%, 2\%, and 5\%, lagged top-1000
54 −universe, regular hours. Pre-2001 levels partly reflect eighth/sixteenth tick
55 −bid--ask bounce.}
77 +universe, regular hours. Pre-2001 levels largely reflect eighth/sixteenth
78 +tick bid--ask bounce.}
56 79 \label{fig:extreme_minutes}
57 80 \end{figure}
58 81
59 82 \begin{figure}[t]
60 83 \centering
61 84 \includegraphics[width=.62\textwidth]{figE8_epps.pdf}
62 −\caption{\textbf{The disappearance of the Epps effect.} Mean pairwise return
85 +\caption{\textbf{The flattening of the Epps curve.} Mean pairwise return
63 86 correlation among the top-100 lagged dollar-volume stocks at 1-, 5-, 15-, and
64 87 30-minute sampling, previous-tick synchronization within day, for 1996, 2005,
65 −2015, and 2024.}
88 +2015, and 2024. Correlations are computed on calendar-clock grids and are
89 +therefore attenuated by nonsynchronous trading at short intervals, most
90 +strongly in 1996.}
66 91 \label{fig:epps}
67 92 \end{figure}
68 93
69 −\subsection{Baseline: idiosyncratic shocks do not propagate}
94 +\subsection{Baseline: isolated shocks and market-wide tail activity}
70 95
71 −Table~\ref{tab:baseline_propagation} reports the paper's baseline estimates of
96 +Table~\ref{tab:baseline_propagation} reports the baseline estimates of
72 97 equation~\eqref{eq:dd} for \NEventsReal{} isolated $|z|\ge6$ source events and
73 −\NEventsPlacebo{} matched placebos. The abnormal breadth response is
74 −economically zero at every horizon: at $h=5$ minutes the point estimate is
75 −\DDBreadthFive$\times10^{-4}$ (bootstrap SE \DDBreadthFiveSE$\times10^{-4}$),
76 −and at $h=30$ minutes \DDBreadthThirty$\times10^{-4}$
77 −(SE \DDBreadthThirtySE$\times10^{-4}$). Against a placebo base rate of roughly
78 −$7\times10^{-2}$, the largest estimate in absolute value implies that an
79 −isolated six-sigma shock changes the probability that any given other security
80 −jumps by less than half of one percent \emph{of the base rate}---with the
81 −wrong sign. Abnormal intensity estimates are equally null.
98 +\NEventsPlacebo{} matched placebos. The difference-in-differences breadth
99 +response is economically negligible at every horizon: at $h=5$ minutes the
100 +point estimate is \DDBreadthFive$\times10^{-4}$ (bootstrap standard error
101 +\DDBreadthFiveSE$\times10^{-4}$), and at $h=30$ minutes
102 +\DDBreadthThirty$\times10^{-4}$ (\DDBreadthThirtySE$\times10^{-4}$). Against
103 +a placebo base rate of about \BaseBreadthShare{} (Table~\ref{tab:design_checks},
104 +Panel~A), the largest estimate in absolute value implies that an isolated
105 +six-sigma move changes the probability that any given other security shows a
106 +tail exceedance by about a third of one percent of the base rate, with a
107 +negative sign that is statistically distinguishable from zero at $h=30$.
108 +Difference-in-differences intensity estimates are of the same order and
109 +alternate in sign.
82 110
83 111 \input{tables/table4_baseline_propagation}
84 112
85 −Three features of Table~\ref{tab:baseline_propagation} deserve emphasis.
86 −First, this is a powered null: standard errors are small enough to detect a
87 −breadth response of $3\times10^{-4}$---one additional responding security per
88 −seventeen events---at conventional levels. Second, the pre-event falsification
89 −passes: real-event outcomes at $t-5$ equal placebo levels, so the null is not
90 −produced by anticipation contaminating the baseline. Third, the null is
91 −uniform: Figure~\ref{fig:irf_decade} shows the same flat response profile in
92 −each of the four decades, so absorption is not a recent acquisition of
93 −electronic markets---though the composition of what is being absorbed changed
94 −radically, as the descriptive results show.
113 +The design diagnostics of Table~\ref{tab:design_checks} qualify this null in
114 +three ways. Panel~A shows that real events do not arrive in ordinary minutes:
115 +five minutes before the event, breadth is \PreBreadthReal$\times10^{-4}$ for
116 +real events against \PreBreadthPlacebo$\times10^{-4}$ for placebos, a
117 +difference of \PreBreadthDiff$\times10^{-4}$ (standard error
118 +\PreBreadthDiffSE), and intensity is higher by \PreIntensityDiff{}
119 +(\PreIntensityDiffSE). A six-sigma idiosyncratic move is thus more likely on
120 +days and in minutes when the whole cross-section is already active, even
121 +though the market factor is calm by construction. This is exactly the
122 +ambient level that the inner difference of~\eqref{eq:dd} removes, and it is
123 +the reason the difference-in-differences, not a post-event comparison, is the
124 +baseline; but it also means that the ``no pre-event response'' statement
125 +sometimes made for designs of this kind does not hold here as a statement
126 +about levels, and we do not make it.
127 +
128 +Panel~C reports the three estimators side by side. Without pre-differencing,
129 +real events show \PostDiffFive$\times10^{-4}$ more breadth than placebos at
130 +$h=5$, almost exactly the pre-event gap, so that the difference-in-differences
131 +is close to zero. Within deciles of the pre-event level, the gap is
132 +\StratFive$\times10^{-4}$ (\StratFiveSE) at $h=5$ and
133 +\StratThirty$\times10^{-4}$ (\StratThirtySE) at $h=30$. As argued in
134 +Section~\ref{sec:design}, the two estimators bracket the market-wide
135 +response: between zero and roughly \StratShareOfBase\% of the base rate.
136 +Either bound is small relative to the network response of
137 +Section~\ref{sec:mechanisms}, and the conclusion we draw is correspondingly
138 +bounded: isolated idiosyncratic extremes raise market-wide tail activity by at
139 +most a few percent of its base rate within two hours.
140 +
141 +Panel~B separates events by the number of source events detected in the same
142 +minute. For the \ShareSingleton\% of events that are alone in their minute,
143 +the difference-in-differences is \DDSingletonFive$\times10^{-4}$
144 +(\DDSingletonFiveSE) at $h=5$ and \DDSingletonThirty$\times10^{-4}$
145 +(\DDSingletonThirtySE) at $h=30$: slightly positive and within two standard
146 +errors of zero. The negative full-sample estimates at longer horizons come
147 +from minutes with three to five co-arriving events
148 +(\DDMultiThirty$\times10^{-4}$, \DDMultiThirtySE, at $h=30$), which are the
149 +minutes with the most elevated ambient activity and the strongest reversion.
150 +The negative sign of the pooled estimate is therefore a mean-reversion
151 +artifact of co-arrival minutes, not evidence that shocks calm the market.
152 +
153 +Three further features of Table~\ref{tab:baseline_propagation} deserve
154 +emphasis. First, this is a powered null: the minimum detectable
155 +difference-in-differences at $h=5$ is about \MDEFive$\times10^{-4}$, one
156 +additional responding security per \MDEEventsPer{} events. Second, the null
157 +is uniform in time: Figure~\ref{fig:irf_decade} shows the same flat response
158 +profile in each of the four decades, and the Online Appendix shows it year by
159 +year, so absorption is not a recent acquisition of electronic markets---though
160 +the composition of what is being absorbed changed radically, as the
161 +descriptive results show. Third, Panel~D of Table~\ref{tab:design_checks}
162 +shows that the standard errors are insensitive to the clustering unit for
163 +the market-wide statistics.
95 164
96 165 \begin{figure}[t]
97 166 \centering
98 167 \includegraphics[width=.6\textwidth]{figF1_irf_decade.pdf}
99 −\caption{\textbf{Abnormal propagation breadth by decade.} Point estimates of
168 +\caption{\textbf{Difference-in-differences breadth by decade.} Point estimates of
100 169 $\Delta\Delta$ breadth (equation~\eqref{eq:dd}) with 95\% day-block bootstrap
101 170 bands, by horizon and decade.}
102 171 \label{fig:irf_decade}
103 172 \end{figure}
104 173
105 −Conditioning does not resurrect average propagation
106 −(Figure~\ref{fig:gradient_state}): estimates by shock sign and by calm versus
107 −stressed market states are all within two standard errors of zero. The only
108 −suggestive pattern is monotonicity in magnitude---the largest shocks
109 −($12\sigma$--$20\sigma$) carry a positive but insignificant point estimate of
110 −\DDGradLargest$\times10^{-4}$ (SE \DDGradLargestSE$\times10^{-4}$)---which
111 −motivates the concentrated-propagation probes of the next section.
174 +\input{tables/table9_design_checks}
175 +
176 +Conditioning does not change the picture (Figure~\ref{fig:gradient_state}):
177 +estimates by shock sign and by calm versus stressed market state are within
178 +two standard errors of zero. Across magnitude bins the point estimates rise
179 +from significantly negative for $6\sigma$--$8\sigma$ events, which are the
180 +most numerous and the most affected by the co-arrival reversion just
181 +described, to a positive but insignificant \DDGradLargest$\times10^{-4}$
182 +(\DDGradLargestSE) for $12\sigma$--$20\sigma$ events. The signed response of
183 +Panel~E is noisy and never significant: the cross-section does not drift in
184 +the direction of the source shock. None of this resurrects market-wide
185 +propagation; it motivates asking where a concentrated response would hide,
186 +which is the subject of the next section.
