The Signal That Ended Quantitative Tightening Came From Repo Markets, Not the Portfolio

Balance sheet reduction is usually explained with a mirror. Easing added securities and created reserves; tightening removes securities and extinguishes reserves; the second is the first run backwards. As an accounting identity the framing holds: assets fall, liabilities fall, and duration absorbed by the central bank returns to private portfolios.

What the mirror does not survive is contact with how the most recent programme ended. The Federal Open Market Committee announced on 29 October 2025 that it would conclude the reduction of its aggregate securities holdings, effective 1 December 2025. Since June 2022 the portfolio had shrunk by more than $2.2 trillion, roughly $1.6 trillion of Treasuries and about $600 billion of agency mortgage-backed securities. Reserve balances stood at $2.85 trillion on 31 December 2025, per the New York Fed's report on open market operations during 2025.

That last figure is where the mirror breaks. The previous runoff stopped in September 2019 with reserves below $1.4 trillion; staff put the mid-September trough at $1.34 trillion, the lowest since 2012. The 2025 programme halted with more than twice that. Either the floor moved by more than a trillion dollars in six years, or the floor was never the variable being measured.

The Surface Reading: A Programme That Stopped Far Above the Old Floor

Each step of the visible sequence is documented. Redemption caps started on 1 June 2022 at $30 billion a month for Treasury securities and $17.5 billion for agency debt and mortgage-backed securities, stepping up after three months to $60 billion and $35 billion. In June 2024 the Treasury cap was cut to $25 billion while the agency cap stayed at $35 billion. In April 2025 the Treasury cap fell again, to $5 billion. In December 2025 the programme stopped entirely, replaced by rolling over all maturing Treasury holdings at auction and reinvesting all agency principal into Treasury bills.

Three tapering steps before a full stop is not a programme running to a known destination. It is a programme feeling for an edge it cannot see.

The liability side gives the reason. Overnight reverse repurchase agreement balances, which had averaged $171.4 billion in December 2024, fell to near zero by September 2025 as money market funds moved into Treasury bills. Once that cushion was gone, every further dollar of runoff came out of reserve balances rather than out of a facility built to absorb surplus cash. Reserves fell $322.3 billion over 2025 to $2.85 trillion.

By 1 July 2026 the picture had inverted. The Federal Reserve's balance sheet stood at about $6.7 trillion, up roughly $150 billion since early January, with $4.492 trillion of Treasury securities and $1.951 trillion of agency debt and mortgage-backed securities. Reserve balances had risen $54 billion to $3.1 trillion. A programme framed as the reverse of easing had, within seven months of concluding, become one of outright purchases.

Where the Asymmetry Actually Lives

Easing pushes; tightening pulls against an unknown

Quantitative easing has no binding lower constraint on the quantity it can add; reserves can be created in whatever volume the central bank chooses. Tightening faces the opposite structure. Reserves are the settlement asset for the payment system, and below some level the system stops functioning smoothly. That level is not published, not stable, and not directly observable — it depends on regulatory liquidity requirements, intraday payment timing, the distribution of reserves across institutions, and dealer willingness to intermediate between holders of cash and those who need it.

This is the asymmetry that matters, and it is not about duration or term premium. It is that easing is bounded by choice and tightening is bounded by a threshold nobody can locate in advance.

The composition of the shrink was never fully controlled

A second asymmetry sits in the portfolio. On the way up the central bank chose what to buy; on the way down it could only choose a cap, and whether a cap binds depends on borrower behaviour. Agency mortgage-backed principal payments averaged $16.2 billion a month during 2025 against a $35 billion cap. The cap was never reached, so it never governed anything, and no reinvestment operations were required through November 2025.

Composition drifted rather than being steered: mortgage holdings ran off at whatever pace prepayment behaviour allowed, which at elevated mortgage rates was slow.

The stopping rule is a spread, not a stock

Because the quantity floor is unobservable, the framework substitutes a price test. The effective federal funds rate trading a basis point or two below the interest rate on reserve balances reads as ample; compression toward zero, with secured rates printing above the administered rate, reads as supply no longer comfortably exceeding demand.

Over 2025 the effective federal funds rate moved from about 7 basis points below the interest rate on reserve balances to about 1 basis point below by year-end, and secured spreads widened over the second half around period-end and Treasury settlement dates. That drift, not any reserve total, is what the programme was reading.

