🔬 Review status
This working paper has not yet gone through the multi-referee review of the UQO working-paper series (blinded referee reports, editorial decision, re-execution, full revision, independent verification). Its results should be read as preliminary. Only reviewed papers of the series are presented on wp.spboucher.ai and www.spboucher.ai.
The Yield Curve at One-Minute Resolution
Title: The Yield Curve at One-Minute Resolution: Nelson–Siegel Factor-Mimicking Portfolios from Treasury Futures, 2008–2026
Author (paper): Guillaume Manseau (Université Laval) — working paper without series number · Project ID: wp47_uqo (internal) ·
Repo: gitsrv:srv/git/wp47_uqo.git · Status: v1.0-rc, paper/main.pdf 45 pages (2026-11-09), after one design review and three
referee-style reviews (see docs/).
Question. What do the three Nelson–Siegel factors (level, slope, curvature) look like minute by minute over the 23-hour CME session, how much of the daily curve variation can Treasury futures reproduce, where in the day and on which days is yield-curve factor risk realized, and does an announcement move last?
Data. FinData OS (FirstRate Data, private licence): 1-minute bars of every individual Treasury futures contract for ZT, ZF, ZN,
TN, US, UB (2008–2026; most-traded contract per session, no roll splice), plus SR3 and five Treasury ETFs for robustness; public:
FRED constant-maturity yields, Fed GSW curve, FOMC and BLS/ALFRED release calendars (data/external/, downloaded 2026-11-09).
Construction. Daily NS factors (λ = 0.7308) from 11 CMT maturities → rolling quarterly OLS of daily factor changes on daily futures returns (3-year window, strictly out of sample) → weights applied to one-minute returns → minute-level factor changes on 3 963 yield days (5.0 M minutes, 2011–2026), realized (co)variances, event windows, local projections, 5-minute VAR, volatility connectedness, HAR forecasts.
Headline numbers (current results/metrics/). Cumulated minute series vs daily CMT factor changes: R² 0.93 / 0.80 / 0.75
(level / slope / curvature); implied 2y–30y yields 0.86–0.94, 3-month yield 0.09 (the futures do not see the bill segment or the
NS short rate). 30-minute FOMC window = 18.5 % of the day's realized level variance, Employment report 30.4 %, benchmark 2.2 %;
since 2022 CPI days move the 2-year sector and the curvature more than FOMC days. FOMC 2-year moves persist (1.67 bp per bp at
10 days), CPI moves partly reverse (0.62). 1-minute level–slope correlation −0.93 vs tick-noise benchmark −0.84. HAR beats GARCH(1,1)
under QLIKE (7/9 Holm-significant) but not under MSE (1/9). Volatility connectedness 56.5 %, curvature a net receiver.
Hypotheses (pre-registered in docs/DESIGN.md, tag design-frozen). H1 fidelity · H2 announcement anatomy · H3 intraday
periodicity · H4 realized factor correlations · H5 HAR vs daily benchmarks. Outcomes and every deviation from the plan:
docs/CHANGELOG-DESIGN.md, paper Appendix A.
Review workflow. Design review before the event study (docs/DESIGN_REVIEW.md → docs/DESIGN_RESPONSE.md); three referee-style
reports on the first complete draft (docs/REFEREE_econometric.md, docs/REFEREE_integrity.md, docs/REFEREE_exhibits.md) →
point-by-point response and list of what was not done (docs/RESPONSE_TO_REFEREES.md). The main consequence of the reviews was the
rebuild of the return panel from individual contracts.
Layout. src/wp47/ library · scripts/ pipeline (01–14, order in docs/REPLICATION.md) · data/external/ public inputs ·
data/cache/ node-only Parquet (git-ignored) · results/metrics/ aggregates · results/figdata/ figure inputs · paper/ LaTeX
(12 figures, 13 tables, references.bib with 52 Crossref-verified entries) · docs/ brief, design, reviews, responses, changelog,
replication, data dictionary, methodology decisions (D-01…D-25), numbers index.
Run. Computation on M3U96a ~/wp47_uqo (see docs/REPLICATION.md); tables, text macros, figures and LaTeX on the laptop
(cd paper && latexmk -pdf main.tex).
Licence. No raw or derived minute bars leave the node or enter git; only aggregates are published.