Case studies · 2005
The State Reserve Bureau copper short
A Chinese state trader built a large short position on LME copper betting prices would fall, and got caught in a squeeze as prices kept rising.
What happened
A trained arbitrageur
Liu Qibing was chief trader for the Import and Export Department of the State Regulation Centre for Supply Reserves (SRCSR), the trading arm of China's State Reserve Bureau (SRB), the agency that manages the country's strategic stockpiles of commodities including copper. The SRB had sent him to the London Metal Exchange (LME) for training in 1995, and between 2002 and 2004 he ran a genuine arbitrage between the Shanghai Futures Exchange (SHFE) and the LME, buying copper on one exchange and selling it on the other to capture the price gap between them. That trade reportedly earned the Chinese state around US$300 million over the period. It was hedged and spread-based, a long position on one exchange balanced against a short position on the other.
The bet turns naked
In spring 2005, Liu changed strategy. He began building a large, unhedged short position in LME copper futures, betting that prices would fall. His reasoning, as later reported, rested on two beliefs: that Chinese lending curbs aimed at cooling the property market would soften industrial demand for copper, and that the SRB's own planned sales from its stockpile would add supply and push the price down. By July or August 2005 the short position had reached about 130,000 tonnes for December 2005 delivery. By early September it had grown to roughly 220,000 tonnes (one account puts it at 180,000 tonnes net of a 50,000-tonne buyback), at an average price of about US$3,500 per tonne, due mostly on 21 December 2005.
The price moves the wrong way
Copper did not fall. It rose, and kept rising, through the second half of 2005. Around the same time, China's National Audit Office issued an unusual public rebuke of the SRB, and the agency moved roughly 40,000 tonnes of copper toward the SHFE, which other traders read as an attempt to influence prices. As the price climbed further against Liu's short, other market participants began to suspect, correctly, that a single very large seller stood behind the position and might not be able to deliver 220,000 tonnes of physical copper when the contracts matured. Then, on 14 November 2005, Liu disappeared from public view. SRB officials initially denied knowing him and denied that the agency had been short selling at all.
The state defends the position
The SRB faced a choice, with the position still open and the price still rising: let it default, or defend it. It chose to defend it. On 16, 23 and 30 November 2005 the agency held a series of domestic auctions, releasing stockpiled physical copper onto the market, about 20,000 tonnes on each of the first two dates. The auctions revealed that the SRB's true reserves were far larger than the market had assumed, around 1.3 million tonnes, against market guesses of only about 250,000 tonnes. Even so, LME copper kept climbing through the auctions, from about US$4,100 per tonne before the first one to about US$4,400 per tonne after the fourth. The state ultimately stood behind Liu's contracts rather than default on them, absorbing a large loss instead.
The 2008 trial
The story does not end with the 2005 unwind. In March 2008, Liu Qibing and a supervisor went on trial before the Beijing No. 1 Intermediate People's Court. The court found that Liu's overseas copper positions between December 1999 and October 2005, a cumulative figure covering nearly six years, had produced losses of US$606 million. He was sentenced to seven years in prison. That US$606 million figure is a different number, over a different window, from the roughly US$150 million loss that contemporary press reporting attached to the 2005 short specifically. Both numbers are genuinely sourced; they answer different questions, and the two should be kept separate rather than collapsed into one headline.
| Date | Event |
|---|---|
| 1995 | Liu Qibing is sent to the LME for training by the SRCSR. |
| 2002–2004 | Liu runs a hedged SHFE–LME copper arbitrage, reportedly earning the Chinese state around US$300 million over the period. |
| Spring 2005 | Liu starts building an unhedged short position in LME copper, betting that lending curbs and planned SRB sales will push prices down. |
| Jul–Aug 2005 | The short position reaches about 130,000 tonnes for December 2005 delivery. |
| Early Sep 2005 | The short position reaches about 220,000 tonnes, average price about US$3,500/tonne, due mostly 21 December 2005. |
| Sep 2005 | China's National Audit Office publicly rebukes the SRB; the agency moves about 40,000 tonnes of copper toward the SHFE. |
| 14 Nov 2005 | Liu Qibing disappears from public view as copper keeps rising; SRB officials deny knowing him and deny the agency was short selling. |
| 16, 23, 30 Nov 2005 | The SRB holds domestic copper auctions, releasing stockpiled physical metal and revealing reserves of about 1.3 million tonnes, far above the market's estimate of about 250,000 tonnes. |
| Late Nov–Dec 2005 | LME copper rises from about US$4,100/tonne before the auctions to about US$4,400/tonne after the fourth; the state stands behind Liu's contracts rather than default. |
| Mar 2008 | Liu Qibing is tried before the Beijing No. 1 Intermediate People's Court, found to have run up US$606 million in cumulative overseas copper losses (Dec 1999–Oct 2005), and sentenced to seven years in prison. |
The mechanics, in course language
A short position in plain futures terms
Liu sold, or went short, standard 25-tonne LME copper lots for December 2005 and later delivery. A short futures position gains if the price falls below the contracted price and loses if the price rises above it. Liu was betting on a fall; the market delivered the opposite.