112 187
113 188 \begin{figure}[t]
114 189 \centering
modified paper/sections/07_mechanisms.tex +129 −59
@@ -1,98 +1,168 @@
1 1 \section{Mechanisms}
2 2 \label{sec:mechanisms}
3 3
4 −The baseline null says that isolated idiosyncratic shocks do not raise
5 −market-wide jump activity. This section asks where transmission lives if not
6 −there: in the shocked stock's network neighborhood, in transient responses
7 −between measurement points, and in the systemic co-jump minutes that the
8 −idiosyncratic design deliberately sets aside.
4 +The baseline bounds the market-wide response to isolated idiosyncratic
5 +extremes at a few percent of the base rate. This section asks where
6 +transmission lives if not there: in the shocked stock's network neighborhood,
7 +in transient responses between measurement points, and in the systemic
8 +co-jump minutes that the idiosyncratic design deliberately sets aside.
9 9
10 −\subsection{Network-concentrated propagation --- and its intensification}
10 +\subsection{Network-concentrated propagation}
11 11
12 −Market-wide breadth spreads any response across five hundred receivers; a
12 +Market-wide breadth spreads any response across several hundred receivers; a
13 13 transmission channel operating through economic or statistical linkages
14 14 concentrates it among a handful. For every event we therefore compare the
15 15 response of the source's twenty closest lagged-correlation neighbors with the
16 16 response of all other receivers, real minus placebo
17 17 (Table~\ref{tab:network_mechanism}).
18 18
19 −The differential is large, precisely estimated, and horizon-decaying: five
20 −minutes after an isolated $|z|\ge6$ shock, the source's neighbors are
21 −\NetDiffFive$\times10^{-4}$ more likely than non-neighbors to exceed
22 −$|z|>2$---relative to placebo---with a bootstrap standard error of
23 −\NetDiffFiveSE$\times10^{-4}$ ($t\approx34$); the differential declines to
24 −\NetDiffThirty$\times10^{-4}$ at thirty minutes. Against the $\sim700\times
25 −10^{-4}$ base rate, a neighbor's jump probability rises by roughly nine
26 −percent of its unconditional level. Idiosyncratic shocks do propagate---to the
27 −connected few, not to the diffuse many, which is precisely why market-wide
28 −breadth cannot detect them.
19 +The network differential is large relative to the market-wide estimates,
20 +precisely estimated, and decays slowly with the horizon: five minutes after
21 +an isolated $|z|\ge6$ move, the source's neighbors are
22 +\NetDiffFive$\times10^{-4}$ more likely than other receivers to show a tail
23 +exceedance, net of the same contrast for placebos, with a day-block bootstrap
24 +standard error of \NetDiffFiveSE$\times10^{-4}$ and a ticker-cluster
25 +standard error of \NetDiffFiveTickerSE$\times10^{-4}$; the differential
26 +declines to \NetDiffFifteen$\times10^{-4}$ at fifteen minutes and
27 +\NetDiffThirty$\times10^{-4}$ at thirty. Against a neighbor base rate of
28 +about \NeighborBaseShare{}, a neighbor's exceedance probability rises by
29 +roughly six percent of its unconditional level. The response is symmetric in
30 +the sign of the shock (\NetDiffNeg$\times10^{-4}$ after negative and
31 +\NetDiffPos$\times10^{-4}$ after positive moves). Idiosyncratic extremes do
32 +propagate---to the connected few, not to the diffuse many, which is why
33 +market-wide breadth cannot detect them.
29 34
30 35 \input{tables/table7_network}
31 36
32 −The evolution across decades is the paper's sharpest dynamic fact
33 −(Figure~\ref{fig:network}): the five-minute neighbor differential was
34 −\NetDiffFiveNineties$\times10^{-4}$ in 1992--1999 and
35 −\NetDiffFiveOughts$\times10^{-4}$ in 2000--2009, then jumps to
37 +Table~\ref{tab:network_components} decomposes the differential into its
38 +four components. Over the full sample the contrast is driven by the real
39 +arm: neighbors of a shocked stock show about $61\times10^{-4}$ more
40 +exceedances than other receivers, while the placebo contrast is
41 +$-2.7\times10^{-4}$. By decade, however, both arms move. In the 1990s,
42 +neighbors show \NetRealNdNineties$\times10^{-4}$ more exceedances than
43 +other receivers after a real event and \NetPlaceboNdNineties$\times10^{-4}$
44 +more at placebo minutes; in the 2010s the real contrast is
45 +\NetRealNdTens$\times10^{-4}$ and the placebo contrast is
46 +\NetPlaceboNdTens$\times10^{-4}$. The placebo contrast is thus far from zero
47 +within decades even though it averages to nearly zero over the sample: in the 1990s, the twenty most correlated names are simply the most
48 +actively traded names, which show more exceedances in every minute; after
49 +2000, the neighbors of a stock that is quiet at the placebo minute are
50 +themselves quieter than the rest of the universe, as correlation implies.
51 +The network differential is therefore legitimately a difference of two
52 +contrasts, and only the real-minus-placebo column of
53 +Table~\ref{tab:network_components} carries a propagation interpretation.
54 +Panel~B of Table~\ref{tab:design_checks} adds that the differential survives
55 +the restriction to minutes in which the source is the only $|z|\ge6$ event
56 +(\NetDiffSingleton$\times10^{-4}$, \NetDiffSingletonSE), so it is not an
57 +artifact of correlated pairs moving in the same minute and remaining active
58 +five minutes later, although co-arrivals do account for about a fifth of the
59 +pooled estimate.
60 +
61 +\input{tables/table10_network_components}
62 +
63 +The evolution across decades (Figure~\ref{fig:network}) is the paper's
64 +sharpest dynamic fact and its most delicate one. The five-minute network
65 +differential was \NetDiffFiveNineties$\times10^{-4}$ in 1992--1999 and
66 +\NetDiffFiveOughts$\times10^{-4}$ in 2000--2009, then rises to
36 67 \NetDiffFiveTens$\times10^{-4}$ in 2010--2019 and
37 −\NetDiffFiveTwenties$\times10^{-4}$ in 2020--2026---a tripling concentrated
38 −exactly in the era of algorithmic liquidity provision, correlation-aware
39 −execution, and index intermediation. Propagation along the statistical network
40 −became roughly three times stronger while market-wide spillover remained zero:
41 −transmission grew \emph{sharper}, not broader.
68 +\NetDiffFiveTwenties$\times10^{-4}$ in 2020--2026, a threefold increase that
69 +coincides with the era of algorithmic liquidity provision, correlation-aware
70 +execution, and index intermediation, and that occurs while market-wide
71 +spillover stays bounded near zero. Two readings are consistent with this
72 +pattern and the data do not separate them. Under the first, transmission
73 +along economic linkages became stronger and faster---basket arbitrage and
74 +statistical-arbitrage strategies convert a move in one name into trades in
75 +its peers within minutes, as \citet{bendavid2018etfs} and
76 +\citet{foucault2016news} would predict. Under the second, transmission was
77 +always present but the one-minute correlations from which we identify
78 +neighbors were too noisy in the 1990s to find the true peers: with the Epps
79 +effect of Figure~\ref{fig:epps}, the prior-month one-minute residual
80 +correlation of a genuine industry pair in 1996 is barely distinguishable from
81 +sampling noise, and a neighbor set contaminated by false peers dilutes the
82 +contrast mechanically. The decline of the 1990s differential at $h=5$
83 +relative to $h=30$ (Table~\ref{tab:network_mechanism}), the opposite of the
84 +decay seen in later decades, is consistent with the second reading, since
85 +slow comovement is what an Epps-era network would transmit. The honest
86 +summary is that propagation along the measured network at least tripled, and
87 +that transmission became sharper rather than broader; how much of the
88 +increase is a change in the market and how much a change in the measurement
89 +requires neighbor sets built on coarser-sampled or transaction-time
90 +correlation estimators \citep{hayashi2005covariance}, an extension we leave
91 +for future work.
42 92
43 93 \begin{figure}[t]
44 94 \centering
45 95 \includegraphics[width=.62\textwidth]{figF4_network.pdf}
46 96 \caption{\textbf{Network-concentrated propagation by decade.} Neighbor minus
47 −non-neighbor $\Delta\Delta$ breadth at $h\in\{5,15,30\}$ with 95\% day-block
48 −bootstrap intervals. Neighbors are the source's top-20 absolute-correlation
49 −partners estimated on prior-month residuals.}
97 +other-receiver breadth at $h\in\{5,15,30\}$, real minus placebo, with 95\%
98 +day-block bootstrap intervals. Neighbors are the source's top-20
99 +absolute-correlation partners estimated on prior-month one-minute residuals.}
50 100 \label{fig:network}
51 101 \end{figure}
52 102
53 103 \subsection{Cumulative responses}
54 104
55 −Point-in-time breadth at $t+h$ misses a receiver that jumps at $t+3$ and is
105 +Point-in-time breadth at $t+h$ misses a receiver that moves at $t+3$ and is
56 106 quiet at $t+15$. Cumulative breadth over $(t,t+h]$, differenced against its
57 107 mirrored pre-window and the placebo, remains null at the market level (at
58 −thirty minutes: \CumThirty$\times10^{-4}$, SE \CumThirtySE$\times10^{-4}$).
59 −The market-wide null is therefore not an artifact of sampling the response at
60 −isolated instants; transient market-wide cascades after isolated idiosyncratic
61 −shocks are absent as well.
108 +thirty minutes: \CumThirty$\times10^{-4}$, standard error
109 +\CumThirtySE$\times10^{-4}$, against a base rate near
110 +\CumBaseFifteen\% at fifteen minutes). The market-wide bound is therefore not
111 +an artifact of sampling the response at isolated instants; transient
112 +market-wide cascades after isolated idiosyncratic extremes are absent as
113 +well.
