Reserves Built $30.5bn on a $24.1bn Cheque. The Residual Flipped to $6.4bn, and a Third of the Two-Week Money Still Never Arrived.
Three balances explain the plumbing. Reserves are the cash banks keep on deposit at the Federal Reserve. The Treasury's general account is the government's chequebook at the same institution: money spent from it arrives in the banking system as reserves, money raised into it is pulled back out. The gap between the two is the residual, and it is the only place where a forecast can be marked wrong. For two weeks it ran one way: $43.1bn of a $79.4bn cheque failed to arrive in the seven days to 4 November, the second consecutive miss in that direction. This week it ran the other way, and the size of the correction is the number that matters. Ahead: the $6.4bn overshoot that flipped the sign, the two-week accounting that is still a third short, why the parking lot cannot be blamed this time, and why a 1.9% fall in the S&P against repaired funding is not a liquidity story - and which variable is actually doing the work.
The build repeated, and this time the arithmetic worked in the other direction
Reserve balances were $2,960.6bn as of 11 Nov, from $2,930.1bn as of 4 Nov. That is $30.5bn on the week, after $36.3bn the week before: the first back-to-back builds since 30 September and 7 October.
The chequebook fell $24.1bn over the same seven days, $936.2bn at the 4 Nov close to $912.1bn at the 10 Nov close - the last print before the reserve date, since Treasury published no balance for 11 November. A one-for-one model puts reserves at $2,954.2bn. The print is $2,960.6bn. The residual is +6.4bn.
The chart starts at 14 October for a reason: before that the residual was not small. The 16 September week printed +170.5bn and the following week -128.0bn, on quarter-end and tax-date flows. What changed is not that the model got worse. It is that the residual went quiet for three consecutive weeks - -7.5bn, -0.2bn, +1.6bn - then missed by $43.1bn and printed +6.4bn.
A model that is exact for three weeks and then wrong by $43.1bn is not a broken model. It is a model with an unobserved term that is usually small. September's two prints put the ordinary size of that term above $100bn in a quarter-end week, so a $6.4bn residual is small by the standard of this year rather than by the standard of the last month.
Two weeks, and a third of the cheque still never arrived
The weekly sign is not the whole accounting. From 28 October to 11 November the chequebook paid out $103.5bn, $1,015.6bn down to $912.1bn, and reserves took $66.8bn, $2,893.8bn up to $2,960.6bn. The one-for-one model again points to $103.5bn of build. The shortfall is $36.7bn, 35.4% of the cheque.
Last week's piece booked $43.1bn missing on the first of those weeks. This week's +6.4bn returns part of it, and the two-week number is now the honest one. The transmission is not one-for-one over a fortnight, even when the weekly direction alternates. The residual is not a leak; it is a lag with a fat tail on the payout side.
It is also worth marking what this week retired. The ceiling this column set on 3 November was $2,953bn. The print of 11 November cleared it by $7.6bn. A ceiling that gets broken in the direction the model did not expect is a ceiling that was never structural, and it should not be defended.
The rebuild is a round trip, not a cushion
$2,960.6bn is $2.0bn below the 21 October level of $2,962.6bn and $69.1bn below the 7 October peak of $3,029.7bn. The rebuild has not recovered the month; it has arrived back at where reserves stood on 21 October, having been $68.8bn lower in the interim.
The longer comparison is less comforting. The 2014 high was $2,809.0bn, set on 30 July of that year. This print is $151.6bn above it, in a much larger banking system. On the level alone, reserves are still being managed near the floor of the post-2008 range rather than in the middle of it.
That is the argument for reading the build as a stabilisation rather than a policy turn. A system this close to its floor produces residuals like the last two weeks as a matter of course, because small balance-sheet decisions have large arithmetic consequences at the margin.
The parking lot is empty, so the residual has nowhere to hide
Reverse repo take-up was $0.182bn on 16 November, $0.325bn on 13 November and $0.250bn on 12 November. Nine of the ten sessions since 2 November printed below $0.5bn. The exception is the $5.087bn month-end print on 2 November. The 2026 low is $0.030bn, set on 20 July.
