Eurodollar market

A eurodollar is a US-dollar-denominated deposit liability of a bank office located outside the United States. The "euro-" prefix is a historical accident (the early centre was London); it has nothing to do with the euro currency, and the market long ago spread to Singapore, Hong Kong, Tokyo, the Gulf and the Caribbean. The precise modern term is the offshore dollar market, measured by the BIS in the locational banking statistics as positions of banks by residence.

What makes it interesting is not geography but balance sheets. A eurodollar is a dollar claim created by a bank that is not subject to US bank regulation, holds no Federal Reserve account, cannot obtain Fed liquidity directly, and creates dollar credit by the identical mechanism a domestic bank does: it lends, and in lending, it creates a deposit. MS1.01 established that loans create deposits; the eurodollar system is that sentence with "and the bank need not be American, supervised by the Fed, or able to reach it" appended.

How is a eurodollar created?

Work the balance sheet, because the section collapses into it (Worked Example 1 does it in full). A dollar deposit at a New York bank moves to a London bank. The New York bank's balance sheet does not shrink by a cent, the ownership of the deposit changes to the London bank, which now holds it as a correspondent (nostro) balance. But the London bank's balance sheet expands: an asset (a claim on the New York bank) and a liability (a dollar time deposit). A dollar claim exists that did not exist a moment before. When the London bank lends dollars, it creates another dollar deposit and expands again. Total US bank deposits are unchanged throughout; total Fed reserves are unchanged throughout; dollar credit outstanding has grown.

How far can eurodollar creation go?

How far can this go? In the textbook telling, a reserve requirement caps it. That telling is doubly obsolete. US reserve requirements on Eurocurrency liabilities were set to zero with effect from 27 December 1990; and on 15 March 2020 the Federal Reserve Board announced that all reserve requirement ratios would fall to zero percent effective 26 March 2020 (Federal Reserve Board, Reserve Requirements). There is now no reserve-requirement ceiling on dollar creation anywhere, onshore or off. What constrains an offshore bank is prudential: correspondent balances held for payment purposes, capital, the liquidity coverage ratio, internal limits, and (the binding one in a crisis) the ability to roll dollar funding it does not print.

The arithmetic is unchanged; only the ratio's interpretation is. If an offshore bank holds a fraction r of its dollar deposits as correspondent balances, the system supports deposits of base ÷ r. Verified: with a base of $500 million and r = 20%, the offshore system supports $2,500 million of dollar deposits and therefore $2,000 million of dollar credit, the geometric series 500 + 400 + 320 + 256 + 204.8 + … converging exactly to 2,500. Halve the ratio to 10% and the same base supports $5,000m of deposits and $4,500m of credit. Nothing in that arithmetic requires any American institution's permission.

Why does the eurodollar market exist, and can the Fed back it?

Why the market exists, in descending order of durability: regulatory arbitrage (Regulation Q ceilings, reserve requirements, deposit-insurance assessments), the original driver, now mostly competed away; convenience and time zones (a Japanese exporter, a Brazilian treasurer and a Gulf fund all need dollar balances managed in their own working day); and the dollar's invoicing and funding dominance, which is self-reinforcing. The BIS global liquidity indicators put US dollar credit to non-bank borrowers outside the United States at $14.3 trillion at end-December 2025, growing 8.5% over the year (the fastest since Q3 2014) with $4.3 trillion owed by emerging market and developing economies (BIS, Global liquidity indicators at end-December 2025, released 30 April 2026). None of that credit is a liability of the Federal Reserve, insured by the FDIC, or reachable by an ordinary Fed operation.

Now ask question 5: who is the backstop? The Fed can create dollar reserves at will. It cannot create a eurodollar, because a eurodollar is somebody else's liability, and no central bank can issue another institution's liability; the Fed supplies reserves only to institutions with Fed accounts. So when the offshore system cannot roll its dollar funding, the Fed's ordinary tools do not reach it, the hole that swap lines and the FIMA repo facility exist to fill (see central bank swap lines). That is the single most important structural asymmetry in the global monetary system, and almost every "dollar shortage" headline is a consequence of it.

How do you measure eurodollar credit in Python?

Modern Data Analysis, the Python way. The offshore-dollar picture is measurable: the BIS Data Portal exposes the locational and consolidated banking statistics and the global liquidity indicators as bulk CSV and via a public SDMX REST API, so pandas.read_csv on a bulk file, or requests.get against the SDMX endpoint parsed into a DataFrame, gets you cross-border dollar claims by counterparty country, sector and reporting-bank nationality, quarterly, back decades. The useful analytics are two lines each: .pivot_table(index="period", columns="counterparty_sector", values="value") to separate bank-to-bank from bank-to-non-bank dollar credit, and .pct_change(4) for year-on-year growth, which is what actually turns in a funding squeeze. Three traps make this worth doing by hand once: BIS series are in millions of US dollars with an explicit unit-multiplier column you must read rather than assume; the LBS include an exchange-rate-adjusted change series that is not the difference of the stocks and is the correct one for growth analysis; and "cross-border" and "foreign currency" are different cuts a careless filter will silently mix. The quant-code region (DA/QD) builds the full pipeline; this node's job is that you know which cut answers which question and can state, before running anything, what the answer should look like, so a wrong join announces itself.