62 114
63 −\subsection{Systemic co-jump minutes: where fragility lives}
115 +\subsection{Systemic co-jump minutes}
64 116
65 117 The isolation filters exclude minutes in which more than 2\% of the universe
66 −jumps simultaneously. These minutes are the natural home of the fragility
67 −hypothesis, and their time series (Figure~\ref{fig:cojumps}) delivers two
68 −messages. First, the ordinary cross-section \emph{de-synchronized} secularly:
69 −with idiosyncratic tail activity a hundred times rarer, minutes breaching the
70 −fixed 2\% threshold fell from near-ubiquity in the sparse, wide-tick 1990s to
71 −about one minute in eight today---a comparability warning we take seriously,
72 −and the reason the fixed-threshold count is not itself a fragility measure.
73 −Second, conditional severity moves the other way in stress: within co-jump
74 −minutes, the mean share of names beyond $4\sigma$ averages
75 −\CoJumpSevCalm\% in calm recent years but reaches \CoJumpSevGFC\% in 2008 and
76 −\CoJumpSevCovid\% in 2020, and severe-synchronization minutes (more than 10\%
77 −of the universe jumping together) spike from \SeverePreCovid{} per 1,000 in
78 −2017--2019 to \SevereCovid{} per 1,000 in 2020. Combined with the
79 −figE5 evidence that $|r|>5\%$ minutes rose by a factor of
80 −\ExtremeCovidFactor{} in 2020, the pattern is consistent across measures:
81 −modern stress arrives as market-wide synchronized minutes, not as
82 −single-name events.
118 +moves beyond four robust standard deviations simultaneously. These minutes
119 +are the natural home of the fragility hypothesis, and their time series
120 +(Figure~\ref{fig:cojumps}) delivers two messages, each with a comparability
121 +limit. First, the ordinary cross-section de-synchronized secularly: with
122 +idiosyncratic tail activity two orders of magnitude rarer, minutes breaching
123 +the fixed 2\% threshold fell from near-ubiquity in the sparse, wide-tick
124 +1990s to about \CoJumpPerThousandRecent{} per 1,000, roughly one minute in
125 +eight, in 2021--2025. Under the same standardization the early 1990s also
126 +produce \SevereEarlyNineties{} severe minutes per 1,000 and a mean affected
127 +share of \CoJumpSevEarlyNineties\% within co-jump minutes---both above
128 +2008---which we read as discreteness rather than fragility: when most minute
129 +returns are zero or one tick, a two-tick move in many names at once breaches
130 +a four-MAD threshold without any economic synchronization. The fixed-threshold
131 +count and the pre-2001 severity series are therefore not fragility measures,
132 +and no claim in this paper compares synchronization across the
133 +decimalization boundary.
134 +
135 +Second, within the post-decimalization era, conditional severity moves
136 +sharply in stress. Within co-jump minutes, the mean share of names beyond
137 +$4\sigma$ averages \CoJumpSevCalm\% in calm recent years but reaches
138 +\CoJumpSevGFC\% in 2008 and \CoJumpSevCovid\% in 2020, and
139 +severe-synchronization minutes (more than 10\% of the universe moving
140 +together) rise from \SeverePreCovid{} per 1,000 in 2017--2019 to
141 +\SevereCovid{} per 1,000 in 2020, with \SevereGFC{} per 1,000 in 2008
142 +against fewer than 6 per 1,000 in 2006 and 2019: an increase of roughly an
143 +order of magnitude in both crises. Combined with the
144 +evidence of Figure~\ref{fig:extreme_minutes} that $|r|>5\%$ minutes rose by a
145 +factor of \ExtremeCovidFactor{} in 2020, the pattern is consistent across
146 +measures and with the collective-synchronization events documented by
147 +\citet{calcagnile2018collective} on a shorter sample: modern stress arrives
148 +as market-wide synchronized minutes, not as single-name events. Whether those
149 +minutes are common news arrivals or endogenous synchronization of the kind
150 +modelled by \citet{bormetti2015modelling} is not identified by bars alone.
83 151
84 152 \begin{figure}[t]
85 153 \centering
86 154 \includegraphics[width=\textwidth]{figF5_cojumps.pdf}
87 155 \caption{\textbf{Systemic co-jump minutes, 1992--2026.} (a) Minutes per 1,000
88 −in which more than 10\% of eligible top-500 names exceed $|z|>4$
156 +in which more than 10\% of observable top-500 names exceed $|z|>4$
89 157 simultaneously (log scale). (b) Mean affected share within co-jump minutes
90 −($>$2\% threshold).}
158 +($>$2\% threshold). Pre-2001 levels are inflated by price discreteness and
159 +are not comparable with later years.}
91 160 \label{fig:cojumps}
92 161 \end{figure}
93 162
94 −Taken together, the mechanism evidence resolves the baseline null into a
95 −sharper statement: idiosyncratic shocks propagate narrowly along the lagged
96 −correlation network, that channel strengthened threefold in the algorithmic
97 −era, and market-wide fragility is the province of common-shock minutes whose
98 −severity---not frequency---is the binding object.
163 +Taken together, the mechanism evidence sharpens the baseline into a bounded
164 +statement: idiosyncratic extremes propagate narrowly along the lagged
165 +correlation network, that measured channel at least tripled in the
166 +algorithmic era, and market-wide tail risk in the post-decimalization era is
167 +the province of common-shock minutes whose severity in crises, not their
168 +frequency in calm times, is the binding object.
modified paper/sections/08_robustness.tex +54 −24
@@ -1,29 +1,59 @@
1 1 \section{Robustness}
2 2 \label{sec:robustness}
3 3
4 −The Internet Appendix reports the full battery specified in the analysis
5 −protocol. The main text highlights the checks that discipline
6 −interpretation. \emph{Tick-regime}: pre-2001 minute returns embed
7 −eighth/sixteenth-tick bid--ask bounce; all secular comparisons are re-run on
8 −range-based measures, on 5- and 15-minute aggregation, and within
9 −post-decimalization subsamples. \emph{Coverage}: the 2020 source-coverage
10 −expansion is absorbed by fixed-rule lagged top-$N$ universes; results are
11 −reported for $N\in\{500,1000,3000\}$. \emph{Universe composition}: ETFs and
12 −likely non-common instruments (ticker-shape heuristic) are excluded in
13 −robustness runs. \emph{Corporate actions}: split windows are excluded
14 −throughout; sensitivity to window width and detection thresholds is
15 −tabulated. \emph{Thresholds}: source shocks at $|z|\ge4,5,6$ and receiver
16 −thresholds at $|z|>2,3$ leave the baseline null unchanged. \emph{Inference}:
17 −day-block and month-block bootstrap agree. Table~\ref{tab:robustness}
18 −collects the headline checks: the market-wide null and the $+64\times10^{-4}$
19 −network differential are stable across the post-decimalization subsample,
20 −mid-day-only events, announcement-window exclusions, calm and crisis years,
21 −higher source thresholds, and non-common-ticker exclusions, and the neighbor
22 −differential measured on placebo events alone is zero ($-2.7\pm2.8\times
23 −10^{-4}$)---the permuted-benchmark falsification the design commits to. The
24 −one systematic pattern is that the network differential \emph{declines} with
25 −source magnitude ($+64$ at $|z|\ge6$ to $+34\times10^{-4}$ at $|z|\ge12$),
26 −consistent with the largest events being firm-specific news that neighbors
27 −correctly do not price.
4 +Table~\ref{tab:robustness} re-estimates the market-wide
5 +difference-in-differences at $h\in\{5,30\}$ and the network differential at
6 +$h=5$ under eleven designs; Tables~\ref{tab:design_checks}
7 +and~\ref{tab:network_components} report the design diagnostics discussed
8 +above. We describe only checks that were executed; analyses that the
9 +replication package does not contain are listed as such in the Online
10 +Appendix and in the concluding section.
11 +
12 +\emph{Tick regime.} Pre-2001 minute returns embed eighth- and sixteenth-tick
13 +bid--ask bounce. Restricting events to 2002 onward leaves the market-wide
14 +estimates unchanged ($-1.03$ and $-3.04\times10^{-4}$) and the network
15 +differential at $65.8\times10^{-4}$. \emph{Time of day.} Restricting events
16 +to 10:00--15:30 ET, which removes the opening and closing periods where
17 +seasonal scaling is most demanding, and separately excluding the 10:00 and
18 +14:00--14:30 ET windows that host most scheduled macroeconomic releases,
19 +changes no estimate by more than one standard error. \emph{Market state.}
20 +Excluding 2008--2009 and 2020 leaves both objects unchanged; within the
21 +crisis years alone the market-wide estimate is imprecise
22 +($2.7\pm5.1\times10^{-4}$) and the network differential is, if anything,
23 +larger ($68\times10^{-4}$). \emph{Source magnitude.} Raising the source
24 +threshold to $|z|\ge8$ and $|z|\ge12$ moves the market-wide point estimates
25 +toward zero or slightly positive, consistent with the co-arrival pattern of
26 +Table~\ref{tab:design_checks}, and lowers the network differential from
27 +$64$ to $50$ and $34\times10^{-4}$. The decline with magnitude is the one
28 +systematic pattern in the table; it is consistent with the largest moves
29 +being firm-specific news that peers correctly do not price, and also with
30 +the largest moves being the ones most often accompanied by halts or
31 +erroneous prints that bars cannot flag. \emph{Universe composition.}
32 +Excluding source securities flagged as likely non-common by the
33 +ticker-shape heuristic leaves every estimate within rounding of the
34 +baseline; ETFs cannot be identified and are not excluded. \emph{Inference.}
35 +Month-cluster resampling raises the standard error of the network
36 +differential from $1.8$ to $3.4\times10^{-4}$ and ticker-cluster resampling
37 +to \NetDiffFiveTickerSE$\times10^{-4}$, leaving a $t$-statistic above
38 +fourteen in every case; the market-wide standard errors are insensitive to
39 +the clustering unit. \emph{Placebo network.} The neighbor-minus-other
40 +contrast computed on placebo events alone is $-2.7\pm2.8\times10^{-4}$ over
41 +the full sample. As Table~\ref{tab:network_components} shows, this average
42 +conceals decade-level offsets of opposite signs, which is why every network
43 +statistic in the paper is a real-minus-placebo difference and never a
44 +real-arm contrast on its own.
