The ampleness check

1. Reserves: the level and the share of GDPWATCH

History percentile: reserves/GDP sits at percentile 26 of its own history (expanding percentile, from the board's turn engine).

Thresholddesk editorial
AMPLE at or above the board's fitted kink ratio (11.40% of GDP today), SCARCE below 8% of GDP, WATCH in between. The kink ratio is fitted, not chosen. The 8% floor is the desk's judgement: reserves broke at roughly 6 to 7 percent of GDP in 2019 and the demand curve has shifted up since, so the floor is set above the 2019 mark rather than at it.

What 2018-19 says this level meant
Reserves were about $1.4 trillion when overnight repo broke in September 2019, which was roughly 6 to 7 percent of GDP at the time. The dollar figure does not carry across seven years of growth in the payments system, the balance sheet and bank size, which is why the share is the comparable measure and the fitted kink is the live one. Reserve demand has plainly shifted up since 2019: the board's own fit puts the bend well above where 2019 bit, so treat the 2019 ratio as a floor marker in the record, not as today's threshold.

2. Reserves as a share of bank assetsNOT AVAILABLE TODAY

What 2018-19 says this level meant
The Fed's ample-reserves discussion uses reserves against bank assets alongside reserves against GDP, because the demand for reserves is a demand by banks and it scales with their balance sheets rather than with the economy. It is the measure most likely to explain why the 2019 ratio does not transfer: banks are much larger now than they were then.

3. Distance to the fitted reserve demand kinkSCARCE

History percentile: the payload carries no history percentile for the distance itself; the fit quality and the model against market check above are the gates that stand in for one.

Thresholddesk editorial
AMPLE more than $200B above the fitted kink, WATCH inside that cushion, SCARCE at or below the kink. The kink location is fitted from the data. The $200B cushion is the desk's judgement, about 6 percent of the current reserve balance: the fitted breakpoint is a point estimate on a hinge regression, and a band around it is more honest than treating it as a wire.

What 2018-19 says this level meant
The reserve demand curve is flat while reserves are abundant and bends steeply near scarcity, which is why the crossing in 2018-19 was so quiet. Nothing looked wrong on the price side for months, and then one morning in September 2019 the spread went from single basis points to hundreds. Distance to the bend is the early part of that story: it says how much of a shock the system can absorb before the price starts responding to the quantity at all.

4. The NY Fed's reserve demand elasticity printWATCH

History percentile: the official series publishes a median and two bands rather than a history percentile, and the bands are the better test.

Threshold
AMPLE when zero sits inside their 68% band (the curve is flat, which is their own definition of ample), WATCH when zero is outside the 68% band but inside the 95%, SCARCE when zero is outside both. The cut points are the bands the New York Fed publishes with the measure, not numbers chosen here.

What 2018-19 says this level meant
The New York Fed built this measure so the ampleness argument would not have to be settled by anecdote, and its own reading of the 2018-19 period is that the demand curve had already left the flat region before the September 2019 break. That is the useful property: the elasticity moved while the price still looked calm. Their definition of ample is a flat curve, so zero inside the band is the ample reading and a band that excludes zero is the curve saying reserve changes now move rates.

5. SOFR minus IORBAMPLE

History percentile: percentile 76 of its own expanding history, from the breakwater engine's replay of the same spread.

Threshold
AMPLE below zero (SOFR under IORB), WATCH from zero to 10 bp above it, SCARCE above that. The scarce cut is the SRF offering rate itself, currently IORB plus 10 bp, so it is the Fed's administered ceiling rather than a number chosen here.

What 2018-19 says this level meant
Through 2018 and into 2019 SOFR moved from printing below IORB to printing above it, first only on settlement dates, then routinely, and finally by hundreds of basis points on 17 September 2019. A negative print is the ample-regime signature: cash lenders are content to leave money at the Fed rather than pay up in the market, which is only true when reserves are plentiful enough that nobody is bidding for them.

6. The share of repo volume printing above the ceilingAMPLE

History percentile: this line IS the percentile frame: the readings above are the 75th and 99th percentiles of the repo distribution against the administered ceiling.

Thresholddesk editorial
SCARCE on any 75th percentile breach at all, because that print proves at least a quarter of repo volume paid above the Fed's own ceiling. WATCH at 3 or more sessions in the last 20 with a 99th percentile breach. AMPLE below that. The bound logic is the feed's, not the desk's: only the 1st, 25th, 75th and 99th percentiles are published, so the share above the ceiling is bounded rather than measured, and this line says bounded. The 3 session cut is the desk's judgement about when isolated prints stop being isolated.