Why the loss demanded cash immediately
The position could not simply sit and wait for the bet to come right. Futures contracts are marked to market and settled daily against the current price, the same margin mechanism introduced under Lecture 2. As copper kept rising through 2005, the short side of every contract had to post more variation margin to its clearing broker each day, even though no cash had actually changed hands in the underlying trade. A paper loss on an open position becomes an immediate cash demand through the margining system, whether or not the position is later proven right. The same mechanic recurs in the GameStop-Robinhood-NSCC episode and in the 2022 LME nickel squeeze involving Tsingshan, both covered elsewhere in this course.
How the squeeze took hold
Once other traders suspected, correctly, that a single very large player stood behind the short and might not be able to deliver 220,000 tonnes of physical copper on the December delivery dates, they had an incentive to buy contracts against that position and hold them into delivery. That forces the short side either to buy back at ever higher prices or to source physical metal it does not have. A concentrated, poorly hidden short position meets a market that can identify it and press against it. This is the same structural failure mode as the 2022 nickel episode, seventeen years later, with a sovereign buyer standing behind the loser instead of a private company.
Why transparency rules exist
The episode is also a case for why exchanges care about position transparency and large-trader reporting. The market could infer the rough shape and size of the SRB-linked short from open interest and delivery patterns, which is exactly the information asymmetry that large-position reporting rules exist to reduce. The position was ultimately backed by a sovereign entity's real stockpile, some 1.3 million tonnes by the SRB's own November 2005 disclosure, so in the end it was absorbable. Absorbing it still cost the state dearly: it had to reveal stockpile information it would otherwise have kept confidential, and it still took a large realised loss.
Hedge and bet, same contract
Liu's two trades sit side by side as the real teaching point here. The 2002–2004 SHFE–LME arbitrage was a genuine cash-and-carry-style trade, built on the cost-of-carry and basis relationship between two related but geographically separate futures markets, which is Lecture 5 territory. The 2005 loss came from abandoning that hedged, basis-driven position for a large, unhedged directional bet in the same instrument. Same market, same contract, but a completely different risk profile once a spread becomes a naked short: a position is only as hedged as its actual structure, not its resemblance to a trade that used to be hedged.
The mathematics
No published source gives the exact realised loss on the 2005 position contract by contract, so what follows is an illustrative, order-of-magnitude mark-to-market calculation, using the sourced position size, the sourced average entry price, and the sourced price levels around the November 2005 auctions. It should be treated as an estimate, not a reconstruction of the actual trading book.
The mark-to-market loss on a short position is the size of the position multiplied by how far the price has risen above the entry price.
$$\text{MTM loss} = Q \times (P_{market} - P_{entry})$$
where \(Q\) is the position size in tonnes, \(P_{entry}\) is the average short (entry) price, and \(P_{market}\) is the current market price.
The sourced figures give \(Q = 220{,}000\) tonnes, \(P_{entry} = \$3{,}500\)/tonne, and the two price levels around the auctions.
$$\text{MTM loss (before auctions)} = 220{,}000 \times (4{,}100 - 3{,}500) = \$132\text{ million}$$
$$\text{MTM loss (after 4th auction)} = 220{,}000 \times (4{,}400 - 3{,}500) = \$198\text{ million}$$
position_tonnes = 220_000 # tonnes short
avg_short_price = 3500 # USD/tonne, average entry price
price_before_auctions = 4100 # USD/tonne, just before the Nov 2005 auctions
price_after_auctions = 4400 # USD/tonne, after the fourth auction
contract_size = 25 # tonnes per LME copper lot
mtm_loss_before = position_tonnes * (price_before_auctions - avg_short_price)
mtm_loss_after = position_tonnes * (price_after_auctions - avg_short_price)
num_contracts = position_tonnes / contract_size
print(mtm_loss_before)
print(mtm_loss_after)
print(num_contracts)Output
132000000
198000000
8800.0This range, US$132m–198m, brackets the widely reported "roughly US$150 million" contemporary loss estimate for the 2005 episode; the two are consistent. The court's US$606 million figure, quoted elsewhere, is not a competing estimate of the same thing: it covers cumulative overseas copper losses from December 1999 to October 2005, a much longer window and a different accounting exercise, so the two should not be averaged together. The 220,000-tonne position works out to about 8,800 standard LME copper contracts at the 25-tonne lot size (this page's own conversion, not a reported fact). The price needed only a small move to inflict serious damage: from US$3,500 to about US$4,182/tonne, a rise of just 19.5%, was enough to produce a US$150 million loss on this position size. That is the lesson in miniature: a fairly modest percentage move in the underlying, once a position is this large, produces a nine-figure loss.