28 45
29 46 \input{tables/table8_robustness}
47 +
48 +Several checks that a complete treatment would include were not executed
49 +and are not claimed. The pipeline stores events at the $|z|\ge6$ threshold
50 +and the receiver threshold $|z|>2$ only, so lower source thresholds
51 +($|z|\ge4,5$), a higher receiver threshold ($|z|>3$), and the top-1000 and
52 +top-3000 universes would require re-running the detection stage on the raw
53 +panel. Neighbor sets built on 5- or 30-minute residual correlations, or on a
54 +transaction-time covariance estimator, and the pre-event ($t-5$) network
55 +contrast, are the two extensions most needed to settle the interpretation of
56 +the secular increase in the network differential. Permuted-network placebos
57 +that preserve industry and liquidity composition, a separate estimate of
58 +propagation inside announcement windows, and a Hawkes-type model of the
59 +co-jump minute series are natural further steps.
modified paper/sections/09_conclusion.tex +42 −20
@@ -1,24 +1,46 @@
1 1 \section{Conclusion}
2 2 \label{sec:conclusion}
3 3
4 −Three decades of minute-level data deliver a sharper statement of intraday
5 −market fragility than either the efficiency or the fragility narrative alone.
6 −Isolated idiosyncratic shocks---even six-sigma events in the largest five
7 −hundred securities---do not propagate: their abnormal cross-sectional
8 −footprint is statistically and economically zero at every horizon from one
9 −minute to two hours, in every decade. What changed is the market's handling of
10 −common variation: information that in the mid-1990s took half an hour to
11 −appear in cross-stock comovement now arrives within the minute, ordinary
12 −extreme moves became an order of magnitude rarer, and extreme synchronization
13 −concentrated in systemic co-jump minutes whose frequency explodes in stress
14 −episodes. Fragility migrated from securities to minutes.
4 +Three decades of minute-level data on the full U.S.\ equity cross-section
5 +allow the title's question to be answered in its three parts. Faster: common
6 +information that in the mid-1990s took tens of minutes to appear in
7 +cross-stock comovement now appears within the minute, whether because
8 +incorporation accelerated or because the mechanical attenuation of
9 +nonsynchronous trading receded, and most likely both. Wider: no. Isolated
10 +idiosyncratic extremes---six-sigma one-minute moves in the largest five
11 +hundred securities---raise market-wide tail activity by at most a few percent
12 +of its base rate within two hours, in every decade and for either sign; what
13 +transmission exists is concentrated among the shocked stock's twenty closest
14 +lagged-correlation neighbors, and that concentrated channel became roughly
15 +three times larger between the floor-trading and algorithmic eras, so
16 +transmission became sharper rather than broader. More fragile: only in a
17 +bounded sense. Within the post-decimalization era, ordinary extreme minutes
18 +became an order of magnitude rarer while minutes in which a tenth of the
19 +universe moves together are roughly ten times more frequent in 2008 and 2020
20 +than in calm years; across the decimalization boundary the question cannot be
21 +answered with threshold-based measures, because price discreteness
22 +manufactures synchronization in the 1990s.
15 23
16 −The evidence is deliberately associational where identification runs out:
17 −OHLCV bars do not observe quotes, depth, or order flow, so liquidity operates
18 −through proxies; tickers are not permanent identities, so long-run statements
19 −run within screened spells; and the co-jump series conflates common news with
20 −endogenous synchronization---separating the two requires timestamped news and
21 −quote data, which we leave as the natural extension. For market design and
22 −risk monitoring, the practical implication is direct: surveillance aimed at
23 −single-name events monitors the part of the system that modern markets already
24 −absorb; the binding constraint is the market-wide minute.
24 +The evidence is deliberately associational where identification runs out,
25 +and the paper reports the diagnostics that qualify its own headline numbers.
26 +Real events arrive in minutes of elevated ambient activity, so the
27 +market-wide response is bracketed rather than pinned at zero; the network
28 +differential is a difference between two non-zero contrasts, and its secular
29 +increase combines a change in transmission with a change in how precisely
30 +one-minute correlations identify economic neighbors; and the co-jump series
31 +conflates common news with endogenous synchronization. OHLCV bars do not
32 +observe quotes, depth, order flow, or security type, so liquidity operates
33 +through proxies and ETFs cannot be isolated; tickers are not permanent
34 +identities, so long-run statements run within screened spells. The
35 +extensions that would most sharpen the conclusions are, in order, neighbor
36 +sets estimated on coarser-sampled or transaction-time correlations, the
37 +pre-event network contrast, lower source and higher receiver thresholds on
38 +wider universes, and the attribution of co-jump minutes to news using
39 +timestamped announcement data.
40 +
41 +For market design and risk monitoring the practical implication is direct.
42 +Surveillance aimed at single-name events monitors the part of the system that
43 +modern markets absorb, or transmit only to a handful of linked names; the
44 +binding constraint for market-wide stability is the market-wide minute, and
45 +the margin that moved over three decades is the sharpness of transmission
46 +along statistical linkages, not its breadth.
modified paper/tables/tabP1_pilot_summary.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/02_pilot_figures_tables.py — do not edit numbers by hand.
2 −\begin{table}[t]
2 +\begin{table}[htbp]
3 3 \centering
4 4 \begin{adjustbox}{max width=\textwidth}
5 5 \begin{threeparttable}
added paper/tables/table10_network_components.tex +23 −0
@@ -0,0 +1,23 @@
1 +% Auto-generated by scripts/17_revision_checks.py — do not edit numbers.
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}\begin{threeparttable}
3 +\caption{Components of the network differential at $h=5$, by decade}
4 +\label{tab:network_components}
5 +\small\begin{tabular}{lrrrrrrr}
6 +\toprule
7 + & \multicolumn{3}{c}{Real events} & \multicolumn{3}{c}{Placebo events} & \\
8 +\cmidrule(lr){2-4}\cmidrule(lr){5-7}
9 +Sample & Neighbors & Others & Diff. & Neighbors & Others & Diff. & Real $-$ placebo \\ \midrule
10 +Full sample & 1,103 & 1,043 & 61.3 (3.4) & 1,004 & 1,006 & -2.7 (3.1) & 64.0 (1.8) \\
11 +1992–1999 & 1,191 & 1,006 & 228.3 (7.7) & 1,121 & 932 & 200.0 (7.9) & 28.3 (6.9) \\
12 +2000–2009 & 1,099 & 1,110 & -10.7 (4.3) & 1,036 & 1,076 & -40.6 (4.1) & 30.0 (2.2) \\
13 +2010–2019 & 1,060 & 1,023 & 37.0 (6.2) & 941 & 985 & -43.7 (5.2) & 80.6 (3.7) \\
14 +2020–2026 & 1,144 & 987 & 157.1 (7.2) & 1,023 & 955 & 67.5 (5.7) & 89.6 (4.3) \\
15 +\bottomrule\end{tabular}
16 +\begin{tablenotes}[flushleft]\footnotesize
17 +\item Notes: Share of receivers with $|z|>2$ five minutes after the event minute, among
18 +the source's twenty lagged-correlation neighbors and among all other receivers, for real
19 +events and for their matched placebos ($\times10^{-4}$). `Diff.' is neighbors minus
20 +others within an arm; the last column is the network differential of
21 +Table~\ref{tab:network_mechanism} (real minus placebo). Day-block bootstrap SEs (500
22 +replications) in parentheses.
23 +\end{tablenotes}\end{threeparttable}\end{adjustbox}\end{table}
\ No newline at end of file
modified paper/tables/table1_sample_construction.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Sample construction by decade}\label{tab:sample_construction}
5 5 \small\begin{tabular}{lrrrr}
modified paper/tables/table2_summary_by_decade.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Summary statistics by decade, top-1000 universe}
5 5 \label{tab:summary_by_decade}
modified paper/tables/table4_baseline_propagation.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/12_stage_f_baseline.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Baseline abnormal propagation after idiosyncratic shocks}
5 5 \label{tab:baseline_propagation}
modified paper/tables/table7_network.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/14_stage_f2_results.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Network mechanism: neighbor minus non-neighbor abnormal response}
5 5 \label{tab:network_mechanism}
modified paper/tables/table8_robustness.tex +1 −1
@@ -1,5 +1,5 @@
1 1 % Auto-generated by scripts/16_robustness.py — do not edit numbers.
2 −\begin{table}[t]\centering\begin{adjustbox}{max width=\textwidth}
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}
3 3 \begin{threeparttable}
4 4 \caption{Robustness of the market-wide null and the network channel}
5 5 \label{tab:robustness}
added paper/tables/table9_design_checks.tex +47 −0
@@ -0,0 +1,47 @@
1 +% Auto-generated by scripts/17_revision_checks.py — do not edit numbers.
