Key Highlights

  • The Fed's balance sheet has grown to 6.7 trillion dollars, or nearly 21% of GDP, with Treasuries making up about two-thirds of holdings.
  • Bank reserves have risen from around 10 billion dollars before the financial crisis to about 3.1 trillion dollars today.
  • A prior unwinding attempt between 2017 and 2019 caused overnight interest rates to spike in September 2019 when reserves ran short.
  • The Fed expanded its balance sheet again from mid-December to mid-April, buying Treasury bills at roughly 40 billion dollars a month.

A Balance Sheet Built Through Successive Crises

During the global financial crisis of 2007-09, the Federal Reserve first began buying mortgage-backed securities to steady the housing market, later expanding into long-dated Treasuries to stimulate the broader economy through a policy known as quantitative easing. Three rounds of QE between 2009 and 2014, followed by another during the covid-19 pandemic in 2020, have swelled the Fed's balance sheet to 6.7 trillion dollars, or nearly 21% of GDP, with Treasuries accounting for about two-thirds of total holdings. Fed Chairman Kevin Warsh, long suspicious of central bank expansion, previously criticised the large balance sheet as a proxy for the Fed's growing influence over the economy.

In pursuit of a smaller balance sheet, Warsh has assembled a task force including Harvard economist Karen Dynan, former Reserve Bank of India head Raghuram Rajan, and former Fed governor Jeremy Stein, tasked with determining whether the reduction can be accomplished safely.

Weighing the Costs and Risks of a Large Balance Sheet

Several objections have been raised against the Fed's expansive holdings. If quantitative easing works as a macroeconomic stimulus tool, a question economists continue to debate, it does so by distorting the market for long-term debt, and a prolonged intervention sustains that distortion. Holding long-term bonds has also proven costly as interest rates have risen, since the Fed pays commercial banks a floating rate on the reserves it creates to purchase bonds, while the yields it earns on those bonds were locked in at the time of purchase, often at ultra-low pandemic-era levels.

Warsh's primary concern, however, appears more institutional in nature. Although the Fed purchases government bonds in secondary markets rather than directly from the Treasury, holding trillions of dollars' worth can still create the appearance of functioning as the government's piggy bank. One academic economist noted that these criticisms currently remain mostly about perceptions rather than substantive risk, though there is concern that such perceptions could eventually become reality if left unaddressed.

A Banking System Now Structurally Dependent on Abundant Reserves

For the balance sheet to shrink, its liabilities must shrink correspondingly, and the analysis shows limited room to manoeuvre. Currency in circulation makes up about 2.5 trillion dollars of the Fed's liabilities and cannot be reduced, while the Treasury's deposit account at the Fed adds roughly 800 billion dollars and may offer some flexibility. The largest liability by far is reserves, which have risen from around 10 billion dollars before the financial crisis to about 3.1 trillion dollars today, meaning any meaningful balance sheet reduction requires draining reserves from the banking system.

The difficulty is that the banking system has become structurally built around abundant reserves, much as a city might gradually reshape itself around a newly abundant water supply. Before the financial crisis, reserves paid no interest, giving banks incentive to lend excess reserves to each other at prevailing short-term rates. As QE flooded the system with reserves, the Fed maintained control over short-term rates by paying interest directly on reserves, reducing banks' incentive to lend them elsewhere, while new liquidity regulations requiring banks to demonstrate they can meet their needs using their own liquid assets pushed in the same direction. Banks also adapted their broader business models to this new environment: when the Fed purchased Treasuries from non-bank investors, the proceeds often landed as uninsured deposits in those investors' bank accounts, funding that in turn allowed banks to expand their investment portfolios and extend more credit, creating what one economist described as a ratcheting effect in which each successive round of QE left banks dependent on progressively higher reserve levels.

Past Attempts Reveal the Risks of Draining Too Far

Judging how low reserves can fall without causing dysfunction has proven genuinely difficult in practice. During the Fed's first attempt to unwind its balance sheet between 2017 and 2019, it reduced reserves gradually, hoping to leave sufficient levels for the financial system to function smoothly. In September 2019, however, reserves ran short and overnight interest rates spiked sharply, exposing the limits of that approach. Similar strains reappeared more recently as reserves approached what some viewed as a low-water mark, prompting the Fed to respond by purchasing Treasury bills at a pace of around 40 billion dollars a month from mid-December to mid-April, once again expanding its balance sheet to replenish the reserve pool.

Various proposals have emerged to address the underlying dynamic. One former Fed governor and critic of balance-sheet expansion has proposed cutting the interest rate paid on reserves below what banks could earn lending in the federal-funds market, while another academic favours updating Fed payment settlement software to offset incoming and outgoing payments before final settlement, potentially allowing banks to operate with smaller reserve buffers, an approach already used by the Bank of England and European Central Bank. Rajan has separately floated the idea of allowing the first institution caught in a crisis to fail, as a way of disciplining banks against excessive risk-taking in an environment where abundant reserves encourage credit growth.

Conclusion

Even Warsh has acknowledged that a balance sheet expanded gradually over 18 years will take considerably longer than 18 weeks to meaningfully shrink, and given the structural dependence the banking system has developed around abundant reserves, a full reversal appears unlikely in the near term. Rather than draining the balance sheet outright, Warsh may instead focus on changing its composition, replacing long-dated bonds with short-term Treasury bills as they mature, an approach that would reduce the Fed's influence over longer-term borrowing costs without necessarily reducing the balance sheet's overall size.