The Trade That Grew Bigger Than Its Own Market

In December 2011 JPMorgan Chase told its Chief Investment Office to reduce risk-weighted assets. The CIO managed roughly $350 billion of the bank's surplus deposits, and inside it sat something called the Synthetic Credit Portfolio — a book of credit derivatives originally intended as a hedge against a corporate credit blow-up.

There were two ways to obey the instruction. Sell the book, or buy offsetting positions until the net risk measured smaller. The desk chose the second. Twelve months later the loss was at least $6.2 billion, five regulators had fined the bank roughly $1 billion, and the chief executive had had his pay halved.

What the position actually was

The book traded standardised credit default swap indices and the tranches built on them. Through the first quarter of 2012 the desk went long credit risk — selling protection — principally on an index called CDX.NA.IG.9, the ten-year investment-grade series, while simultaneously buying protection on high-yield indices. In plain terms: it was collecting premium for insuring investment-grade corporate debt against default, and paying premium to insure junk debt.

The theory was a relative-value trade on the relationship between those instruments — the basis between an index and the single names inside it, and between investment grade and high yield. The theory was not absurd. The execution had a fatal property: to make the reported risk number go down, the desk had to make the actual position go up. The Senate investigation put it plainly: by taking the portfolio net long, the strategy “eliminated the hedging protections the SCP was originally supposed to provide.”

Roughly $40 billion of long notional was added in March 2012 alone — after the book was already losing money.

The position ate its own market

This is the part that transfers directly to a retail account, only with different zeros. On 29 February 2012 the desk sold $7.17 billion of the IG9 ten-year index net — its biggest day ever, with $4.6 billion of it inside a three-hour window — to defend a position that was already moving against it.

The regulator's findings on what that meant, all of them admitted by the bank:

• Those sales were more than 90% of the entire market's net volume that day.
• They were 15% of the market's net volume for the whole month.
• They were roughly eleven times the desk's own average daily volume that February.
• Over 27–29 February the desk's net volume was about one third of everything every other participant traded in that index in the entire month.

By late February the net position in that single index had reached $65 billion. When you are ninety percent of the volume, you are not trading a market price any more. You are printing it, and the print is a lie you will have to buy back later at whatever the truth turns out to cost.

JPMorgan's own people described the book as of “a perilous size,” because a small adverse move would translate into an enormous loss. And the structural failure behind it is almost comically simple: the bank had no limit on the notional size of the position. Concentration limits that were standard in JPMorgan's investment bank, for the very same instruments, were simply absent in the CIO.

Making the risk number smaller instead of the risk

On 16–19 January 2012 the desk breached its own value-at-risk limit and the bank-wide VaR limit for four consecutive days. On 27 January, mid-breach, a new VaR model went live. Reported risk fell by roughly half overnight — the bank's own internal email put the reduction at 44%, the Senate report at 50% — which ended the breach and freed the desk to trade more.

The new model was later found to have been improperly implemented, dependent on error-prone manual data entry, and to contain formula and calculation errors. On 10 May 2012 the bank revoked it and reinstated the old one. It had bought the desk three and a half months.

The wider alarm record is the detail worth keeping: between 1 January and 30 April 2012, CIO risk limits and advisories were breached more than 330 times. One credit-spread limit was exceeded by 100% in January, 270% in early February, and by more than 1,000% in mid-April. In February an internal risk measure warned the portfolio could lose $6.3 billion in a year. It was dismissed internally as “garbage.” It turned out to be the most accurate number anyone produced.

Marking the book to hope

A derivatives position must be valued somewhere inside the bid-ask spread, at the point most representative of fair value. Early in January the desk marked at or near the mid. From late January it began marking at the most favourable end of the daily range, and in some March cases outside every dealer quote it had received that day.

The junior trader keeping the parallel spreadsheet of real values wrote to his supervisor: “I am not marking at mids as per a previous conversation.” The next day the senior trader wrote: “I can't keep this going … I don't know where he wants to stop, but it's getting idiotic.”

Reality arrived through counterparties, who valued the same positions $500 million lower and issued collateral calls; the peak dispute reached $690 million. On 13 July JPMorgan restated its first-quarter results, cutting pre-tax income by $660 million ($459 million after tax), on the finding that the London marks had not been good-faith estimates of fair value.

The timeline of denial

The $5.8bn and $6.2bn figures are not in conflict: one is the loss through 30 June, the other through 31 December.

What it cost, and who paid

On 19 September 2013 four regulators settled simultaneously: the SEC ($200 million, with the bank admitting the facts and that it violated federal securities law), the Office of the Comptroller of the Currency ($300 million), the Federal Reserve ($200 million) and the UK's Financial Conduct Authority (£137.61 million). A month later the CFTC added $100 million in the first case brought under the Dodd-Frank anti-manipulation provision, with the bank admitting its traders had acted recklessly. Total: roughly $1.02 billion.

Two London traders were indicted in September 2013. Neither was ever tried: extradition from Spain was refused in 2015, extradition from France was judged futile, and in July 2017 the Department of Justice moved to dismiss all charges, stating that it could no longer rely on the testimony of its intended witness. The civil case was dismissed with prejudice a month later. Nobody was convicted. The trader the press had nicknamed the London Whale was never charged at all.

The four mistakes, at any account size

Nothing in this story requires a balance sheet. Every failure here is available to a trader with a five-figure account and a futures platform.

1. Hedging by adding, not by reducing. The book was told to carry less risk and responded by opening more positions. Retail version: a losing long “hedged” with a short in a correlated instrument, producing two positions, two sets of costs, and a net exposure nobody can now describe in a sentence.

2. Defending a position with size. Adding into a loser feels like conviction and is arithmetically indistinguishable from doubling risk at the worst available price. The desk's biggest ever trading day was a defence, not an opportunity.

3. Size that exceeds the liquidity you will need to exit. If your position is a meaningful share of what trades, your exit is your own worst counterparty. Test it the boring way: what fraction of the average volume at your typical exit time is your position?

4. Marking to the number you want. A retail trader has no bid-ask spread to abuse, but has something better: an unrealised loss, which can be quietly excluded from “how I'm doing” until it is realised. A journal that records open positions at the mark, every day, is the small-account version of the control that JPMorgan did not have.