2 +\begin{table}[htbp]\centering\begin{adjustbox}{max width=\textwidth}\begin{threeparttable}
3 +\caption{Design diagnostics: pre-event levels, co-arrivals, mean reversion, and clustering}
4 +\label{tab:design_checks}
5 +\small\begin{tabular}{lrrr}
6 +\toprule
7 +\multicolumn{4}{l}{\emph{Panel A: Pre-event levels at $t-5$ ($\times10^{-4}$)}} \\
8 + & Real & Placebo & Real $-$ placebo \\ \midrule
9 +Breadth & 1,044 & 1,004 & 40.25 (1.18) \\
10 +Intensity & 13,226 & 12,488 & 737.77 (115.12) \\
11 +\addlinespace
12 +\multicolumn{4}{l}{\emph{Panel B: By number of source events in the same minute}} \\
13 +Events in minute (share of events) & $\Delta\Delta$Breadth$_{5}$ & $\Delta\Delta$Breadth$_{30}$ & Network diff$_{5}$ \\ \midrule
14 +1 (21.7\%) & 1.74 (1.67) & 3.69 (1.89) & 51.01 (2.55) \\
15 +2 (27.1\%) & -2.97 (1.80) & -2.29 (1.98) & 64.17 (2.67) \\
16 +3–5 (43.6\%) & -3.01 (1.79) & -8.35 (2.13) & 66.87 (2.99) \\
17 +6+ (7.7\%) & 9.12 (5.19) & 1.39 (5.68) & 76.49 (7.40) \\
18 +\addlinespace
19 +\multicolumn{4}{l}{\emph{Panel C: Market-wide breadth, alternative estimands ($\times10^{-4}$)}} \\
20 +Horizon & Post level, real $-$ placebo & $\Delta\Delta$ (baseline) & Pre-level stratified \\ \midrule
21 +$h=5$ & 39.63 (1.28) & -1.04 (0.87) & 19.48 (0.88) \\
22 +$h=15$ & 39.46 (1.36) & -1.46 (1.00) & 19.59 (0.98) \\
23 +$h=30$ & 37.93 (1.42) & -3.10 (1.07) & 18.41 (1.04) \\
24 +\addlinespace
25 +\multicolumn{4}{l}{\emph{Panel D: Bootstrap standard errors by clustering unit ($\times10^{-4}$)}} \\
26 +Statistic & Estimate & Day clusters & Ticker clusters \\ \midrule
27 +$\Delta\Delta$Breadth$_{5}$ & -1.04 & (0.90) & (0.96) \\
28 +$\Delta\Delta$Breadth$_{30}$ & -3.10 & (1.06) & (0.93) \\
29 +Network diff$_{5}$ & 64.02 & (1.89) & (4.72) \\
30 +\addlinespace
31 +\multicolumn{4}{l}{\emph{Panel E: Signed response, real $-$ placebo ($\times10^{-4}$)}} \\
32 +Horizon & $h=1$ & $h=5$ & $h=30$ \\ \midrule
33 +sign$(e_{i,t})\times$ mean $z_{j,t+h}$ & 151.33 (104.90) & 9.57 (91.95) & -65.70 (89.23) \\
34 +\bottomrule\end{tabular}
35 +\begin{tablenotes}[flushleft]\footnotesize
36 +\item Notes: All statistics from the baseline event files (isolated $|z|\ge6$ source
37 +shocks, lagged top-500 universe, 1992--2026; matched within-stock placebos).
38 +Day-block bootstrap standard errors (500 replications) in parentheses unless stated.
39 +Panel A: levels of the outcomes five minutes before the event minute. Panel B: events
40 +classified by the number of source events detected in the same minute (the isolation
41 +filters cap the share of simultaneous $|z|>4$ names at 2\%, not the number of $|z|\ge6$
42 +events); each placebo inherits the class of its real event. Panel C: `post level' is
43 +the real-minus-placebo difference in breadth at $t+h$ without pre-differencing; `pre-level
44 +stratified' reweights the same difference within deciles of the pre-event breadth level.
45 +Panel D: standard errors from resampling trading days versus resampling source tickers.
46 +Panel E: cross-sectional mean standardized return of receivers signed by the source shock.
47 +\end{tablenotes}\end{threeparttable}\end{adjustbox}\end{table}
\ No newline at end of file
modified scripts/02_pilot_figures_tables.py +1 −1
@@ -195,7 +195,7 @@ def latex_table(stats: pl.DataFrame) -> str:
195 195 "2025-12": "December 2025"}
196 196 lines = [
197 197 "% Auto-generated by scripts/02_pilot_figures_tables.py — do not edit numbers by hand.",
198 − "\\begin{table}[t]",
198 + "\\begin{table}[htbp]",
199 199 "\\centering",
200 200 "\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
201 201 "\\caption{Summary statistics, pilot months}",
modified scripts/10_stage_e_figures.py +2 −2
@@ -299,7 +299,7 @@ def table1_sample_construction():
299 299
300 300 lines = [
301 301 "% Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.",
302 − "\\begin{table}[t]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
302 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
303 303 "\\caption{Sample construction by decade}\\label{tab:sample_construction}",
304 304 "\\small\\begin{tabular}{lrrrr}",
305 305 "\\toprule",
@@ -376,7 +376,7 @@ def table2_summary_by_decade():
376 376 fmt = lambda v: f"{v:,.2f}"
377 377 lines = [
378 378 "% Auto-generated by scripts/10_stage_e_figures.py — do not edit numbers.",
379 − "\\begin{table}[t]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
379 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
380 380 "\\caption{Summary statistics by decade, top-1000 universe}",
381 381 "\\label{tab:summary_by_decade}",
382 382 "\\small\\begin{tabular}{llrrrrrrrr}",
modified scripts/12_stage_f_baseline.py +1 −1
@@ -106,7 +106,7 @@ def main() -> None:
106 106 fmt = lambda v, s: f"{v*1e4:,.2f} & ({s*1e4:,.2f})"
107 107 lines = [
108 108 "% Auto-generated by scripts/12_stage_f_baseline.py — do not edit numbers.",
109 − "\\begin{table}[t]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
109 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
110 110 "\\caption{Baseline abnormal propagation after idiosyncratic shocks}",
111 111 "\\label{tab:baseline_propagation}",
112 112 "\\small\\begin{tabular}{lrrrrrrrr}",
modified scripts/14_stage_f2_results.py +1 −1
@@ -106,7 +106,7 @@ def main() -> None:
106 106 # ---- Table VII: network differential by decade (h=15)
107 107 lines = [
108 108 "% Auto-generated by scripts/14_stage_f2_results.py — do not edit numbers.",
109 − "\\begin{table}[t]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
109 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
110 110 "\\caption{Network mechanism: neighbor minus non-neighbor abnormal response}",
111 111 "\\label{tab:network_mechanism}",
112 112 "\\small\\begin{tabular}{lrrr}",
modified scripts/15_build_results_macros.py +70 −0
@@ -109,6 +109,76 @@ def main() -> None:
109 109 f"{cj.filter(pl.col('year')==2020)['severe_per1000'][0]:.1f}")
110 110 macro("SeverePreCovid",
111 111 f"{cj.filter(pl.col('year').is_in([2017,2018,2019]))['severe_per1000'].mean():.2f}")
112 + macro("SevereGFC", f"{cj.filter(pl.col('year')==2008)['severe_per1000'][0]:.1f}")
113 + macro("SevereEarlyNineties",
114 + f"{cj.filter(pl.col('year').is_between(1992,1997))['severe_per1000'].mean():.0f}")
115 + macro("CoJumpSevEarlyNineties",
116 + f"{cj.filter(pl.col('year').is_between(1992,1997))['mean_share4'].mean()*100:.1f}")
117 + macro("CoJumpPerThousandRecent",
118 + f"{cj.filter(pl.col('year').is_between(2021,2025))['per1000'].mean():.0f}")
119 +
120 +
121 + # ---- zero-return share (fig E7 source) and post-decimalization extremes
122 + e7 = pl.read_csv(OUTE / "figures/data/figE7_zeroret.csv")
123 + macro("ZeroRetNinetySix", f"{e7.filter(pl.col('year')==1996)['zero_pct'][0]:.0f}")
124 + macro("ZeroRetTwentyFour", f"{e7.filter(pl.col('year')==2024)['zero_pct'][0]:.0f}")
125 + macro("ExtremeGtOneTwoThousandTwo", num(e5.filter(pl.col("year") == 2002)["gt1"][0] * 1e4, 0))
126 + macro("ExtremeGtOneNinetyTwo", num(e5.filter(pl.col("year") == 1992)["gt1"][0] * 1e4, 0))
127 +
128 + # ---- revision checks (scripts/17_revision_checks.py, 2026-10-17)
129 + rcp = OUTE / "estimates/revision_checks.json"
130 + if rcp.exists():
131 + rc = json.loads(rcp.read_text())
132 + pb = rc["pre_event"]["breadth"]
133 + macro("PreBreadthReal", num(pb["real"] * 1e4, 0))
134 + macro("PreBreadthPlacebo", num(pb["placebo"] * 1e4, 0))
135 + macro("PreBreadthDiff", f"{pb['diff']*1e4:.1f}")
136 + macro("PreBreadthDiffSE", f"{pb['se']*1e4:.1f}")
137 + pi_ = rc["pre_event"]["intensity"]
138 + macro("PreIntensityDiff", f"{pi_['diff']:.3f}")
139 + macro("PreIntensityDiffSE", f"{pi_['se']:.3f}")
140 + br = rc["base_rates"]
141 + macro("BaseBreadthShare", f"{br['placebo_breadth_5']:.2f}")
142 + macro("NEligMean", num(br["nelig_5_mean"], 0))
143 + macro("NEligNineties", num(br["nelig_5_by_decade"]["1992–1999"], 0))
144 + macro("EventsPerStockWeek", f"{br['events_per_stock_week']:.1f}")
