# Quantitative Tightening **Quantitative tightening** is the reduction of the central bank's securities holdings, extinguishing reserves and reversing the portfolio effects of [[Quantitative Easing]]. It is not symmetric with QE, and the asymmetry is the whole subject: easing can be done in whatever size seems useful, while tightening walks the banking system down a demand curve toward a cliff whose location cannot be observed in advance. ## Mechanism Two methods exist in principle — passive **runoff**, in which the Fed declines to reinvest principal from maturing Treasuries and MBS prepayments, and **outright sales** into the secondary market. > [!important] The Fed has never conducted outright sales > Both QT episodes — 2017–19 and 2022–25 — were pure runoff. The 2022 Plan describes only adjustments to reinvestment and contains no provision for sales. This is the single most frequently misstated operational fact about QT. Both episodes used **monthly redemption caps**: principal up to the cap runs off, principal above it is reinvested. Caps make the pace predictable and insulate it from a lumpy, rate-sensitive maturity schedule — which matters because MBS prepayments accelerate when rates *fall*, so uncapped runoff would be fastest exactly when the Fed least wants tightening. In practice the opposite bound bit: once rates rose in 2022, prepayments collapsed and MBS runoff persistently ran *below* its $35 billion cap, which is why MBS holdings stayed large (roughly $2.1 trillion as of October 2025) long after Treasury holdings had fallen sharply. ## The first episode, 2017–2019 — and the accident Designed in the **Addendum to the Policy Normalization Principles and Plans, 14 June 2017**: Treasury caps starting at $6bn/month and rising in $6bn steps to $30bn; agency debt and MBS starting at $4bn and rising to $20bn. Runoff began **October 2017**. In March 2019 the Committee announced it would halve the Treasury cap in May and conclude in September; at the July meeting it ended runoff two months early, effective **1 August 2019**. ### September 2019 On **16–17 September 2019** overnight funding rates spiked violently. SOFR rose above 5%. Critically, the pressure escaped the repo market and hit the Fed's own policy rate: the **effective federal funds rate reached 2.3%, above the top of the target range.** A central bank losing control of its policy rate is an institutional failure, and that is what made this more than a plumbing story. The proximate causes were two large, entirely predictable, simultaneous cash drains on 16 September — roughly $120 billion in corporate quarterly tax payments, and settlement of $54 billion in long-term Treasury debt that dealers had to finance. Together they cut reserves by about $120 billion over two business days. The underlying condition was **reserve scarcity**. Reserves had fallen below $1.4 trillion, down from roughly $2.8 trillion in October 2014. The Fed responded with overnight repo operations of up to $75 billion a day from 17 September — the first since the financial crisis — a technical cut to the interest rate on excess reserves on 19 September, and from 11 October, Treasury bill purchases of about $60 billion a month. ⭐ **What it revealed, and why every later decision follows from it:** 1. **The reserve demand curve is non-linear.** Money-market rates are insensitive to reserve levels across a wide range, then become violently sensitive past a threshold. There is no advance warning. 2. **The Fed cannot observe the threshold.** The "lowest comfortable level of reserves" is not calculable; it must be inferred from market signals, and by the time the signal is unambiguous, control is already lost. 3. **Aggregates hide distribution.** Reserves were concentrated in a few large banks whose intraday liquidity and balance-sheet constraints made them unwilling to lend into repo at nearly any rate. 4. **A backstop was missing.** The absence of a standing facility forced ad hoc improvisation — which produced the [[Standing Repo Facility]] in July 2021. The October 2019 bill purchases also set a definitional precedent the Fed still relies on: it insisted these were **reserve management, not QE**, because buying short-dated bills extracts no duration and compresses no [[Term Premium]]. Critics called it "QE4." The distinction is analytically defensible, and it governs the 2026 operations below. ## The second episode, 2022–2025 The **Plan of 4 May 2022** started runoff on **1 June 2022** at $30bn/month Treasuries and $17.5bn agency for three months, stepping up to **$60bn and $35bn** — roughly double the maximum pace of 2017–19, from a far larger balance sheet. Caps were cut twice: Treasuries from $60bn to $25bn effective **1 June 2024** (with agency principal above its cap now reinvested into Treasuries, a deliberate move toward an all-Treasury portfolio), and from $25bn to $5bn effective **1 April 2025**. Governor Waller dissented from the 2025 slowdown, preferring to maintain the pace — his view being that debt-ceiling dynamics were inflating reserves artificially and the Fed was reacting to a measurement artifact rather than a real constraint. **The [[Overnight Reverse Repo Facility]] absorbed the first eighteen months.