Macro Liquidity Monitor — USD Funding & Repo Plumbing

The one question this answers

Is the banking system's overnight funding loose, normal, tightening, or stressed — and what's the next pressure point?

This is NOT the equity-warning radar (macro-warning). That one watches valuation / VIX / sentiment. This one watches the plumbing: where overnight cash actually clears, how much buffer is left, and whether anyone is tapping the Fed's backstop.

Two lenses, both served by the same data:

• "Too loose → bubble" (the SLR-relief / deregulation worry): watch for SOFR

printing below IORB while the RRP buffer is gone — cash with nowhere to go.

• "Too tight → funding stress" (a repo crisis): watch for SOFR above IORB,

reserves near LCLoR, and SRF takeup spiking.

When to invoke

• Manual: "流动性怎么样 / liquidity check / SOFR-IORB / when does liquidity tighten".
• Twice-daily batch (GitHub Actions, weekdays): 08:30 ET morning preview

(prior-day SOFR/EFFR) + 16:30 ET post-close confirm (same-day ON RRP + SRF). Both push a Telegram digest; a 🔔 REGIME CHANGED banner is prepended on a band flip (tracked via committed state.json).

• Around month-end / quarter-end and big Treasury-issuance / debt-ceiling weeks.

⭐ Canonical data pull: scripts/liquidity_pull.py

Only dependency is requests. No API key. Run it:

PY=/tmp/.insider_venv/bin/python; [ -x "$PY" ] || PY=python3
$PY ~/.claude/skills/macro-liquidity-monitor/scripts/liquidity_pull.py            # human card
$PY ~/.claude/skills/macro-liquidity-monitor/scripts/liquidity_pull.py --json-only
$PY ~/.claude/skills/macro-liquidity-monitor/scripts/liquidity_pull.py --telegram # send if regime changed
TEST_MODE=1 $PY .../liquidity_pull.py --telegram --force   # print the Telegram msg, don't send

Data sources (all public, no key)

FRED rejects a Chrome User-Agent — the script calls FRED with the default requests UA (browser=False). NY Fed needs the browser UA.

No public feed → manual event flags: SLR / eSLR rule changes, GSE balance shifts, debt-ceiling / issuance calendar. These are the structural drivers; note them in the wiki page investing/wiki/dollar-liquidity-plumbing.md when they happen.

The 0-100 Liquidity Score (higher = more abundant / looser)

Additive from a neutral 50. SOFR−IORB is the spine; the rest nudge it:

Bands → 🟢 ABUNDANT (≥80) · 🟡 AMPLE (60–79) · ⚪ BALANCED (45–59) · 🟠 TIGHTENING (25–44) · 🔴 STRESS (<25). Override: SRF ≥ $10B, or SOFR > IORB+5bp with RRP < $10B, forces 🔴.

⭐ The forward-looking factor. A 2yr backtest (scripts/backtest_liquidity.py) showed every level gauge here is coincident (≈0 correlation to next-week returns), but the 13-week CHANGE in net liquidity (WALCL−TGA−RRP) leads forward-3m returns (corr +0.26 SPX / +0.33 NDX). The net-liquidity level is a spurious −0.71 trap. So this row is the one that tilts the score toward what's coming; the rest describe what is. Regime-bucketed fwd-3m: 🟢/🟡 ≈ +4–6%, 🔴 ≈ NDX −1% (value is downside avoidance).

Why an empty ON RRP does NOT drag the score down: that cash already flowed into the system (loose), so it isn't current tightness. An empty buffer is a forward fragility flag — surfaced as a trigger, not subtracted from the abundance number. The per-layer layers{} block still scores RRP 0–3 as evidence.

How to read it (the mental model)

1. SOFR − IORB is the headline. SOFR is what the market pays to borrow cash

against Treasuries overnight; IORB is what the Fed pays banks. SOFR far below IORB = so much cash banks lend it out below the risk-free admin rate = too loose. SOFR climbing to/above IORB = cash is getting scarce = tightening.

1. ON RRP is the shock absorber. It was $2.5T in 2023; near $0 now. With the

sponge empty, any drain (TGA rebuild, QT) hits reserves directly, so SOFR spikes get sharper and faster. This is the key fragility today.

1. TGA + WALCL set the direction. Treasury issuing debt and rebuilding its TGA

account drains reserves; spending it adds. QT (WALCL falling) drains. Net liquidity ≈ WALCL − TGA − RRP.

1. SRF takeup > a few $B = the alarm. Dealers only tap the Fed's backstop when

private repo is too tight. Quarter-ends produce transient spikes; a persistent one means genuine scarcity.

Output contract

scoring.liquidity_score (0-100), scoring.regime, scoring.score_detail{}, scoring.layers{...}, scoring.triggers[], scoring.headline_metrics{liquidity_score, sofr,iorb,spread_bp,effr,rrp_bn,reserves_t,tga_bn,srf_bn,net_liquidity_bn}, calendar{month_end,days_to_month_end,is_quarter_end_month}.

Hard rules

1. Never claim timing certainty — say "tightening risk rising", not "crunch on X date".
2. Always show SOFR−IORB and the RRP buffer; they're the two that matter most.
3. Tag the data date — the plumbing moves daily and month/quarter-ends distort it.
4. Distinguish a month/quarter-end technical spike from a structural drain.
5. Cite the series — every number above has a verifiable NY Fed / FRED source.

Companion

• investing/wiki/dollar-liquidity-plumbing.md — the methodology + standing read.
• macro-warning — equity-valuation/sentiment regime (different question).
• SLR / deregulation context lives in the wiki page as a manual event log.