145 + macro("EventDays", num(br["event_days"]))
146 + macro("NeighborBaseShare", f"{br['placebo_nbh_5']:.2f}")
147 + macro("MatchRate", f"{br['match_rate']*100:.1f}")
148 + macro("MatchRateNineties", f"{br['match_rate_by_decade']['1992–1999']*100:.0f}")
149 + macro("MatchRateMinPost", f"{min(v for k, v in br['match_rate_by_decade'].items() if k != '1992–1999')*100:.0f}")
150 + macro("CumBaseFifteen", f"{br['placebo_cum_15']*100:.0f}")
151 + mde = 2.8 * f1["baseline"][2]["se_breadth"] # h=5, 5% size / 80% power
152 + macro("MDEFive", f"{mde*1e4:.1f}")
153 + macro("MDEEventsPer", num(1.0 / (mde * br["nelig_5_mean"]), 0))
154 + ca = rc["coarrival"]
155 + macro("ShareSingleton", f"{ca['share_by_class']['1']*100:.0f}")
156 + macro("MeanCoArrival", f"{ca['mean_events_per_event_minute']:.1f}")
157 + cls = {r["cls"]: r for r in ca["by_class"]}
158 + macro("DDSingletonFive", f"{cls['1']['db5']*1e4:.1f}")
159 + macro("DDSingletonFiveSE", f"{cls['1']['db5_se']*1e4:.1f}")
160 + macro("DDSingletonThirty", f"{cls['1']['db30']*1e4:.1f}")
161 + macro("DDSingletonThirtySE", f"{cls['1']['db30_se']*1e4:.1f}")
162 + macro("DDMultiThirty", f"{cls['3–5']['db30']*1e4:.1f}")
163 + macro("DDMultiThirtySE", f"{cls['3–5']['db30_se']*1e4:.1f}")
164 + macro("NetDiffSingleton", f"{cls['1']['nd5']*1e4:.0f}")
165 + macro("NetDiffSingletonSE", f"{cls['1']['nd5_se']*1e4:.1f}")
166 + st = {r["h"]: r for r in rc["stratified"]}
167 + macro("PostDiffFive", f"{st[5]['post_diff']*1e4:.0f}")
168 + macro("StratFive", f"{st[5]['strat']*1e4:.0f}")
169 + macro("StratFiveSE", f"{st[5]['strat_se']*1e4:.1f}")
170 + macro("StratThirty", f"{st[30]['strat']*1e4:.0f}")
171 + macro("StratThirtySE", f"{st[30]['strat_se']*1e4:.1f}")
172 + macro("StratShareOfBase", f"{st[5]['strat']/br['placebo_breadth_5']*100:.0f}")
173 + macro("NetDiffFiveTickerSE", f"{rc['cluster']['nd5']['se_ticker']*1e4:.1f}")
174 + bs = {r["sign"]: r for r in rc["network_by_sign"]}
175 + macro("NetDiffNeg", f"{bs['negative']['dd']*1e4:.0f}")
176 + macro("NetDiffPos", f"{bs['positive']['dd']*1e4:.0f}")
177 + nc = {r["scope"]: r for r in rc["network_components"]}
178 + macro("NetRealNdNineties", f"{nc['1992–1999']['real_nd_5']*1e4:.0f}")
179 + macro("NetPlaceboNdNineties", f"{nc['1992–1999']['placebo_nd_5']*1e4:.0f}")
180 + macro("NetRealNdTens", f"{nc['2010–2019']['real_nd_5']*1e4:.0f}")
181 + macro("NetPlaceboNdTens", f"{nc['2010–2019']['placebo_nd_5']*1e4:.0f}")
112 182
113 183 tex = "% Auto-generated by scripts/15_build_results_macros.py — DO NOT EDIT.\n"
114 184 tex += "\n".join(M) + "\n"
modified scripts/16_robustness.py +1 −1
@@ -139,7 +139,7 @@ def main() -> None:
139 139 return f"{v*1e4:,.2f} ({s*1e4:,.2f})"
140 140 lines = [
141 141 "% Auto-generated by scripts/16_robustness.py — do not edit numbers.",
142 − "\\begin{table}[t]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
142 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
143 143 "\\caption{Robustness of the market-wide null and the network channel}",
144 144 "\\label{tab:robustness}",
145 145 "\\small\\begin{tabular}{lrrr}",
added scripts/17_revision_checks.py +335 −0
@@ -0,0 +1,335 @@
1 +"""Revision checks (editorial-board review, 2026-10-17) from stored event files.
2 +
3 +Everything here is recomputed from outputs/estimates/events*/ (stage F/F2
4 +outputs); no raw minute data are needed. Day-block bootstrap, B=500, same seed.
5 +
6 + A. Pre-event levels: real vs placebo breadth/intensity at t-5 (the "no
7 + pre-event response" falsification quantified; by decade as well).
8 + B. Base rates: placebo breadth, eligible receivers, neighbor / non-neighbor
9 + levels by arm and decade (components of the network differential).
10 + C. Co-arrivals: share of real events that share their minute with other
11 + events; network differential and market-wide ΔΔ by co-arrival class.
12 + D. Pre-level-stratified post comparison (mean-reversion check): real minus
13 + placebo breadth at t+h within deciles of the pre-event breadth level.
14 + E. Alternative clustering: ticker-cluster bootstrap SEs.
15 + F. Signed response (real minus placebo), network differential by shock sign.
16 +
17 +Outputs: outputs/estimates/revision_checks.json,
18 + outputs/tables/table9_design_checks.{tex,csv},
19 + outputs/tables/table10_network_components.{tex,csv}
20 + (+ copies in paper/tables/).
21 +"""
22 +
23 +from __future__ import annotations
24 +
25 +import json
26 +from pathlib import Path
27 +
28 +import numpy as np
29 +import polars as pl
30 +
31 +ROOT = Path(__file__).resolve().parents[1]
32 +EV = ROOT / "outputs" / "estimates" / "events"
33 +EV2 = ROOT / "outputs" / "estimates" / "events2"
34 +TAB = ROOT / "outputs" / "tables"
35 +PTAB = ROOT / "paper" / "tables"
36 +B = 500
37 +SEED = 20260713
38 +DEC = {"1992–1999": (1992, 1999), "2000–2009": (2000, 2009),
39 + "2010–2019": (2010, 2019), "2020–2026": (2020, 2026)}
40 +
41 +
42 +def load(folder: Path, prefix: str) -> pl.DataFrame:
43 + """Concatenate monthly event files, keeping the real→placebo row order and
44 + labelling each placebo with the minute of the real event it matches."""
45 + frames = []
46 + for p in sorted(folder.glob(f"{prefix}_*.parquet")):
47 + if p.stat().st_size <= 500:
48 + continue
49 + d = pl.read_parquet(p).with_row_index("i")
50 + prev = d.select((pl.col("i") + 1).alias("i"), pl.col("kind").alias("pk"),
51 + pl.col("ticker").alias("pt"), pl.col("z").alias("pz"),
52 + pl.col("date").alias("ev_date"), pl.col("mod").alias("ev_mod"))
53 + d = d.join(prev, on="i", how="left")
54 + ok = (pl.col("kind") == "real") | (
55 + (pl.col("pk") == "real") & (pl.col("pt") == pl.col("ticker"))
56 + & (pl.col("pz") == pl.col("z")))
57 + d = d.filter(ok).with_columns(
58 + pl.when(pl.col("kind") == "real").then(pl.col("date")).otherwise(pl.col("ev_date")).alias("ev_date"),
59 + pl.when(pl.col("kind") == "real").then(pl.col("mod")).otherwise(pl.col("ev_mod")).alias("ev_mod"),
60 + ).drop("i", "pk", "pt", "pz")
61 + frames.append(d)
62 + df = pl.concat(frames, how="diagonal")
63 + return df.with_columns(pl.col("date").str.slice(0, 4).cast(pl.Int32).alias("year"))
64 +
65 +
66 +def dd_boot(df: pl.DataFrame, col: str, rng, cluster: str = "date"):
67 + """Real-minus-placebo difference of means with cluster bootstrap SE."""
68 + g = (df.filter(pl.col(col).is_not_null())
69 + .group_by(cluster, "kind")
70 + .agg(pl.col(col).sum().alias("s"), pl.len().alias("n"))
71 + .pivot(on="kind", index=cluster, values=["s", "n"]))
72 + need = ("s_real", "n_real", "s_placebo", "n_placebo")
73 + if any(c not in g.columns for c in need) or g.height < 10:
74 + return np.nan, np.nan, 0
75 + g = g.fill_null(0).sort(cluster) # deterministic cluster order → reproducible draws
76 + sr, nr = g["s_real"].to_numpy(), g["n_real"].to_numpy()
77 + sp, npb = g["s_placebo"].to_numpy(), g["n_placebo"].to_numpy()
78 + est = sr.sum() / max(nr.sum(), 1) - sp.sum() / max(npb.sum(), 1)
79 + idx = rng.integers(0, len(sr), size=(B, len(sr)))
80 + bs = (sr[idx].sum(1) / np.maximum(nr[idx].sum(1), 1)
81 + - sp[idx].sum(1) / np.maximum(npb[idx].sum(1), 1))
82 + return float(est), float(bs.std()), int(nr.sum())
83 +
84 +
85 +def level_boot(df: pl.DataFrame, col: str, kind: str, rng):
86 + g = (df.filter((pl.col("kind") == kind) & pl.col(col).is_not_null())
87 + .group_by("date").agg(pl.col(col).sum().alias("s"), pl.len().alias("n")).sort("date"))
88 + s, n = g["s"].to_numpy(), g["n"].to_numpy()
89 + est = s.sum() / max(n.sum(), 1)
90 + idx = rng.integers(0, len(s), size=(B, len(s)))
91 + bs = s[idx].sum(1) / np.maximum(n[idx].sum(1), 1)
92 + return float(est), float(bs.std()), int(n.sum())
93 +
94 +
95 +def stratified_boot(df: pl.DataFrame, post: str, pre: str, nbins: int, rng):
96 + """Real minus placebo mean of `post` within quantile bins of `pre`,
97 + weighted by the real-arm bin counts; day bootstrap."""