The stigma engine's own caveats, verbatim: only P1/P25/P75/P99 are published, so mass above the ceiling is bounded, not measured: a P99 breach proves at least 1 percent of volume paid up, a P75 breach proves at least 25 percent; the true share in between is unobservable from this feed; percentiles are volume weighted rate levels: bp above the ceiling is rate distance, not dollar volume.

What 2018-19 says this level meant
In the run-up to September 2019 the upper percentiles went first. The 99th percentile of repo cleared the administered rate for months while the median still looked perfectly calm, and only then did the whole distribution follow. Anyone watching the average alone would have seen nothing until the morning it was too late, which is the case for reading the distribution rather than the print.

7. EFFR minus IORBAMPLE

History percentile: the payload carries no history percentile for this spread; the sign against IORB is the test that mattered in 2018-19 and it needs no percentile.

Thresholddesk editorial
AMPLE below zero (EFFR under IORB), WATCH from zero to 5 bp above it, SCARCE above that. Zero is not the desk's line: the Fed's own 2018-19 practice was to make a technical adjustment when the effective rate pressed the administered rate. The 5 bp width is the desk's judgement, taken from the size of those adjustments, which were 5 bp each.

What 2018-19 says this level meant
EFFR climbed toward and then through IOER during the 2018-19 runoff, and the Fed answered with a sequence of 5 bp technical adjustments to the administered rate rather than a policy move. That sequence is the cleanest signal the last cycle gave: when the effective rate presses the rate the Fed pays on reserves, the plumbing is saying reserves are getting tight, and it says it in the one market the Fed targets.

8. SRF and discount window take-upAMPLE

History percentile: the payload carries no history percentile for take-up; the engine's de minimis, notable and material bands are the frame it publishes instead.

Thresholddesk editorial
One token for the pair, and it is the worse of the two. SRF take-up is graded on the stigma engine's own published scale: AMPLE below $1B (de minimis), WATCH to $25B, SCARCE at or above. The discount window is graded AMPLE below $10B, WATCH to $25B, SCARCE above, and those cuts are the desk's judgement: they are set wide deliberately because the post-2023 push to pre-position and test the window puts a few billion of routine borrowing on the tape that has nothing to do with stress.

The pair grades as: SRF AMPLE, the discount window AMPLE.

What 2018-19 says this level meant
There was no standing repo facility in 2019: it was created in July 2021 precisely because of what happened. The discount window did exist, and it was barely touched through the squeeze, which is the stigma problem in a single fact. Take-up is therefore a confession indicator and the size matters less than the willingness to be seen. One thing has changed since: from 2023 the Fed has pushed banks to pre-position collateral and to test the window, so small steady use now reads as hygiene rather than distress, and this line is graded with that in mind.

9. The ON RRP buffer left to drainSCARCE

History percentile: the payload carries no history percentile for the RRP balance; near the floor a percentile would be the least informative way to say it.

Thresholddesk editorial
AMPLE above $100B, WATCH from $25B to there, SCARCE below. Both cuts are the desk's judgement. The reasoning is stated so it can be argued with: the facility ran above $2 trillion through 2022 and 2023, and a balance under $25B cannot absorb one heavy settlement week, which makes the next drain a reserve drain by arithmetic rather than by judgement.

What 2018-19 says this level meant
There was no meaningful ON RRP balance in 2018-19, so every dollar drained from the Fed's liabilities came straight out of reserves, and that is the configuration that produced September 2019. Between 2022 and 2024 the facility was a shock absorber measured in trillions, and balance sheet runoff could proceed for two years without touching reserves at all. When that absorber is empty the runoff arithmetic goes back to what it was in 2019.

10. The runway's kink crossing datesSCARCE

History percentile: a projected date has no history percentile; the three scenarios are the uncertainty statement.

Thresholddesk editorial
SCARCE when the base case crosses the fitted kink inside the horizon (or starts below it), WATCH when only a stressed scenario crosses, AMPLE when none does. That ladder is the desk's judgement about how to read a scenario set, and the underlying projection is arithmetic on the engine's stated assumptions rather than a forecast of policy: the Fed can change the drift on any Wednesday and this line would not know until it did.

The runway engine's own caveats, verbatim: arithmetic on stated assumptions, not a forecast of policy; the trailing drift already embeds recent QT, settlement and fiscal flows, so the settlement term enters as a deviation from the calendar's own weekly mean (shape, not level) and only the explicit QT pace can still double count against the drift.

What 2018-19 says this level meant
The timing lesson from 2018-19 is that the crossing is not an event anyone feels. Reserves fell through the level that mattered months before the September break, and the calendar supplied the trigger when a tax date and a settlement date landed together. A projected crossing date is not a forecast of stress. It is the date after which the calendar starts to matter.