A measured reserve decline and a dated settlement have put Seiche’s funding board in STRAIN. But overnight cash still signals abundance, making the settlement a test of whether calendar pressure can become a funding constraint.
The new change in the dollar-funding picture is a reserve drain meeting a dated settlement—not a spike in the price of overnight cash.
Reserves fell $49.3B in the week through 2026-08-12 to $2,944B. The Treasury General Account rebuilt $56.6B and foreign RRP parking rose $39.7B. Seiche’s board reads 45.2, STRAIN, led by the dated reserve path, which contributes 11.0 points and is pinned near its ceiling. The immediate observable test is the $266B auction settlement on 2026-08-20.
The thesis is narrow: dollar-funding capacity may be tightening through the timing and location of balances before an aggregate overnight-cash shortage appears in market prices. The evidence does not establish a funding event. Indeed, the current price of overnight cash argues against one.
The weekly reserve ledger shows the immediate path of the drain. The TGA rebuild and increase in foreign RRP parking were partly offset by a $36.5B release from the residual, producing the $49.3B decline in reserves. Six legs sum to the reserve change to the dollar, although $18.9B of the balance sheet, or 0.28% of assets, remains in an unnamed residual. The data overview therefore identifies the accounting movements without fully resolving every balance-sheet shift.
A separate official-sector measure points in the same direction. Foreign officials cut Fed custody Treasuries by $90.2B over 13 weeks while foreign RRP parking rose $34.8B. The combined official footprint nevertheless fell $55.5B to $2,954B. Seiche calculates that 62% of the custody decline left the official sector rather than moving within it. That is evidence of a smaller official footprint, not proof of a funding shortage.
The mechanism becomes relevant at settlement. Seiche’s reserve-path model places reserves below its estimated demand-curve kink as the $266B settlement arrives. Its central path for 2026-08-20 is $2,935.2B, while its lower error-band outcome is $2,882.1B. A settlement must be financed through dealer and bank balance sheets; the question is whether deploying those balance sheets changes the price or availability of secured funding.
The STRAIN label is a Seiche derivation, not an observed market price. The board has full input coverage, but its largest contribution comes from the dated reserve path rather than a live funding-rate dislocation.
The reserve-demand model estimates a kink at $3,703.1B as of 2026-08-12, putting reserves $759B below that estimate. Its fit has an R-squared of 0.62. Yet the contemporaneous secured-rate evidence diverges: SOFR was 1.7bp below IORB as of 2026-08-12, compared with a model-predicted spread of negative 0.4bp. A negative SOFR-IORB spread is an abundance signal, not a scarcity print.
The same split appears in Seiche’s plumbing-versus-market comparison. Plumbing was at the 64th percentile and market stress at the 27th percentile as of 2026-08-13, a gap of 37 percentile points. The market inputs included VIX at 14.63, high-yield OAS at 2.71%, investment-grade OAS at 0.79%, and 10-year realized volatility at 5.32bp per day. The percentile gap is a Seiche calculation: it says plumbing leads price; it does not say market stress has begun.
The calendar lens is similarly bounded. Its read for a SOFR-IORB pop of at least 10bp within five business days was 9.5% as of 2026-08-13. The model reports that its curve beats climatology out of sample, but also cautions that multi-week cumulative figures are upper bounds because funding pops cluster. The five-business-day reading is a monitored calendar risk, not literal odds.
The strongest counter-case is that funding capacity is not currently tight where a binding constraint should first appear: in the price of overnight cash and use of the official backstop.
SOFR was 3.62% and EFFR was 3.63% as of 2026-08-13, while IORB was 3.65% as of 2026-08-17. Secured overnight cash was trading below the administered rate. Standing Repo Facility accepted take-up was $0.00B as of 2026-08-14. Market stress was at only the 27th percentile of its own history.
This case can defeat the thesis. Reserves can fall, their distribution can shift, and large settlements can clear without a meaningful rise in funding costs if banks and dealers retain enough balance-sheet elasticity. The strain case rests primarily on a modelled reserve path and slow-moving ledger structure; the counter-case rests on current rates and facility use. Seiche’s confidence is therefore guarded.
The test is the settlement window around 2026-08-20. The tightening thesis gains support only if the $266B settlement is followed within five business days by a SOFR-IORB pop of at least 10bp relative to its trailing five-business-day median—the dossier’s defined funding-event statistic—or by Standing Repo Facility accepted take-up above $0.00B.
If neither condition occurs, the claim that this dated reserve drain is currently constraining system dollar-funding capacity fails this test. That outcome would not erase the reserve decline, official-sector changes, or calendar pressure. It would show that private balance sheets absorbed them without observable funding stress.
Later watch dates remain 2026-08-31, with a $24B settlement and a lower-band reserve estimate of $2,849.6B, and 2026-09-15, with a lower-band estimate of $2,831.1B. These are reserve-path scenarios, not forecasts of a dislocation.
Follow the sequence rather than a single print: TGA rebuilding and foreign RRP parking remove or redirect balances; lower reserves meet settlement needs; banks and dealers decide whether to deploy balance sheets; secured rates and official-facility use reveal whether that deployment has become costly.
For now, the chain breaks at the final link. The ledger and calendar describe pressure, while SOFR and SRF usage describe accommodation. That permits a STRAIN reading without a claim that a funding event is under way.
Observed inputs and series are available in Seiche’s data overview and series index. The current board is published in the daily dispatch, with prior publications in the dispatch archive. Definitions for the composite, reserve path, kink estimate, and calendar tools are in the methodology. Desk updates are available through the Liquidity Lab Desk channel.
Research boundary: Seiche’s composite is a deterministic synthesis, not a market price, and its confidence measure reflects agreement among independent signals rather than the probability of a market outcome. The reserve path uses trailing balance-sheet drift and seasonal Treasury-account assumptions with error bands. Historical transforms used in related tools rely on final or current-vintage history, do not reconstruct what was publicly knowable on each historical date, and are not eligible as publication-vintage backtests or real-money evidence.
Seiche covers system dollar-funding capacity. For institution and lender balance-sheet risk, see LiquiLens. For market liquidity, crowding, and exit capacity, see Undertow.
Research and market data, not investment advice.