98 + d = df.filter(pl.col(post).is_not_null() & pl.col(pre).is_not_null())
99 + qs = [d[pre].quantile(q / nbins) for q in range(1, nbins)]
100 + d = d.with_columns(pl.col(pre).cut(qs, labels=[str(k) for k in range(nbins)]).alias("bin"))
101 + g = (d.group_by("date", "bin", "kind")
102 + .agg(pl.col(post).sum().alias("s"), pl.len().alias("n")))
103 + dates = sorted(g["date"].unique().to_list())
104 + di = {x: k for k, x in enumerate(dates)}
105 + S = {k: np.zeros((len(dates), nbins)) for k in ("real", "placebo")}
106 + N = {k: np.zeros((len(dates), nbins)) for k in ("real", "placebo")}
107 + for r in g.iter_rows(named=True):
108 + S[r["kind"]][di[r["date"]], int(r["bin"])] += r["s"]
109 + N[r["kind"]][di[r["date"]], int(r["bin"])] += r["n"]
110 +
111 + def stat(ix):
112 + sr, nr = S["real"][ix].sum(0), N["real"][ix].sum(0)
113 + sp, npb = S["placebo"][ix].sum(0), N["placebo"][ix].sum(0)
114 + w = nr / nr.sum()
115 + return float(np.sum(w * (sr / np.maximum(nr, 1) - sp / np.maximum(npb, 1))))
116 +
117 + est = stat(np.arange(len(dates)))
118 + idx = rng.integers(0, len(dates), size=(B, len(dates)))
119 + bs = np.array([stat(idx[b]) for b in range(B)])
120 + return est, float(bs.std()), int(N["real"].sum())
121 +
122 +
123 +def main() -> None:
124 + rng = np.random.default_rng(SEED)
125 + e1 = load(EV, "events").with_columns(
126 + *[(pl.col(f"breadth_{h}") - pl.col("pre_breadth")).alias(f"db_{h}") for h in (5, 15, 30)],
127 + (pl.col("intensity_5") - pl.col("pre_intensity")).alias("di_5"))
128 + e2 = load(EV2, "events2").with_columns(
129 + *[(pl.col(f"nbh_{h}") - pl.col(f"non_{h}")).alias(f"nd_{h}") for h in (5, 15, 30)])
130 + real1 = e1.filter(pl.col("kind") == "real")
131 + print(f"events: {real1.height:,} real, {e1.height - real1.height:,} paired placebos")
132 +
133 + out: dict = {"params": {"bootstrap": B, "seed": SEED}}
134 +
135 + # ---------------- A. pre-event levels
136 + pre = {}
137 + for col, lab in [("pre_breadth", "breadth"), ("pre_intensity", "intensity")]:
138 + r = level_boot(e1, col, "real", rng)
139 + p = level_boot(e1, col, "placebo", rng)
140 + dlt = dd_boot(e1, col, rng)
141 + pre[lab] = {"real": r[0], "placebo": p[0], "diff": dlt[0], "se": dlt[1], "n": dlt[2]}
142 + print(f"pre {lab}: real {r[0]:.5f} placebo {p[0]:.5f} diff {dlt[0]*1e4:+.2f} ({dlt[1]*1e4:.2f}) x1e-4")
143 + pre["breadth_by_decade"] = []
144 + for dec, (a, b) in DEC.items():
145 + sub = e1.filter(pl.col("year").is_between(a, b))
146 + dlt = dd_boot(sub, "pre_breadth", rng)
147 + pre["breadth_by_decade"].append({"dec": dec, "diff": dlt[0], "se": dlt[1], "n": dlt[2]})
148 + out["pre_event"] = pre
149 +
150 + # ---------------- B. base rates and network components
151 + base = {
152 + "placebo_breadth_5": float(e1.filter(pl.col("kind") == "placebo")["breadth_5"].mean()),
153 + "placebo_breadth_30": float(e1.filter(pl.col("kind") == "placebo")["breadth_30"].mean()),
154 + "nelig_5_mean": float(real1["nelig_5"].mean()),
155 + "nelig_5_by_decade": {d: float(real1.filter(pl.col("year").is_between(a, b))["nelig_5"].mean())
156 + for d, (a, b) in DEC.items()},
157 + "placebo_nbh_5": float(e2.filter(pl.col("kind") == "placebo")["nbh_5"].mean()),
158 + "placebo_non_5": float(e2.filter(pl.col("kind") == "placebo")["non_5"].mean()),
159 + "event_days": int(real1["date"].n_unique()),
160 + "match_rate": float((e1.height - real1.height) / real1.height),
161 + "match_rate_by_decade": {d: float(e1.filter(pl.col("year").is_between(a, b) & (pl.col("kind") == "placebo")).height
162 + / max(real1.filter(pl.col("year").is_between(a, b)).height, 1))
163 + for d, (a, b) in DEC.items()},
164 + "placebo_cum_15": float(e2.filter(pl.col("kind") == "placebo")["cum_15"].mean()),
165 + "events_per_stock_week": float(real1.height / (500 * real1["date"].n_unique() / 5)),
166 + }
167 + comp = []
168 + for scope, (a, b) in [("all", (1992, 2026))] + list(DEC.items()):
169 + sub = e2.filter(pl.col("year").is_between(a, b))
170 + row = {"scope": scope}
171 + for kind in ("real", "placebo"):
172 + for col in ("nbh_5", "non_5", "nd_5"):
173 + est, se, n = level_boot(sub, col, kind, rng)
174 + row[f"{kind}_{col}"] = est
175 + row[f"{kind}_{col}_se"] = se
176 + dd, se, n = dd_boot(sub, "nd_5", rng)
177 + row.update({"dd": dd, "dd_se": se, "n": n})
178 + comp.append(row)
179 + print(f"net components {scope:10} real nd {row['real_nd_5']*1e4:+.1f} placebo nd {row['placebo_nd_5']*1e4:+.1f} dd {dd*1e4:+.1f} ({se*1e4:.1f})")
180 + out["base_rates"] = base
181 + out["network_components"] = comp
182 +
183 + # ---------------- C. co-arrivals
184 + cnt = real1.group_by("date", "mod").len().rename({"len": "n_co"})
185 + e1c = e1.join(cnt, left_on=["ev_date", "ev_mod"], right_on=["date", "mod"], how="left")
186 + e2c = e2.join(cnt, left_on=["ev_date", "ev_mod"], right_on=["date", "mod"], how="left")
187 + dist = (real1.join(cnt, on=["date", "mod"], how="left")
188 + .with_columns(pl.when(pl.col("n_co") == 1).then(pl.lit("1"))
189 + .when(pl.col("n_co") == 2).then(pl.lit("2"))
190 + .when(pl.col("n_co") <= 5).then(pl.lit("3–5"))
191 + .otherwise(pl.lit("6+")).alias("cls"))
192 + .group_by("cls").len().sort("cls"))
193 + tot = dist["len"].sum()
194 + coarr = {"share_by_class": {r["cls"]: r["len"] / tot for r in dist.iter_rows(named=True)},
195 + "mean_events_per_event_minute": float(real1.height / cnt.height),
196 + "by_class": []}
197 + classes = {"1": (1, 1), "2": (2, 2), "3–5": (3, 5), "6+": (6, 10**6)}
198 + for cls, (lo, hi) in classes.items():
199 + s1 = e1c.filter(pl.col("n_co").is_between(lo, hi))
200 + s2 = e2c.filter(pl.col("n_co").is_between(lo, hi))
201 + row = {"cls": cls, "share": coarr["share_by_class"].get(cls, 0.0)}
202 + for h in (5, 30):
203 + est, se, n = dd_boot(s1, f"db_{h}", rng)
204 + row[f"db{h}"], row[f"db{h}_se"], row["n"] = est, se, n
205 + for h in (5, 15, 30):
206 + est, se, n2 = dd_boot(s2, f"nd_{h}", rng)
207 + row[f"nd{h}"], row[f"nd{h}_se"] = est, se
208 + coarr["by_class"].append(row)
209 + print(f"co-arrival {cls:4} share {row['share']:.3f} ΔΔb5 {row['db5']*1e4:+.2f} ({row['db5_se']*1e4:.2f}) nd5 {row['nd5']*1e4:+.1f} ({row['nd5_se']*1e4:.1f})")
210 + out["coarrival"] = coarr
211 +
212 + # ---------------- D. pre-level-stratified post comparison
213 + strat = []
214 + for h in (5, 15, 30):
215 + raw = dd_boot(e1, f"breadth_{h}", rng) # post level, real − placebo
216 + dd = dd_boot(e1, f"db_{h}", rng) # ΔΔ (baseline estimand)
217 + st = stratified_boot(e1, f"breadth_{h}", "pre_breadth", 10, rng)
218 + strat.append({"h": h, "post_diff": raw[0], "post_se": raw[1], "dd": dd[0], "dd_se": dd[1],
219 + "strat": st[0], "strat_se": st[1], "n": st[2]})
220 + print(f"h={h}: post diff {raw[0]*1e4:+.2f} ({raw[1]*1e4:.2f}); ΔΔ {dd[0]*1e4:+.2f} ({dd[1]*1e4:.2f}); pre-stratified {st[0]*1e4:+.2f} ({st[1]*1e4:.2f})")
221 + out["stratified"] = strat
222 +
223 + # ---------------- E. ticker-cluster bootstrap
224 + clus = {}
225 + for lab, df, col in [("db5", e1, "db_5"), ("db30", e1, "db_30"), ("nd5", e2, "nd_5")]:
226 + d_est, d_se, _ = dd_boot(df, col, rng, cluster="date")
227 + t_est, t_se, _ = dd_boot(df, col, rng, cluster="ticker")
228 + clus[lab] = {"est": d_est, "se_day": d_se, "se_ticker": t_se}
229 + print(f"{lab}: est {d_est*1e4:+.2f} day-SE {d_se*1e4:.2f} ticker-SE {t_se*1e4:.2f}")
230 + out["cluster"] = clus
231 +
232 + # ---------------- F. signed response and network differential by sign
233 + signed = []
234 + for h in (1, 5, 15, 30):
235 + est, se, n = dd_boot(e1, f"signed_{h}", rng)
