Bank Reserves and the Plumbing

Bank reserves are the most important form of money you’ve never seen. They exist only as digits in accounts that commercial banks hold at the Federal Reserve — not as bills, not as coins, not as anything you could withdraw from an ATM. When Bank of America transfers $500 million to Citibank to settle a day’s worth of customer payments, those reserves move between their respective accounts at the Fed. When the Treasury Department sends out Social Security checks, reserves flow from the government’s account to bank accounts. Every large payment in the United States ultimately settles through reserve transfers at the Fed.

Think of reserves as the blood supply of the financial system. Individual transactions — your mortgage payment, your paycheck, your credit card bill — might involve multiple intermediaries, but they all ultimately clear through the reserve system. If a bank doesn’t have enough reserves, it can’t settle its obligations. And if enough banks can’t settle, the entire payments system grinds to a halt.

For most of the Fed’s history, reserves were scarce by design. The Fed controlled the money supply by adjusting the tiny pool of reserves available. Banks that needed reserves for the day would borrow them from banks that had a surplus, at the federal funds rate — the most important interest rate in the world. The Fed influenced this rate by adding or draining small amounts of reserves through open market operations. The whole system balanced on a fulcrum of roughly $10–50 billion in total reserves. It was elegant, efficient, and — as 2008 would prove — fragile.

In January 2008, total reserve balances at the Federal Reserve stood at $13 billion. Twelve months later, in January 2009, they stood at $878 billion. Reserves had increased by a factor of 67 in a single year. Nothing like it had ever happened before — not in the Fed’s history, not in any central bank’s history. The entire operating framework of American monetary policy changed overnight.

The reason was QE1. When the Fed bought $1.7 trillion in Treasuries and MBS from banks and dealers, it paid for them by crediting the sellers’ reserve accounts. The money didn’t come from anywhere — the Fed simply typed new digits into existence. This is what “printing money” actually means in the modern era: not running a press, but entering numbers into a database at the Federal Reserve Bank of New York. Each bond purchase added reserves to the system. And because the Fed was buying trillions of dollars in bonds, trillions of dollars in reserves appeared.

The old system of scarce reserves was instantly obsolete. When every bank is swimming in reserves, no bank needs to borrow reserves overnight. The federal funds market — the centerpiece of monetary policy for decades — essentially dried up. Its daily volume fell from $200 billion to under $50 billion. The Fed needed a new tool to control interest rates, so it started paying interest on reserves (IOR). By setting the rate it paid banks to park money at the Fed, it could control short-term rates without needing to fine-tune the supply of reserves. The steering mechanism changed from controlling the quantity of reserves to controlling the price.

Reserves kept growing with each round of QE. After QE2, they crossed $1.5 trillion. After QE3, they peaked at $2.7 trillion in 2014. QT1 brought them back down to $1.4 trillion — and then, as Episode 5 showed, the repo market broke. COVID pushed reserves to an all-time high of $3.86 trillion in January 2022. Even after QT2 drained $2.4 trillion from the balance sheet, reserves in January 2026 still stand at $3.0 trillion — 230 times the pre-2008 level.

The September 2019 repo crisis, which we covered in Episode 5, was the defining stress test for the reserve system. As QT1 drained reserves from $2.2 trillion down to about $1.5 trillion, the system seemed fine. Banks reported ample liquidity. The overnight markets functioned normally. Then, on September 17, 2019, two things happened simultaneously: quarterly corporate tax payments pulled reserves out of the banking system, and a large Treasury settlement required dealers to absorb new government bonds.

The combination was catastrophic. Reserve balances dropped from $1.49 trillion on September 4 to $1.39 trillion on September 18 — a decline of just $100 billion, or about 6.5%. But it was enough. The overnight repo rate, which normally trades at or near the fed funds rate (then 2.0–2.25%), spiked to 10%. Some trades reportedly cleared at even higher rates. For a few hours, the most liquid short-term market in the world had seized up.

The lesson was brutal and immediate: $1.4 trillion in reserves was not enough. The banking system’s minimum comfortable level was somewhere above that number, and the Fed had found the boundary by crashing into it. Within days, the Fed began repo operations to inject reserves back into the system. By December 2019, reserve balances were back above $1.6 trillion. The whole episode lasted weeks, but it permanently reshaped the Fed’s thinking about how much liquidity the system needed.