** Money funds pulled cash out of the facility to buy the flood of Treasury bills issued after the 2023 debt-ceiling resolution, so bank reserves stayed roughly flat even as the balance sheet shrank. QT only began to bite on reserves once the ON RRP approached empty — which is exactly why the Fed's stopping decision came when it did. Powell's figures of 14 October 2025: the balance sheet had fallen **$2.2 trillion since June 2022, from 35% to just under 22% of GDP**. He also offered a retrospective concession worth recording — that with hindsight the Fed "could have — and perhaps should have — stopped asset purchases sooner." **QT ended 1 December 2025**, announced at the 29 October 2025 meeting, on the stated ground that reserves had moved sufficiently close to ample. Duration: about three and a half years and roughly $2.2 trillion. ⭐ **The contrast with 2019 is the point.** This time the Fed stopped *before* a money-market accident, acting on early signs — firming repo rates, temporary pressure on selected dates — and deliberately targeting reserves "somewhat above the level we judge consistent with ample reserve conditions." Stopping before the kink rather than at it. On the evidence so far, the 2019 lesson was learned. ## Where things stand, September 2026 QT is over and the balance sheet is growing again — and the Fed does not call it QE. - **Reserve management purchases** began January 2026. Per the July 2026 Monetary Policy Report, the Fed bought nearly $250 billion in Treasury bills from early January, raising total assets by about $150 billion. - **Total assets** were $6.725 trillion on 1 July 2026; **reserves** roughly $3.1 trillion, judged by the FOMC to be within the ample range. - **ON RRP usage is near zero on most days.** The buffer is gone, so reserves now respond one-for-one to balance-sheet changes. - **RMPs are currently paused** — a second consecutive monthly pause, with the Desk intending no reserve management purchases over the mid-September to mid-October window, citing ample reserves. The Fed is in a genuinely new regime: the balance sheet grows passively with currency demand and trend liability growth, calibrated month to month, with pauses when reserves run ahead of need. This is neither easing nor tightening but **maintenance** — and the Fed's insistence on buying only bills and sub-three-year paper, so as not to extract duration, is the same line it drew in October 2019. ## How the Fed decides when to stop The lowest comfortable level of reserves depends on banks' internal liquidity requirements, regulatory treatment, the distribution of reserves across institutions and payment-system frictions, all of which drift. So the Fed monitors signals rather than computing a number: repo rates relative to the interest rate on reserve balances and the frequency of spikes on month-end, quarter-end and tax dates; the effective funds rate's drift within the target range; ON RRP take-up; [[Standing Repo Facility]] usage; and the dispersion of reserves across banks rather than the aggregate. Published quantitative benchmarks exist but are individual estimates, not Committee policy — Governor Waller's July 2025 speech noted reserves fell below 7% of GDP when 2019 stress emerged and put a comfort threshold nearer 9%, implying roughly $2.7 trillion. Attribute such figures to their authors, never to "the Fed." The fundamental asymmetry governs everything: **the costs of stopping too early are small and diffuse; the costs of stopping too late are sudden and public.** ## See Also - [[Quantitative Easing]] — what this reverses, and the symmetry puzzle - [[Ample Reserves]] — the regime, and the demand curve with the kink in it - [[Standing Repo Facility]] — September 2019's direct institutional legacy - [[Overnight Reverse Repo Facility]] — the buffer that absorbed QT's first half - [[Open Market Operations]] · [[Term Premium]] - [[Jerome Powell]] · [[Lorie Logan]]