Gold's London-COMEX EFP dislocation

Two markets, one arbitrage

Gold trades in two linked places. In London, bullion changes hands over the counter as loco spot gold, in standard "good delivery" bars of about 400 troy ounces. In New York, the COMEX exchange lists a futures contract on 100 troy ounces of gold, physically deliverable into COMEX-approved vaults in either 100-ounce or 1-kilo bars. Ordinarily the two prices are held together by a cash-and-carry arbitrage: a bank can buy a bar in London, ship it to New York, have it recast into COMEX-eligible size, and deliver it into the future, or run the trade in reverse. The gap between the COMEX future and London spot that survives this arbitrage is quoted as the EFP, the Exchange for Physical, and in normal times it costs only about $1.50 an ounce, roughly the cost of financing and moving the metal for a short period.

The chain breaks

In March 2020, that arbitrage stopped working. On 11 March, President Trump announced US travel restrictions on passenger flights from Europe, taking effect on 14 March, grounding much of the transatlantic passenger fleet that normally carries gold bars in its cargo hold. On 23 March, the three big Swiss refineries in the canton of Ticino, Valcambi, Argor-Heraeus and PAMP, which between them recast most London 400-ounce bars into the smaller sizes COMEX accepts, suspended production under a Ticino lockdown order. Shipping and refining, the two physical steps that keep the London and New York gold prices in line, both failed in the same fortnight.

Banks caught short

Banks that were short the COMEX future, expecting to deliver London bars against it, found they could neither move the metal nor reshape it in time. Instead of delivering, they had to buy back their futures positions, and that buying pushed the COMEX price further above London spot. On 24 March, Bloomberg described the COMEX April gold future as trading at its highest premium to London spot in roughly four decades, and the London Bullion Market Association issued a statement acknowledging the "impact on liquidity arising from price volatility in Comex 100oz futures contracts". The EFP, which should have stayed near $1.50, reached about $60 an ounce on 25 March 2020, with intraday quotes on the New York premium touching roughly $70. An even wider range appears in some press coverage for 24 March: futures trading $80 to as much as $100 to $120 above spot. That figure is best read as a snapshot of a more volatile moment on a fast-moving day, possibly using a different quote convention, rather than as a separate fact to reconcile against the case's headline numbers.

CME's fix, and the reversal

CME Group moved quickly. On 24 March 2020, the same day the dislocation was making headlines, it announced a new contract, Gold (Enhanced Delivery), ticker 4GC, that would let a trader deliver a London-standard 400-ounce bar directly into a COMEX position, alongside the existing 100-ounce and kilo bars. That removed the need to reshape the metal before delivery, the step the Swiss refinery shutdown had broken. The new contract began trading on 6 April 2020, with a first expiry that same month. On 1 April, the LBMA and CME Group issued a joint statement reporting London vault holdings of about 8,326 tonnes and COMEX New York depository holdings of about 9.2 million ounces, intended to reassure the market that there was no true physical shortage of gold, only a blocked route between the two locations. The EFP narrowed over the following weeks as Swiss refineries reopened and gold shipments, including record Swiss exports, flowed into New York, though it was still reported around $15 an ounce by 30 April, well above its pre-crisis level. By late May the position had reversed so far that Bloomberg reported the opposite problem: so much gold had flowed into New York that COMEX vaults were running a glut, and traders who had paid to close short positions in March were now struggling to avoid being forced to take delivery themselves.

Why the case matters

Gold is the textbook example of an investment asset with no income and next to no storage cost, the case where both arms of the cash-and-carry arbitrage are meant to work cleanly and the no-arbitrage relationship should hold almost exactly. That is what makes the episode instructive. The formula that ties the COMEX future to London spot did not change and was never in doubt. What temporarily stopped working was the physical machinery, aeroplanes and refineries, that the market relies on to enforce it every day without anyone thinking about it.

Gold is a clean case for the cost-of-carry formula because it pays no income and costs almost nothing to store for a short period. The general cost-of-carry formula collapses close to \(F_0 = S_0 e^{rT}\), and both arms of the no-arbitrage argument are meant to work. If the future trades too high relative to spot, a bank can borrow dollars, buy spot gold in London, and sell the future, a cash-and-carry trade. If the future trades too low, a bank can sell spot short and buy the future, a reverse cash-and-carry trade. Because both directions are available, gold's no-arbitrage relationship is normally an equality, not just a bound, and the tiny residual gap the market actually quotes between the COMEX future and London spot, the EFP, sits close to the pure financing cost.

What running that trade requires in practice is where the story breaks down. The cash-and-carry trade needs three physical steps: hold or buy a 400-ounce bar in London, ship it to New York, usually in the cargo hold of a scheduled passenger flight, and, since COMEX only accepts 100-ounce or 1-kilo bars, recast it at one of a handful of accredited refineries, most of them in Switzerland. In March 2020, steps two and three both failed within the same ten days. Banks that were short the COMEX future could not physically deliver the London metal they were holding against it, so they closed out by buying back futures instead, and that buying is exactly what drove the COMEX price away from London spot.

This is a basis-risk and market-design story built on Lecture 5's core idea: a forward or futures price equals spot carried forward at the net cost of holding the asset, and it is arbitrage activity, not the formula on its own, that forces that relationship to hold day to day. When the arbitrage trade becomes physically impossible to execute, even a genuinely low-risk spread can detach from its formula value, because the mechanism that pins it down has stopped operating. CME's fix worked because it addressed the mechanism directly: allowing direct delivery of the London bar removed the recasting step that had broken, rather than trying to influence the price itself.