What the 2025 Signal Actually Looked Like

The sequence of operational changes is more informative than any single rate print, because each was a response to observed pressure rather than a scheduled step. A morning standing repo operation was added on 26 June 2025, giving counterparties a second daily window instead of one. On 10 December 2025 the aggregate limit on standing overnight repo operations was removed, effective 11 December, moving the facility to full allotment with one proposition per security type per operation, up to $40 billion each. An aggregate cap is removed when it is plausibly binding.

Usage confirms it. Standing repo operations reached about $75 billion on 31 December 2025, the highest since the facility was introduced in 2021, with take-up clustering on reporting dates and when repo traded well above the facility rate.

Federal Reserve staff research published in 2026 makes the diagnostic point explicit. In the 2017–2019 episode the classic scarcity indicators — reserve demand elasticity, late payment shares, daylight overdrafts, domestic institutions borrowing federal funds — deteriorated together before the September 2019 disruption. In 2022–2025 the binding signal was different: the share of repo volume transacting at or above the interest rate on reserve balances became the primary indicator of tightness. That points at dealer intermediation capacity, not bank reserve demand.

The distinction is not academic. If the constraint is how much balance sheet dealers can devote to matched-book repo, reserve totals can look comfortable while the market distributing those reserves is congested. A mirror-image framing, which treats the reserve stock as the state variable, has nowhere to put that observation.

The Part Most Commentary Underweights: The Return Trip Is Also Asymmetric

Attention concentrates on the risk that runoff goes one step too far. Less goes to what happens immediately afterwards. Reserve management purchases began in mid-December 2025 at roughly $40 billion a month, a pace the New York Fed described as temporary and tied to the seasonal Treasury General Account build through the April tax date. All were conducted in Treasury bills, with authority to buy other Treasury securities maturing in three years or less if bill market conditions required. By the July 2026 Monetary Policy Report, cumulative purchases since January totalled about $160 billion.

The rate response then overshot the other way. FOMC minutes for the 16–17 June 2026 meeting record repurchase agreement rates falling to 15 basis points below the interest rate on reserve balances in mid-May 2026, attributed to seasonal Treasury General Account movements, reserve management purchases reducing net bill supply, greater bank intermediation capacity, softer levered repo demand, and higher cash investment by government-sponsored enterprises.

Within roughly six months, secured rates moved from printing above the administered rate to printing well below it. That is not a system easing gradually off a hard boundary. It is a system whose apparent distance from the boundary is highly sensitive to fiscal flows and dealer capacity, both of which move faster than balance sheet policy can.

Staff estimates show how wide the uncertainty band is. Reserve demand estimates cited in 2026 Federal Reserve staff work put the level consistent with a 2 basis point spread between the effective federal funds rate and the interest rate on reserve balances at about $2.65 trillion, roughly $350 billion below prevailing levels, with a confidence band of something like $150 billion to $550 billion. A range that wide is why no quantity target was ever announced.

The same work argues a materially smaller footprint need not require scarce reserves at all. It catalogues fifteen policy pathways — liquidity regulation changes, discount window capacity recognition, supervisory treatment of bills versus reserves, payment system upgrades, sterilising Treasury General Account swings with bill issuance — with simulated aggregate reduction potential of $1.2 trillion to $2.1 trillion. On that reading the binding constraint is regulatory and operational design, not the pace of redemptions.

Three Central Banks, Three Different Unwinds

Cross-country comparison is the cleanest test of whether tightening is one mechanical reversal. It is not.

Institution	Method	Documented parameter
Federal Reserve	Passive redemption only; no outright sales	Caps reduced three times, then concluded 1 December 2025
Bank of England	Maturities plus active gilt sales	Stock target reduced by £70 billion over October 2025 to September 2026, to £488 billion
European Central Bank	Passive redemption; framework redesigned around demand-driven supply	Main refinancing spread over the deposit facility rate narrowed from 50 to 15 basis points, effective 18 September 2024

The Bank of England's schedule for July to September 2026 shows how granular active sales become: three short-maturity auctions of £725 million each in the 3 to 7 year sector, two medium-maturity auctions of £600 million each in the 7 to 20 year sector, and no long-maturity sales. Suspending the long end while continuing elsewhere is a duration decision with no counterpart in a passive programme. The European framework went a third way, narrowing the corridor so balance sheet size becomes an outcome of bank demand rather than an instrument.

