QRA Watch

Treasury Maturity Composition Research

View the public site

QRA Watch is an open, reproducible research and data product measuring whether Treasury's bill-versus-coupon mix operates like a shadow form of balance-sheet policy — changing public duration supply, term premia, and market plumbing. All data is sourced from public APIs and official filings: FiscalData, FRED, Treasury QRA documents, and SEC EDGAR.

The project stays neutral about motive. It treats the underlying proposition as a testable empirical question, then maps it into official measurement, duration/plumbing mechanisms, and reduced-form pricing.

Claim boundary

The project maintains an explicit hierarchy between what the current evidence can and cannot support:

What the evidence supports: Official maturity-composition measurement across the current public history window. Exact official quarter coverage currently spans 2009Q1 through 2025Q4. Headline plumbing and duration supply regressions at headline readiness. A narrow current-sample financing pilot with 14 current-sample financing components, 6 verified pre-release external benchmarks, 5 Tier A components, 8 source-family-exhausted blocked rows, and 0 open benchmark candidates.

What the evidence does not establish: A settled or full-sample causal estimate of Treasury issuance effects on long rates. The pricing layer remains supporting/provisional reduced-form evidence. Scenario translations are illustrative only.

Evidence hierarchy

Lane	Description	Status
Headline measurement & mechanism	Official maturity composition, plumbing regressions, public duration supply	Headline ready
Reduced-form context	Monthly carry-forward Maturity-Tilt Flow specification	Supporting/provisional
Reduced-form credibility anchor	Unique-release fixed-horizon +63bd flow design	Supporting/provisional
Bounded causal pilot	Post-2022Q3 financing-estimates event design	Supporting, bounded
Supporting extensions	Investor allotments, primary dealer, SEC N-MFP	Summary ready

The release-level +63bd design is the credibility anchor for the pricing layer. The monthly flow spec provides supporting context with a longer sample but weaker identification. The causal pilot is narrower and supporting — not the main pricing coefficient source in this round.

Key terminology

• Maturity Tilt — the quarter-level bill-versus-coupon composition object from official Treasury financing arithmetic
• Maturity-Tilt Flow — the signed quarterly flow object measuring coupon shortfall relative to baseline
• Excess Bills Stock — the cumulative stock object derived from the flow series
• Public Duration Supply — the weekly duration-supply construction used in the duration and pricing layers

See docs/DATA_DICTIONARY.md for internal field-level definitions.

Site architecture

The public site at site/index.html is a static frontend (HTML + CSS + vanilla JS) that reads only from published backend artifacts:

• site/data/ — JSON and CSV artifacts mirrored from output/publish/
• site/figures/ — SVG figures mirrored from output/figures/
• site/data/index.json — the artifact manifest used for conditional rendering

The frontend never accesses raw or interim backend data. Numeric values, status counts, coverage windows, claim boundaries, and evidence-hierarchy roles are loaded from backend-generated artifacts at runtime. Some display labels, section prose, and model display ordering remain in frontend code. make site mirrors the current output/publish/ artifacts into the site bundle.

Quickstart

python3 -m venv ~/venvs/qrawatch
"$HOME/venvs/qrawatch/bin/python" -m ensurepip --upgrade
"$HOME/venvs/qrawatch/bin/python" -m pip install -r requirements.txt
"$HOME/venvs/qrawatch/bin/python" -m pip install -e .

If you have a local .env file with environment overrides (e.g. VIRTUAL_ENV), source it before running make commands.

With network access:

make bootstrap
make test
make regenerate
make pricing-figures
make site

make regenerate runs the full backend pipeline. make site mirrors publish artifacts into the frontend bundle.

Repository map

• README.md — project overview, claim boundary, and evidence hierarchy
• DATA_SOURCES.md — source registry
• docs/PRICING_METHODS.md — pricing estimands, panel design, interpretation boundaries
• docs/PRICING_RESULTS_MEMO.md — current coefficients and claim boundary
• docs/BENCHMARK_SEARCH_CLOSURE.md — bounded benchmark-search closure memo
• docs/STATUS_GLOSSARY.md — readiness/source-quality labels
• docs/DATA_DICTIONARY.md — field-level definitions
• docs/REGRESSION_SPECS.md — regression equations and layouts
• docs/QRA_CAPTURE_PROTOCOL.md — official quarter-capture workflow
• src/ati_shadow_policy/ — reusable modules
• scripts/ — pipeline entry points
• tests/ — unit tests

Sign conventions

• Positive Maturity-Tilt Flow = more bill-heavy financing relative to the 18% baseline
• Positive Public Duration Supply = more duration pushed into private hands
• Positive QT contribution = Fed Treasury holdings falling
• Buybacks enter the duration construction with opposite sign (they remove duration from public hands)

What is intentionally provisional

• The pricing layer is reduced-form and published as supporting/provisional
• Extension modules remain supporting context, not headline evidence
• The duration headline combines exact non-bill net supply with a QT proxy
• The QRA event and causal layers remain supporting audit surfaces
• Pre-2009Q1 historical extension remains out of scope

What comes next

• Keep the QRA causal lane bounded rather than reopening broad benchmark hunts
• Freeze the pricing claim boundary after the completed-month monthly lock
• Shift effort toward frontend communication if no remaining backend task moves the claim boundary materially