The specific dynamics revealed why reserves couldn’t return to pre-2008 levels. Post-crisis regulations — the Liquidity Coverage Ratio, the Supplementary Leverage Ratio, internal stress testing requirements — all forced banks to hold far more liquid assets than before. JPMorgan alone needed hundreds of billions in reserves to satisfy its regulatory requirements. The four largest banks collectively held over $800 billion in reserves even when systemwide totals looked comfortable. Reserves had become structural rather than operational — locked up in regulatory buffers rather than freely circulating.

The Fed now describes three possible states for reserves: abundant, ample, and scarce. Understanding the distinction is key to understanding why QT ends when it does.

In the abundant regime, reserves are so plentiful that banks don’t even think about them. There’s no competition for reserves, no pressure on overnight rates, no stress in funding markets. The fed funds rate sits right at the interest-on-reserves rate because banks have no incentive to lend reserves at a lower rate. This was the state during QE and COVID — roughly $2.5 trillion and above.

In the ample regime, reserves are sufficient but not excessive. Banks are comfortable but attentive. Overnight rates may fluctuate slightly around the target. Small shocks — a large Treasury settlement, a tax payment date — can cause temporary pressure, but the system absorbs them. This is where the Fed wants to operate: enough reserves for smooth functioning, but not so many that monetary policy loses traction. The ample regime sits in a band that the Fed estimates at roughly $2.5–3 trillion in today’s system.

In the scarce regime, reserves are insufficient. Banks hoard what they have. Overnight rates spike. Funding markets freeze. This is what happened in September 2019, and it’s what the Fed is desperate to avoid. The scarce regime begins below approximately $1.5 trillion — though nobody knows the exact threshold until they hit it, which is precisely the problem.

TOTRESNS — Total Reserves of Depository Institutions — offers the longest continuous view of reserves in the U.S. banking system. The series begins in 1959 and captures a remarkable story of stasis followed by revolution. For the first fifty years, from 1959 through 2008, total reserves fluctuated in a narrow band between $19 billion and $63 billion. The entire American banking system — processing trillions of dollars in payments — operated on a sliver of reserves thinner than the cash drawer of a single large bank today.

Then came 2009: $860 billion. The chart doesn’t slope upward — it detonates. Every decade of data from 1959 to 2008 looks like a flat line at the bottom of the chart. The fifty-year average of $42 billion is rounding error compared to the post-crisis world. The shift was so dramatic that it’s hard to display both eras on the same chart without making the old regime invisible.

The peak came in January 2022, at $3,871 billion — nearly 86 times the 2008 level. Even after QT2’s $2.4 trillion drain, reserves in January 2026 stand at $2,968 billion — 66 times the pre-crisis norm. The old world of $40 billion in reserves is as distant as the gold standard. It isn’t coming back, and the Fed isn’t trying to bring it back. The “ample reserves” framework is now permanent policy.

The pre-2008 world of $10–50 billion in reserves was a product of a specific regulatory regime, a specific set of banking practices, and a specific level of trust. All three have changed irreversibly.

First, the regulations. The Dodd-Frank Act and Basel III requirements, implemented after 2008, force banks to hold far more liquid assets than before. The Liquidity Coverage Ratio (LCR) requires large banks to hold enough high-quality liquid assets — mainly Treasuries and reserves — to cover 30 days of cash outflows. The Supplementary Leverage Ratio (SLR) counts reserves in its denominator, limiting how much leverage banks can use. These rules effectively lock up hundreds of billions in reserves that can’t be freely circulated.

Second, the structure of banking has changed. The four largest banks — JPMorgan, Bank of America, Citigroup, and Wells Fargo — are collectively far larger relative to the system than in 2008. Their internal risk models require massive reserve buffers. JPMorgan’s internal “lowest comfortable level of reserves” (LCLOR) is reportedly hundreds of billions alone. These aren’t regulatory minimums; they’re self-imposed buffers that reflect the sheer scale of modern banking operations.

Third, the psychology has changed. Before 2008, banks treated reserves as a cost — idle money earning nothing. After the Fed began paying interest on reserves, holding reserves became profitable. Today, with the interest-on-reserve-balances rate at 4.4%, banks earn substantial income simply by parking money at the Fed. There’s no incentive to minimize reserves when they pay a competitive return. The old game of lending every spare dollar overnight and keeping reserve balances as close to zero as possible simply doesn’t exist anymore.