At this size the facility cannot absorb a $24.1bn cheque. Whatever holds the residual is either a bank balance-sheet decision - deposits moving between reserve accounts and non-bank hands - or a Treasury flow that does not touch the general account. Neither is visible in this snapshot, and that is the honest boundary of the read. Two weeks ago this column read the $5.087bn month-end print as a cushion that could absorb part of the miss. At $0.182bn it cannot absorb anything, and that reading is retired.
An empty facility has a second consequence. There is no buffer left to absorb a funding shock, so the next scarcity episode will show up in the secured rate rather than in a drawdown of parked cash. That is why the overnight fix matters more this month than it did in September.
Secured money is at the administered rate, not above it
SOFR fixed at 3.90% on 12 and 13 November against an administered rate of 3.90%: a spread of zero, twice. The 10 November fix was 3.89%, minus 1 basis point.
Volume was $3,138bn on 12 November and $3,092bn on 13 November, so the zero is a cleared market rather than a thin one. The ten-session premium that ended at minus 1 basis point on 6 November has settled flat instead of turning into a discount. Secured money is clearing exactly where the Fed set it, on volumes above $3tn.
Two prints at precisely zero is the most benign reading available. It says the collateral shortage that bid overnight cash above the administered rate has been met by supply, and that no one is so short of cash as to pay up for it. It does not say the system is flush; the reserve level says the opposite.
High yield retraced 25 basis points off a 2026 wide. Investment grade retraced 5.
High-yield spreads were 3.23% as of 13 November, from 3.42% as of 6 November: 19 basis points tighter on the week and 25 inside the 3.48% print of 5 November, the 2026 wide. The 2026 low is 2.60%, set on 28 August, so 63 basis points of widening remain.
Investment-grade spreads were 0.89% as of 13 November, from 0.93%: 4 basis points tighter, and 5 inside the 0.94% print of 2 November, its own 2026 high. The 2026 low is 0.73%, set on 22 January, so 16 basis points remain there.
Both markets have given back all of November's widening, so the retrace itself is not the story. The distance to the year's tight is. High yield is 63 basis points above its August low and investment grade is 16 above its January low, and the ratio between the two spreads has fallen to 3.63 times from 3.68. The speculative end is carrying four times the distance to normal that the high-grade end is.
The stress gauge rose anyway, because volatility did the work
The Credit & Stress sub-index is 0.623, from 0.352 in the 16 November snapshot. Its four inputs: high yield at a 0.508 z-score, investment grade at 0.171, VIX at 1.253 and MOVE at 0.559. The two spread inputs average 0.34. The two volatility inputs average 0.91. The sub-index is a volatility index with a credit label on it.
The headline composite is 0.461. The regime label is TIGHTENING, where it has sat since 13 November after two NEUTRAL prints on 11 and 12 November. The dominant sub-index is Scarcity at 1.46, which is the real 10yr at 2.92% and bitcoin's thirty-session trend at -21.07% doing the work.
Liquidity is the deepest negative at -0.504, with Growth & Inflation at -0.384 and the secured spread at zero. That is the signature of a level problem rather than a flow problem: reserves near the floor of their range and a facility that is already empty. A level problem does not resolve with a good week of Treasury arithmetic.
Risk sold off into a repaired funding market
The S&P closed at 7,416.59, down 145.08, or 1.92%. The NDX closed at 27,197.15, down 746.59, or 2.67%. VIX 22.91 from 19.30, at the 27.3rd percentile. MOVE 101.28 from 98.79, up 8.51 on the week.
Nothing in the funding data explains that. Reserves built, the chequebook released, secured money is flat at the administered rate and the facility is empty. A 1.9% equity decline against that backdrop is a repricing of something other than liquidity, and this file cannot name it. Guessing would be worse than saying so.