236 + signed.append({"h": h, "diff": est, "se": se, "n": n})
237 + print(f"signed h={h}: {est*1e4:+.2f} ({se*1e4:.2f}) x1e-4")
238 + out["signed"] = signed
239 + bysign = []
240 + for sgn, lab in [(-1, "negative"), (1, "positive")]:
241 + est, se, n = dd_boot(e2.filter(pl.col("sign") == sgn), "nd_5", rng)
242 + bysign.append({"sign": lab, "dd": est, "se": se, "n": n})
243 + print(f"network diff, {lab} shocks: {est*1e4:+.1f} ({se*1e4:.1f})")
244 + out["network_by_sign"] = bysign
245 +
246 + (ROOT / "outputs" / "estimates" / "revision_checks.json").write_text(json.dumps(out, indent=1))
247 + (ROOT / "outputs" / "diagnostics" / "revision_checks.json").write_text(json.dumps(out, indent=1))
248 +
249 + # ---------------- Table 9: design diagnostics
250 + f4 = lambda v, s: f"{v*1e4:,.2f} ({s*1e4:,.2f})"
251 + rows9 = []
252 + L = ["% Auto-generated by scripts/17_revision_checks.py — do not edit numbers.",
253 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
254 + "\\caption{Design diagnostics: pre-event levels, co-arrivals, mean reversion, and clustering}",
255 + "\\label{tab:design_checks}",
256 + "\\small\\begin{tabular}{lrrr}", "\\toprule"]
257 + L += ["\\multicolumn{4}{l}{\\emph{Panel A: Pre-event levels at $t-5$ ($\\times10^{-4}$)}} \\\\",
258 + " & Real & Placebo & Real $-$ placebo \\\\ \\midrule"]
259 + for lab, key in [("Breadth", "breadth"), ("Intensity", "intensity")]:
260 + r = pre[key]
261 + L.append(f"{lab} & {r['real']*1e4:,.0f} & {r['placebo']*1e4:,.0f} & {f4(r['diff'], r['se'])} \\\\")
262 + rows9.append({"panel": "A", "row": lab, "real": r["real"], "placebo": r["placebo"], "diff": r["diff"], "se": r["se"]})
263 + L += ["\\addlinespace", "\\multicolumn{4}{l}{\\emph{Panel B: By number of source events in the same minute}} \\\\",
264 + "Events in minute (share of events) & $\\Delta\\Delta$Breadth$_{5}$ & $\\Delta\\Delta$Breadth$_{30}$ & Network diff$_{5}$ \\\\ \\midrule"]
265 + for r in coarr["by_class"]:
266 + L.append(f"{r['cls']} ({r['share']*100:.1f}\\%) & {f4(r['db5'], r['db5_se'])} & {f4(r['db30'], r['db30_se'])} & {f4(r['nd5'], r['nd5_se'])} \\\\")
267 + rows9.append({"panel": "B", "row": r["cls"], "share": r["share"], "db5": r["db5"], "db5_se": r["db5_se"],
268 + "db30": r["db30"], "db30_se": r["db30_se"], "nd5": r["nd5"], "nd5_se": r["nd5_se"]})
269 + L += ["\\addlinespace", "\\multicolumn{4}{l}{\\emph{Panel C: Market-wide breadth, alternative estimands ($\\times10^{-4}$)}} \\\\",
270 + "Horizon & Post level, real $-$ placebo & $\\Delta\\Delta$ (baseline) & Pre-level stratified \\\\ \\midrule"]
271 + for r in strat:
272 + L.append(f"$h={r['h']}$ & {f4(r['post_diff'], r['post_se'])} & {f4(r['dd'], r['dd_se'])} & {f4(r['strat'], r['strat_se'])} \\\\")
273 + rows9.append({"panel": "C", "row": r["h"], "post": r["post_diff"], "post_se": r["post_se"], "dd": r["dd"],
274 + "dd_se": r["dd_se"], "strat": r["strat"], "strat_se": r["strat_se"]})
275 + L += ["\\addlinespace", "\\multicolumn{4}{l}{\\emph{Panel D: Bootstrap standard errors by clustering unit ($\\times10^{-4}$)}} \\\\",
276 + "Statistic & Estimate & Day clusters & Ticker clusters \\\\ \\midrule"]
277 + for lab, key in [("$\\Delta\\Delta$Breadth$_{5}$", "db5"), ("$\\Delta\\Delta$Breadth$_{30}$", "db30"), ("Network diff$_{5}$", "nd5")]:
278 + r = clus[key]
279 + L.append(f"{lab} & {r['est']*1e4:,.2f} & ({r['se_day']*1e4:,.2f}) & ({r['se_ticker']*1e4:,.2f}) \\\\")
280 + rows9.append({"panel": "D", "row": key, "est": r["est"], "se_day": r["se_day"], "se_ticker": r["se_ticker"]})
281 + L += ["\\addlinespace", "\\multicolumn{4}{l}{\\emph{Panel E: Signed response, real $-$ placebo ($\\times10^{-4}$)}} \\\\",
282 + "Horizon & $h=1$ & $h=5$ & $h=30$ \\\\ \\midrule"]
283 + sg = {r["h"]: r for r in signed}
284 + L.append("sign$(e_{i,t})\\times$ mean $z_{j,t+h}$ & " + " & ".join(f4(sg[h]["diff"], sg[h]["se"]) for h in (1, 5, 30)) + " \\\\")
285 + L += ["\\bottomrule\\end{tabular}",
286 + "\\begin{tablenotes}[flushleft]\\footnotesize",
287 + "\\item Notes: All statistics from the baseline event files (isolated $|z|\\ge6$ source",
288 + "shocks, lagged top-500 universe, 1992--2026; matched within-stock placebos).",
289 + "Day-block bootstrap standard errors (500 replications) in parentheses unless stated.",
290 + "Panel A: levels of the outcomes five minutes before the event minute. Panel B: events",
291 + "classified by the number of source events detected in the same minute (the isolation",
292 + "filters cap the share of simultaneous $|z|>4$ names at 2\\%, not the number of $|z|\\ge6$",
293 + "events); each placebo inherits the class of its real event. Panel C: `post level' is",
294 + "the real-minus-placebo difference in breadth at $t+h$ without pre-differencing; `pre-level",
295 + "stratified' reweights the same difference within deciles of the pre-event breadth level.",
296 + "Panel D: standard errors from resampling trading days versus resampling source tickers.",
297 + "Panel E: cross-sectional mean standardized return of receivers signed by the source shock.",
298 + "\\end{tablenotes}\\end{threeparttable}\\end{adjustbox}\\end{table}"]
299 + (TAB / "table9_design_checks.tex").write_text("\n".join(L))
300 + (PTAB / "table9_design_checks.tex").write_text("\n".join(L))
301 + pl.DataFrame(rows9, strict=False).write_csv(TAB / "table9_design_checks.csv")
302 +
303 + # ---------------- Table 10: network components by decade
304 + L = ["% Auto-generated by scripts/17_revision_checks.py — do not edit numbers.",
305 + "\\begin{table}[htbp]\\centering\\begin{adjustbox}{max width=\\textwidth}\\begin{threeparttable}",
306 + "\\caption{Components of the network differential at $h=5$, by decade}",
307 + "\\label{tab:network_components}",
308 + "\\small\\begin{tabular}{lrrrrrrr}", "\\toprule",
309 + " & \\multicolumn{3}{c}{Real events} & \\multicolumn{3}{c}{Placebo events} & \\\\",
310 + "\\cmidrule(lr){2-4}\\cmidrule(lr){5-7}",
311 + "Sample & Neighbors & Others & Diff. & Neighbors & Others & Diff. & Real $-$ placebo \\\\ \\midrule"]
312 + rows10 = []
313 + for r in comp:
314 + lab = "Full sample" if r["scope"] == "all" else r["scope"]
315 + L.append(f"{lab} & {r['real_nbh_5']*1e4:,.0f} & {r['real_non_5']*1e4:,.0f} & {r['real_nd_5']*1e4:,.1f} ({r['real_nd_5_se']*1e4:,.1f})"
316 + f" & {r['placebo_nbh_5']*1e4:,.0f} & {r['placebo_non_5']*1e4:,.0f} & {r['placebo_nd_5']*1e4:,.1f} ({r['placebo_nd_5_se']*1e4:,.1f})"
317 + f" & {r['dd']*1e4:,.1f} ({r['dd_se']*1e4:,.1f}) \\\\")
318 + rows10.append(r)
319 + L += ["\\bottomrule\\end{tabular}",
320 + "\\begin{tablenotes}[flushleft]\\footnotesize",
321 + "\\item Notes: Share of receivers with $|z|>2$ five minutes after the event minute, among",
322 + "the source's twenty lagged-correlation neighbors and among all other receivers, for real",
323 + "events and for their matched placebos ($\\times10^{-4}$). `Diff.' is neighbors minus",
324 + "others within an arm; the last column is the network differential of",
325 + "Table~\\ref{tab:network_mechanism} (real minus placebo). Day-block bootstrap SEs (500",
326 + "replications) in parentheses.",
327 + "\\end{tablenotes}\\end{threeparttable}\\end{adjustbox}\\end{table}"]
328 + (TAB / "table10_network_components.tex").write_text("\n".join(L))
329 + (PTAB / "table10_network_components.tex").write_text("\n".join(L))
330 + pl.DataFrame(rows10).write_csv(TAB / "table10_network_components.csv")
331 + print("revision checks complete")
332 +
333 +
334 +if __name__ == "__main__":
335 + main()
336