Quantitative Easing and Tightening: The Fed's Balance Sheet

The Bank of Japan named the policy. Having cut its policy rate effectively to zero and found the economy still deflating, it announced in March 2001 that it would shift the operating target from the overnight rate to the quantity of reserves in the banking system, and the phrase quantitative easing entered the vocabulary attached to a country most economists then regarded as a special case. Seven years later that stopped. When the Federal Reserve began buying on a large scale in 2008, Ben Bernanke preferred the term credit easing, on the grounds that the identity of the assets bought was what mattered, and the distinction was lost almost immediately. By 2020 nobody called it unconventional.

The mechanics are less mysterious than the name. Knowing them changes how you read both the easing phase and the tightening phase that follows it.

What happens when the Fed buys a bond

The trading desk at the Federal Reserve Bank of New York buys Treasury securities and agency mortgage-backed securities from primary dealers in the open market, acting on an implementation note published with the statement, which directs the desk in plain operational language to purchase securities in the amounts needed to achieve whatever the committee has decided. In March 2020 that instruction was written in terms of the amounts needed to support smooth market functioning, which was an unusual thing for a central bank to say out loud and told you precisely what the problem was that week.

Payment is made by crediting the dealer’s bank with new reserve balances, which the Fed is able to create because reserves are its own liability. Two entries appear. On the asset side, the bond. On the liability side, the reserves. The bank that received them now holds a deposit at the central bank in place of a bond, and the investor who sold holds cash in place of a bond.

Side	Main items	What changes under QE
Assets	Treasury securities, agency mortgage-backed securities, loans	Rise as the portfolio is bought
Liabilities	Bank reserves, currency in circulation, the Treasury General Account, reverse repo balances	Reserves rise by roughly the purchase amount
Capital	A small, fixed sliver	Barely moves

The liability side deserves more attention than it usually gets, because those items compete for space with each other. Currency in circulation is demand-driven and grows slowly. The Treasury General Account is the government’s checking account at the Fed, and when it is filled by tax receipts or debt sales, reserves fall by the same amount without the Fed having done anything at all. The overnight reverse repurchase facility is where money market funds park cash. A given size of balance sheet can therefore sit alongside very different quantities of bank reserves depending on what the other liabilities are doing, which is why the headline total tells you considerably less than the split within it.

Three channels, and what the research supports

The portfolio balance channel is the mechanical one. Long-dated bonds carry duration risk, somebody has to hold it, and when the central bank absorbs a large quantity of it the remaining private holders require less compensation for what is left, so the term premium embedded in long yields compresses. The arithmetic of duration sits in duration and interest rate risk, and the whole channel is an application of it at the scale of a market.

The signalling channel works through expectations. A central bank that has committed to buying bonds for the next year is telling the market it does not intend to raise the policy rate any time soon, and since a long yield is roughly the average of expected short rates plus a term premium, changing the expected path moves the long yield directly, a relationship set out in the yield curve explained. It is also why announcement effects have consistently been larger than the effects of the purchases themselves, a pattern found repeatedly in the academic literature and one that makes intuitive sense: the information arrives with the announcement, and the buying merely carries it out.

The liquidity channel matters in a crisis. Otherwise, hardly at all. In March 2020 the Treasury market, which is supposed to be the deepest market in the world, stopped functioning smoothly as dealers proved unable to absorb the selling, and yields on the most liquid securities moved in ways that made no sense against fundamentals. Large purchases restored two-way trading within days. That episode is the clearest case on record of QE accomplishing something no other tool could have accomplished, and it is a narrower claim than the one usually made for the policy.

What the body of research broadly supports is that each round of purchases moved long yields by a modest amount, that the effect was largest when markets were stressed and smallest when they were calm, and that later rounds achieved less than earlier ones. Proportionality does not survive the evidence.

What it does to a portfolio

Lower long yields raise the present value of distant cash flows, which is the same discount-rate arithmetic set out in how interest rates affect stocks. They also lower the return available from safe bonds, which pushes investors along the risk spectrum into credit, equities and property, and that push is the point of the exercise.

The bond side can be made concrete. A 10-year Treasury note with a modified duration near 8 gains roughly 8 x 0.25 = 2% in price for a quarter-point fall in its yield. Holders of long bonds are therefore paid immediately when a purchase program is announced, before any of the intended effects on employment have had time to occur, which is the substance of the complaint that QE reaches asset owners first and wage earners much later. That complaint is about timing. The arithmetic is hard to argue with.

Tightening by attrition

QT is the unwind. It is normally done by not doing something. No bonds are sold into the market. The Fed simply stops reinvesting the principal repaid on securities as they mature, up to a monthly cap, so the portfolio shrinks passively while the schedule is published in advance and the market can plan around it. The Fed set out this approach in a 2017 addendum to its normalization principles, and again in plans published in May 2022, both times with the caps written down before the process began. Announcing the mechanism in advance is deliberate, because the alternative is having the market guess.

The arithmetic fits on paper. A cap of $60 billion a month permits 60 x 12 = $720 billion of runoff a year, assuming enough securities mature to fill it. Taking a portfolio down by $1.5 trillion at that pace requires 1,500 / 60 = 25 months. Caps are typically reduced as the process advances, which stretches the timeline further, and mortgage-backed securities run off more slowly than scheduled whenever high rates discourage refinancing, since prepayments are what retire them early.

On the liability side, runoff destroys reserves. Every dollar of maturing principal the Fed declines to reinvest is a dollar the Treasury pays out of its account at the Fed, funded by issuing new debt to private buyers, and reserves in the banking system fall accordingly. That is the part which requires care. Here the exit stops resembling the entry.

Why the exit is harder than the entry

Nobody knows how many reserves the banking system needs until the day it has too few, since banks hold reserves for payment settlement and to satisfy liquidity regulation, and that demand shifts with the rules, with the structure of the system and with how nervous treasurers happen to be.

The lesson arrived in September 2019, before the pandemic and in an otherwise unremarkable week. Reserves had been drained far enough that overnight repo rates spiked well above the fed funds target and stayed there for several days, which is not supposed to be possible in a system with a floor. The Fed injected liquidity and bought Treasury bills. The episode established that the floor for reserves was materially higher than the committee had assumed, and it is the reason a standing repo facility now exists as a release valve. Anyone tempted to treat the balance sheet as a dial that can be turned smoothly in either direction should keep that fortnight in mind, and should watch the front of the curve on the yield curve tool when the question comes round again.

There is a fiscal dimension as well. The Fed earns interest on the bonds it holds and pays interest on the reserves it created. When short rates rise above the average yield of a portfolio assembled in a low-rate period, it pays out more than it takes in, remittances to the Treasury stop, and the shortfall is carried as a deferred asset until earnings recover. This has no bearing on the Fed’s ability to conduct policy, and it reliably attracts congressional attention, which are two separate facts that often get argued as one.

As for whether any of this causes inflation, the honest answer is that the record does not settle it, because the programs following the 2008 crisis ran alongside inflation that stayed below target for most of a decade, which is hard to square with a simple money-creation story. The pandemic expansion coincided with large direct transfers to households and with broken supply chains, so attributing what followed to asset purchases requires separating effects that arrived in the same eighteen months. Anyone who tells you the question is closed is arguing from conviction.

See how the balance sheet sits alongside the rate decision in how the Federal Reserve works, or look at what signals the end of a cycle in recession indicators, and the Fed and the economy quiz covers the balance sheet mechanics.