The variable that did move with it is the real yield. The 10yr nominal fell 3.0 basis points to 5.120%, and the breakeven fell 4 to 2.20%. The real 10yr is therefore +0.7 basis points higher at 2.921%. Every basis point of the nominal decline came out of inflation compensation, and then some. That is not a growth scare. It is a market cutting the inflation it is willing to pay for.
Claims turned up from a low base
Initial claims were 200,000 in the week to 7 November, from 191,000 the week before: 9,000 higher. Continuing claims were 1,697,000 as of 31 October, from 1,717,000 as of 24 October, 20,000 lower.
Both are low in absolute terms. The last four initial prints sit between 191,000 and 200,000, which is a range, not a trend. The 2026 high is 230,000, set on 31 January, so a 200,000 print is 30,000 below anything that has this year looked like a turn.
The continuing series falling while the initial series rises is the ordinary noise of two series measured a week apart. It is also the pair most likely to be revised, so it should not carry a thesis on its own.
The coin fell 11.4% and the wrapper fell 28.0%
Bitcoin is $64,490.32 at 07:00 UTC, down $6,609.99 on the week, 11.4%, and 21.07% over thirty sessions. Ether $1,824.23, down 11.2%, at the 10.7th percentile of its range.
MSTR is $103.03, down $22.56 on the week, 28.0%, at the 13.5th percentile. The implied beta to the coin is 2.45.
The preferred stack held. STRC $99.12 from $99.88, at the 97.4th percentile. STRF $99.96 from $100.35. STRK $64.22 from $66.16.
That is the balance-sheet hierarchy doing what it is supposed to do. The instrument closest to a bond keeps its price. The instrument carrying the residual claim takes the whole move and more than the coin's own move. In a week when the funding market was repaired and the equity market was not, the wrapper traded like equity and the preferred traded like credit. Buying the complex for coin exposure bought the coin's drawdown at roughly two and a half times the size.
What would change the read
The 18 November reserve print is the first test. The chequebook has rebuilt $12.5bn since the reserve date, $912.1bn on 10 November to $924.6bn on 13 November, so the next week has no payout behind it. A third consecutive build without a cheque to explain it would break the one-for-one model in the other direction. A flat or lower print would restore it.
The second is the spread. SOFR above the administered rate for five consecutive sessions would say the premium is back as a level rather than an episode. Two prints at exactly zero say the market is clearing where the Fed set it, which is the intended state and not a stress signal.
The third is credit. High yield through 3.48% with investment grade unchanged would make November's widening a preview rather than a spike. Investment grade through 0.94% with high yield unchanged would make it a duration event, and duration events do not stay in the credit market.
The fourth is the residual itself. If 11 November's +6.4bn is the start of a payback, the next two weeks should print positive residuals totalling near $36.7bn. If it is the return to the quiet regime the four weeks to 28 October described, then the two-week shortfall stays where it is and the model remains a bound on the sign of the move rather than an estimate of its size.
The fifth is the equity tape against this plumbing. Two more sessions of equity weakness with reserves building, spreads tight and secured money flat would move the story out of this column's scope. Funding is not driving it, and a piece that insisted otherwise would be inventing a transmission channel to keep its own framework intact.
Notes
1. Reserve balances $2,960.6bn as of 2026-11-11 (WRESBAL 2,960,600m), $2,930.1bn (2026-11-04), $2,893.8bn (2026-10-28), $2,962.6bn (2026-10-21), $3,029.7bn (2026-10-07); 2014 high $2,809.0bn (2014-07-30). Weekly change +30.5. Source: FRED WRESBAL, as of 11 Nov - a six-day lag on the post date.
2. Treasury general account closing balances: $924.6bn (13 Nov), $912.1bn (10 Nov), $922.6bn (9 Nov), $910.3bn (6 Nov), $936.2bn (4 Nov), $1,015.6bn (28 Oct), $945.2bn (21 Oct). No 11 November balance: federal holiday. Change over the week to the reserve date -24.1bn, measured to the 10 Nov close. Source: US Treasury Fiscal Data, Daily Treasury Statement, operating cash balance, latest 13 Nov.
