The Great Unwind That Wasn’t

After three years of holding its balance sheet steady at around $4.5 trillion — the “reinvestment phase” where maturing bonds were replaced with new purchases — the Fed announced in September 2017 that it would begin balance sheet normalization. Starting in October, it would let bonds mature without reinvesting the proceeds, effectively draining reserves from the banking system at a measured pace.

The plan was modest by design. In the first month, the Fed would let just $10 billion roll off ($6 billion in Treasuries, $4 billion in MBS). Each quarter, the cap would rise by $10 billion until reaching a maximum pace of $50 billion per month. At that rate, it would take roughly five years to return the balance sheet to something approaching pre-crisis normality — perhaps $2.5 to $3 trillion.

Janet Yellen, who had designed the plan before handing the chair to Jerome Powell in February 2018, described it as something that should run “quietly in the background,” like “watching paint dry.” The metaphor was deliberate: after the drama of QE, the Fed wanted the reverse to be as boring as possible. Markets would barely notice.

For the first year, the paint-drying analogy held. The balance sheet declined from $4.46 trillion in October 2017 to $4.14 trillion by October 2018 — a smooth $320 billion reduction over twelve months. The runoff hit its maximum pace of $50 billion per month by Q4 2018. Stocks were rising. Growth was solid. Nobody paid attention.

Then the market cracked. In December 2018, the S&P 500 plunged nearly 20% from its September peak. The proximate cause was a trade war with China and hawkish Fed rhetoric (Powell had called rates “a long way from neutral” in October). But many participants blamed QT. The combination of rate hikes and balance sheet shrinkage was tightening financial conditions from both ends simultaneously.

President Trump was furious, tweeting that the Fed was the “only problem our economy has” and comparing Powell unfavorably to a golfer who “can’t putt.” Trump’s sustained public attacks on the Fed were unprecedented in modern times — and while Powell maintained the institution’s independence, the political pressure was real.

In January 2019, the Fed blinked. Powell signaled that rate hikes were on hold and that the Fed would be “patient” with further tightening. In March, the Fed announced that QT would end by September 2019 — much earlier than originally planned. The balance sheet would settle at approximately $3.5 trillion, not the $2.5 trillion some had expected. The unwind was being unwound.

On September 17, 2019, the overnight repurchase agreement (“repo”) rate — the interest rate at which banks and financial institutions borrow cash overnight using Treasury securities as collateral — spiked from its usual 2% to 10%. Some trades reportedly executed at rates as high as 15%. For a market that normally moves in basis points (hundredths of a percent), this was the equivalent of an earthquake.

The repo market is the plumbing of the financial system. It’s where banks, money market funds, and hedge funds borrow and lend cash every night, usually in amounts measured in hundreds of billions of dollars. When the repo rate spikes, it means someone needs cash badly and can’t get it. When it spikes to 10%, it means the system is breaking.

What happened? Two things collided on the same day: quarterly corporate tax payments (which drained about $100 billion from bank accounts to the Treasury) and a large settlement of Treasury auctions (which required dealers to pay for newly purchased government bonds). Both events sucked cash out of the banking system simultaneously. Normally, banks would have enough excess reserves to absorb these flows. But QT had been draining reserves for two years, and they had fallen from about $2.2 trillion in late 2017 to $1.39 trillion in the week of September 18.

The level wasn’t zero — far from it. But reserves were distributed unevenly. The largest banks (JPMorgan, Bank of America, Citigroup) held the lion’s share and weren’t willing to lend them in the repo market at 2% when they could earn 2.4% on IOER at the Fed. Post-crisis regulations also discouraged banks from deploying reserves into short-term lending. The result was a plumbing failure: there was enough water in the system, but it couldn’t get to where it was needed.

The Fed responded immediately. On September 17, the New York Fed began conducting overnight repo operations — lending cash to dealers against Treasury collateral — for the first time since the financial crisis. Initially intended as a one-time intervention, the repo operations became a daily fixture. The Fed was once again injecting cash into the system, barely six months after it had stopped draining it.

In October 2019, the Fed went further. It announced it would begin purchasing $60 billion per month in Treasury bills — not QE, officials insisted, but “reserve management operations.” The distinction was semantic. The balance sheet, which had fallen to $3.76 trillion in September, began growing again. By January 2020, it was back up to $4.17 trillion. The repo crisis had forced the Fed to abort QT and begin expanding the balance sheet just five months after the formal end of tightening.

The lesson was brutal and simple: the Fed didn’t know how many reserves the banking system needed. For two years, it had been draining reserves on the assumption that $1.5 trillion or so would be “enough.” The repo crisis proved that the minimum comfortable level was considerably higher — perhaps $1.5 to $1.7 trillion — and that the distribution of reserves mattered as much as the total. The Fed had been flying blind.

QT1 revealed a fundamental truth about the post-QE financial system: the balance sheet expansion is largely irreversible. The Fed inflated its balance sheet by $3.6 trillion over six years. It managed to unwind just $700 billion — less than 20% of the total — before the system broke. The remaining $3 trillion had become structural: the banking system had reorganized itself around abundant reserves, and removing them faster than the system could adapt caused acute stress.

This had profound implications. It meant that the “new normal” for the Fed’s balance sheet was not $900 billion (pre-crisis) or even $2.5 trillion (the original target). It was something north of $4 trillion — a floor that would only rise over time as the economy grew and new regulations required banks to hold more liquid assets. The balance sheet had ratcheted up permanently.

In January 2019, the Fed formally adopted the “ample reserves” framework, acknowledging that it would operate with a much larger balance sheet than it had before the crisis. The old system of “scarce reserves,” where the Fed fine-tuned short-term rates by adding or removing small quantities of reserves, was dead. The new system would keep reserves abundant and control rates through administered rates (IOER and the reverse repo facility) rather than market operations.

QT1 was supposed to prove that QE was reversible — that the extraordinary measures of the crisis could be unwound and the system returned to normal. Instead, it proved the opposite. The balance sheet was a one-way ratchet. And then COVID arrived, and the ratchet cranked up to levels nobody had imagined possible.