Conditions That Reverse the Signal

• If the binding constraint is dealer balance sheet rather than reserve quantity, repo spreads carry information about intermediation capacity and little about how much further redemptions could have run. The programme may then have stopped early relative to any true reserve floor.
• If fiscal flows dominate, the symptoms mimic scarcity without scarcity being present. September 2019 combined a corporate tax date on 16 September with $54 billion of long-term Treasury settlement the same day, draining more than $100 billion of reserves in two days. A comparable Treasury General Account swing produces the same rate print regardless of the reserve level.
• If a full-allotment standing facility works as designed, secured rate spikes are capped by construction and stop conveying scarcity information. The indicator that ended the programme is then muted by the tool built to manage it, and monitoring has to shift to facility usage volumes rather than rate levels.
• If regulatory or operational reform reduces precautionary reserve demand, a materially smaller balance sheet becomes compatible with stable overnight rates, and the 2025 stopping point looks like a function of the rulebook rather than a structural limit.
• The symmetric view retains force at the long end. On duration and term premium, adding and removing holdings plausibly work in roughly opposite directions, even if magnitudes differ. The asymmetry set out here concerns the money market and the reserve floor, not the whole transmission channel; treating it as a general claim about yields would overreach.

Concrete Framework — Reading It in Sequence

1. Start with the spread, not the stock. Track the effective federal funds rate against the interest on reserve balances rate. The 2025 drift from roughly 7 basis points below to 1 basis point below is the move that preceded the policy change; compression toward zero or a positive print is the alert condition.
2. Check secured rates separately. Repo persistently at or above the administered rate carries different information than a reserve level. The indicator is two-sided: mid-May 2026 saw repo 15 basis points below the interest on reserve balances rate.
3. Read facility usage as an intensity gauge. Take-up on quarter-end and reporting dates is routine; the same on ordinary mid-month days is not. The 31 December 2025 peak near $75 billion is the reference point for a stressed reporting date.
4. Separate the fiscal driver from the policy driver. Before attributing a rate move to balance sheet policy, check the Treasury General Account path and the settlement calendar. Both moved rates in September 2019 and both were cited again in the June 2026 minutes.
5. Treat the reserve level as a wide band, not a number. Published confidence bands run to several hundred billion dollars. Any commentary naming a precise floor asserts more precision than the research supports.
6. Check whether purchases are operational or expansionary. Bill purchases that offset growth in non-reserve liabilities are a rate-control operation; purchases of longer duration intended to compress term premium are a policy stance. Conflating them produces the wrong conclusion about the direction of policy.

What to Watch Next Week

• The weekly H.4.1 release. Published Thursdays, it gives reserve balances and the Treasury General Account. The item of interest is whether reserves hold near the $3.1 trillion reported for 1 July 2026 or resume drifting.
• Daily standing repo operation results. Negligible take-up outside reporting dates is consistent with the abundance signal seen in mid-2026; recurring mid-month take-up would be the first sign the balance moved back.
• The monthly reserve management purchase schedule. Amounts are announced on or around the ninth business day of each month, with purchase periods running mid-month to mid-month so agency principal payment data can be incorporated. A pace well above or below the roughly $40 billion monthly level of early 2026 signals a changed read on liability growth.
• Overnight reverse repo take-up. After sitting near zero from September 2025, renewed usage indicates the floor is being tested from below rather than the ceiling from above.
• Administered rate settings. The 29 July 2026 implementation note set the target range at 3.50 to 3.75 percent, interest on reserve balances at 3.65 percent, standing repo operations at 3.75 percent, and reverse repo at 3.50 percent with a $160 billion per-counterparty limit. A change to any of these independent of the target range would signal a technical adjustment rather than a policy shift.

One test would settle the question and is unlikely to be run. If reserves fell well below the 2025 stopping level without secured rates moving above the administered rate, the claim that the programme met a genuine constraint would weaken sharply. Absent that test, the reserve floor stays an estimate rather than a measurement.

The honest summary is narrow. Easing and tightening differ in what limits them, how quickly the limit announces itself, and who controls composition. On duration the mirror may be serviceable. On the money market it is not, and the record between October 2025 and July 2026 is the clearest available demonstration.

This article is analysis of monetary policy mechanics and is not investment or financial advice.