Data and facts
| Quantity | Value | Source |
|---|---|---|
| Short position, Jul–Aug 2005 | ≈130,000 tonnes | Wikipedia, attributed to China Daily |
| Short position, early Sep 2005 | ≈220,000 tonnes (or 180,000 tonnes net of a 50,000-tonne buyback) | Wikipedia (China Daily); SCMP, 25 Nov 2005 |
| Average entry price of the short | ≈US$3,500/tonne | Wikipedia, attributed to China Daily |
| LME copper price before the Nov 2005 auctions | ≈US$4,100/tonne | Wikipedia timeline, period press incl. Washington Post |
| LME copper price after the fourth auction | ≈US$4,400/tonne | Wikipedia timeline |
| SRB stockpile auctioned, 16 & 23 Nov 2005 | ≈20,000 tonnes each date | Wikipedia timeline; CFR blog, 21 Nov 2005 |
| SRB declared stockpile vs market estimate | ≈1.3m tonnes vs ≈250,000 tonnes | Wikipedia timeline, period press |
| Date Liu disappeared from public view | 14 Nov 2005 | Washington Post, 25 Nov 2005 |
| Loss on the 2005 episode (contemporary press estimate) | ≈US$150 million | Wikipedia; cross-checked vs Washington Post |
| Cumulative overseas copper losses, Dec 1999–Oct 2005 (2008 court finding) | US$606 million | SCMP, 1 May 2008 |
| Trial outcome | Seven-year prison sentence | SCMP, 1 May 2008 |
| Copper price, start 2005 → 17 Apr 2006 (per the court's figures) | ≈US$3,000/tonne → ≈US$8,880/tonne | SCMP, 1 May 2008 |
| LME copper contract size | 25 tonnes/lot | LME contract specifications |
The lesson
- Margin turns a paper loss into an immediate cash problem. Every day copper rose against Liu's short, the position was marked to market and margin calls followed, the same mechanism behind GameStop/NSCC and LME nickel/Tsingshan elsewhere in this course, just run through a state balance sheet instead of a company's.
- Size and secrecy invite a squeeze. A short position large enough to be noticed, and opaque enough that the market could not verify whether the seller could really deliver, gave other traders a clear incentive to buy and hold into delivery. Concentrated, hidden positions are structurally exposed to being pressed.
- A hedge and a bet can use the same contract but carry opposite risks. Liu's earlier SHFE–LME arbitrage was genuinely hedged and basis-driven; the 2005 short was a directional bet dressed in the same instrument. Always ask whether a position is actually hedged, or whether it just looks like the trades that used to be hedged.
- A state guarantee can absorb a loss, but it cannot undo the market's information. Once the SRB stepped in to defend the position, it had to reveal the true scale of its physical stockpile, information it would otherwise have kept confidential, in order to make its domestic auctions credible.
- Concealment inside an organisation compounds trading losses. Later reporting around the 2008 trial describes forged documents and reluctance to escalate losses. When a losing position is hidden from supervisors instead of reported and closed, the eventual loss tends to be measured in multiples of what an early, honest stop would have cost.
Where it appears in the course
Think about it
- Liu's 2002–2004 SHFE-LME trade and his 2005 LME short both used the same contract. What made one a hedge and the other a naked bet, and what would you look for on a trading book to tell them apart?
- The SRB chose to defend Liu's position with real stockpile rather than let it default. What would have happened to the LME copper market's confidence in Chinese state trading if it had defaulted instead, and why might that have mattered more than the money?
- Large-trader position reporting exists partly to stop this kind of squeeze. If the LME had required Liu's short to be disclosed earlier in 2005, how do you think the rest of the market would have traded differently, and would that have made the squeeze better or worse for the SRB?
Sources
- Wikipedia contributors, "State Reserves Bureau copper scandal", Wikipedia (used here as a pointer to the underlying press sources it cites, principally China Daily and the Washington Post; facts are attributed to those primary sources, not to Wikipedia itself). en.wikipedia.org
- Peter S. Goodman, "New China Stumbles Into Old-Fashioned Trade Scandal", The Washington Post, 25 November 2005. washingtonpost.com
- South China Morning Post, "Copper trader a scapegoat, not rogue", 25 November 2005. scmp.com
- South China Morning Post, "Justice catches up with rogue copper trader", 1 May 2008 (reports the Beijing No. 1 Intermediate People's Court trial, the US$606 million cumulative loss figure, the seven-year sentence, and the 2005–2006 copper price trajectory used at trial). scmp.com
- Council on Foreign Relations (Brad Setser), "Rogue Copper Trader?", CFR blog, 21 November 2005 (contemporary analysis of the SRB's cross-market activity and the National Audit Office rebuke). cfr.org
- Fastmarkets, "SRB REVEALED: Shedding light on China's State Reserve Bureau in the copper market", 31 July 2015 (retrospective on the scandal's aftermath and the SRCSR's restructuring). fastmarkets.com
- Geoffrey Poitras, Commodity Risk Management: Theory and Application, Routledge, 2013 (academic treatment of the trial and sentencing outcome, cited here as corroboration).