3. Overnight reverse repo take-up: $0.182bn (16 Nov), $0.325bn (13 Nov), $0.250bn (12 Nov), $0.152bn (10 Nov), $0.252bn (9 Nov), $0.150bn (6 Nov), $0.125bn (5 Nov), $0.225bn (4 Nov), $0.427bn (3 Nov), $5.087bn (2 Nov). 2026 low $0.030bn (20 Jul). Source: FRED RRPONTSYD, as of 16 Nov.
4. SOFR 3.90% (13 Nov), 3.90% (12 Nov), 3.89% (10 Nov); volumes $3,092bn, $3,138bn and $3,159bn. IORB 3.90%. Spread 0 basis points on 12 and 13 Nov, minus 1 on 10 Nov. Source: Federal Reserve Bank of New York, as of 13 Nov; IORB from FRED, as of 17 Nov.
5. High-yield OAS 3.23% (13 Nov), 3.42% (6 Nov), 3.48% (5 Nov, 2026 high), 2.60% (2026-08-28, 2026 low). Source: FRED BAMLH0A0HYM2, as of 13 Nov.
6. Investment-grade OAS 0.89% (13 Nov), 0.93% (6 Nov), 0.94% (2 Nov, 2026 high), 0.73% (2026-01-22, 2026 low); 0.94% was also printed on 16 March, so 2 Nov is a co-high. Source: FRED BAMLC0A0CM, as of 13 Nov.
7. Initial claims 200,000 (week to 7 Nov), 191,000 (31 Oct), 195,000 (24 Oct), 195,000 (17 Oct); continuing claims 1,697,000 (31 Oct), 1,717,000 (24 Oct), 1,695,000 (17 Oct). Source: FRED ICSA and CCSA, as of 7 Nov and 31 Oct respectively.
8. Marks are 16 November closes except FX, metals and digital assets, which are 17 November 07:00 UTC: 10yr 5.120%, 2yr 4.83%, 2s10s 29.1bp, breakeven 2.20%, real 10yr 2.921%. Source: latest.json 07:00 UTC 17 Nov; breakeven from FRED T10YIE, as of 16 Nov.
9. Equities, volatility and credit ETFs are 16 November closes: SPX 7,416.59, NDX 27,197.15, VIX 22.91, MOVE 101.28, HYG 76.73, LQD 102.80. Percentiles are against each series' stored history, not a fixed window.
10. MSTR, STRC, STRF and STRK are 16 November closes. BTC and ETH are 17 November 07:00 UTC marks. MSTR beta is the ratio of its weekly percentage change to bitcoin's, -28.0% against -11.4% - a single-week ratio, not a fitted beta.
11. Composite sub-indices and the regime label are from latest.json at 07:00 UTC 17 Nov: headline 0.461, TIGHTENING, dominant Scarcity 1.46. Each sub-index is the average of its non-null inputs: Credit & Stress (0.508, 0.171, 1.253, 0.559) to 0.623; Liquidity (-0.595, -0.918, 0.000) to -0.504; Growth & Inflation over four of five, core CPI being null, to -0.384; Scarcity over two of three, the gold trend being null, to 1.46. The 16 November comparison for Credit & Stress is that day's snapshot, whose Liquidity and Growth inputs differ from today's; only the credit sub-index is compared across the two.
12. Core CPI is null in this snapshot, so no inflation rate is quoted. The breakeven and real yield carry the inflation read instead, and both are dated above.
| TENOR | LEVEL | 1D Δ | Z (3Y) | RANGE %ILE |
|---|---|---|---|---|
| 2Y | 4.800 | -0.030 | +1.73 | 87 |
| 5Y | 4.900 | -0.030 | +2.38 | 90 |
| 10Y | 5.100 | -0.020 | +2.63 | 89 |
| 30Y | 5.460 | 0.000 | +2.19 | 88 |