Repo might be even bigger than we thought

Finance is anthropological

That's Zoltan Pozsar, the Hungarian-American economist who mapped the plumbing of modern money before most people knew there was plumbing to map. When he said it to Bloomberg in 2019, he was trying to explain why repo markets (the overnight lending infrastructure that lubricates trillions in daily transactions) had just seized up in ways the Federal Reserve didn't anticipate.

I've written about Pozsar's work before, particularly his "Bretton Woods III" thesis about the shifting role of the dollar. But his earlier research on shadow banking and repo markets feels increasingly relevant as we enter 2026. In December 2025, the Office of Financial Research published new data on the size of the U.S. repo market. The number: $12.6 trillion in average daily exposures. That's roughly $700 billion larger than previous estimates; a measurement error roughly the size of the entire Swiss banking system.

Where did the extra $700 billion come from? Mostly from what the OFR calls "non-centrally cleared bilateral repo," or NCCBR; the segment of the market that doesn't flow through clearinghouses or the tri-party platforms that regulators can easily observe. This bilateral segment alone accounts for $5 trillion. Until the OFR's new transaction-level data collection, which only reached full implementation in July 2025, much of this activity was essentially invisible.

This matters because repo is not a peripheral market. It is the market through which cash-rich institutions lend to cash-poor ones, every single day, against collateral. Money market funds, hedge funds, broker-dealers, asset managers, banks. When repo works, it's invisible. When it doesn't, as in September 2019, overnight rates spike and the Fed scrambles to inject liquidity.

On December 31, 2025, eligible financial firms borrowed a record $74.6 billion from the Fed's Standing Repo Facility, which is the highest since its launch in 2021. The Fed had just eliminated the $500 billion daily cap on this facility, a quiet acknowledgment that the ceiling might actually matter. Quantitative tightening officially ended on December 1, 2025. Reserves had fallen to $2.8 trillion, their lowest in four years.

The plumbing was straining again.

Pozsar's 2014 OFR paper, "Shadow Banking: The Money View," introduced a framework that still haunts anyone who reads it carefully. At its core is a hierarchy of money. Currency sits at the top, the liability of the sovereign. Below that: bank deposits, insured and backstopped by the FDIC. Below that: repo, secured by collateral but not by any explicit government guarantee. Below that: the constant-NAV shares of money market funds, which promise par redemption but rest on layers of private credit puts, reputational commitments, and the fragile assumption that nothing will go wrong simultaneously.

The key insight is that what counts as "money" depends on where you sit in this hierarchy. For a retail depositor, money is an insured bank balance. For a corporate treasurer managing $50 billion in cash, money begins where M2 ends—in repo, in money fund shares, in instruments that offer some semblance of safety at scale but lack the explicit backstops that smaller depositors take for granted.

Pozsar called these institutions "cash pools"—the corporate treasuries, sovereign wealth funds, and asset managers whose cash balances are too large to fit within the insured deposit system. They need money-like instruments, but the supply of truly safe assets (Treasury bills, insured deposits) is inelastic. So they reach for the next best thing: shadow money claims backed by private collateral and private liquidity puts.

Now, the new OFR data reveals that $5 trillion of daily repo activity, roughly 40% of the market, occurs in bilateral arrangements that, until recently, were largely opaque to regulators. The collateral backing this activity is 61.8% Treasuries, but that leaves substantial room for corporate bonds, agency MBS, and other assets that can gap in value during stress.

Pozsar's 2019 Global Money Notes described the repo market as a hierarchy with dealers at the center and the Fed at the top, operating as a "dealer of last resort" when private balance sheets reach their limits. The Standing Repo Facility was supposed to institutionalize this role, providing a ceiling on overnight rates by offering funding at a known price.

The facility sat unused for years while reserves were abundant. Now, as reserves decline, usage is spiking at quarter-ends and year-ends, exactly when balance sheet constraints bind hardest. The question Pozsar raised in 2019 remains unanswered: can the Fed operate a standing repo facility that polices the top of its target range without losing control over its balance sheet size? Or will it be forced, eventually, to monetize excess collateral on a scale that looks a lot like QE by another name?

There's a concept in infrastructure studies called "seamful design": the idea that making the seams of a system visible can improve rather than degrade the user experience. GPS, for instance, became more useful when designers surfaced uncertainty estimates rather than hiding them.

The repo market is the opposite: seamless by design, invisible until it fails. The OFR's new data collection is, in some sense, an attempt to add seams, to make visible what was hidden, to understand the shape of the beast before the next crisis. But measurement is not control. Knowing the market is $12.6 trillion doesn't tell you what happens when a major counterparty fails, or when a category of collateral suddenly trades at distressed prices, or when the behavioral assumptions embedded in banks' liquidity models turn out to be wrong.

Pozsar understood this intuitively. His famous map of the shadow banking system, posted in the New York Fed's briefing room, required zooming in seven or eight times to read any detail. Colleagues who didn't take the time to study it, he warned, were looking at "10% of the picture."