Famous Derivatives Disasters: When Quant Finance Goes Wrong

All the math in the world does not help if the people using it are reckless, clueless, or dishonest. Chapter 44 is Wilmott’s tour through the greatest hits of derivatives disasters. These are not abstract case studies. Real people lost real billions, institutions collapsed, and careers ended. Some of these stories are tragic, some are farcical, and a few are both.

Orange County: The Treasurer Who Bet the Farm

Orange County, California. Early 1990s. The County Treasurer was a man named Robert Citron. He managed the county investment fund, a pool of taxpayer money. From 1991 to early 1994, he made about $750 million. That is 400 basis points above government rates. Very impressive returns.

How? Leveraged inverse floating rate notes. The coupon on these things goes UP when interest rates go DOWN. While rates were low in the early 90s, Citron was collecting fat coupons. The formula was something like $\max(\alpha r_f - \beta r_L, 0)$ where $\beta > 1$, meaning the exposure was leveraged. Rates go up a little, your coupon drops a lot.

Citron bet that rates would stay low. They did not. In mid-1994, US rates rose by about 3%. The leveraging kicked in hard. Orange County lost $1.6 billion.

The county did not actually need to declare bankruptcy. They were still solvent and money was still flowing in. But they declared bankruptcy anyway on December 6, 1994, as a legal tactic in their lawsuit against the brokers. At the time of filing, S&P rated Orange County AA and Moody’s rated them Aa1. Very high ratings.

The instrument itself is not complicated. It is just $\max(\alpha r_f - \beta r_L, 0)$. The risks should be obvious. But during sentencing, psychologists found that Citron had “the math skills of a seventh grader” and was “in the lowest 5% of the population in terms of ability to think and reason.” He was sentenced to one year of community service.

And who won? The counterparties selling the floaters were US government housing agencies. So taxpayer money flowed from Orange County taxpayers to US taxpayers everywhere. Wilmott finds this amusing.

Proctor and Gamble: The Swap That Went Sideways

P&G is a giant multinational. They have big exposure to interest rates. In late 1993 they wanted a swap from fixed to floating, betting that rates (then low) would stay low. Their counterparty was Bankers Trust (BT).

The deal was a five-year swap on $200 million notional. For the first six months, P&G pays a fixed rate. After that, the rate was defined by a formula involving the five-year Treasury yield and the 30-year Treasury bond price. The formula was locked in on May 2, 1994, based on rates on that date.

Best case for P&G: rates stay near November 1993 levels for a few months, and they save about $7.5 million. Not a huge amount relative to their size.

But if rates rose between November and May, the formula worked against them. And rise they did. By May 1994, five-year and 30-year rates had gone up by over 150 basis points on average. P&G lost close to $200 million.

Here is the devastating part. Wilmott did the math on a spreadsheet. Using 10 years of historical data available at the time the deal was signed, there was a 14% chance of rates rising enough for P&G to start losing money. The expected profit over the life of the deal was negative $8.7 million, not positive $7.5 million. Ten minutes on a spreadsheet would have saved $200 million.

P&G sued BT for failing to disclose information. They settled out of court, with BT absorbing about 83% of the losses.

Metallgesellschaft: Hedged on Paper, Broke in Practice

Metallgesellschaft is a German conglomerate. Its US subsidiary MGRM sold 10-year forward contracts locking in the price of oil and gasoline. The total volume was 180 million barrels, equivalent to 85 days of Kuwait’s output.

MGRM hedged this long-term short position with short-term futures on NYMEX, rolling them over every few months. In theory, this should work if you have a good model for interest rates and the cost of carry.

Then oil prices fell during late 1993. MGRM was long the futures, so a price drop meant margin calls. Daily margin calls. The extra margin requirements hit $900 million during 1993.

On paper, MGRM was fine. The long-term forward contracts were gaining value as oil prices fell. But those were forwards, not futures. The gains were not realized until delivery. So MGRM was losing real cash on the futures side while only gaining theoretical value on the forward side.

The parent company panicked, installed new management, and started closing positions. Then oil prices rose in early 1994, so the remaining long-term contracts started losing. More positions were closed. Total losses: $1.3 billion.

Were they actually losses? If MGRM had waited, the positions might have worked out. Marking to model, the net position should have been roughly flat regardless of oil price. Think margin hedging from CrashMetrics, covered in the previous chapter. MGRM was delta hedged but not margin hedged.

There was also basis risk. The oil market shifted from backwardation (downward-sloping forward curve, profitable for rollovers) to contango (upward-sloping, costly for rollovers), adding $20 million per month in rollover losses.

Gibson Greetings: When Your Bank Lies to You

Gibson Greetings makes greeting cards. They issued $50 million in bonds at 9.33%. When rates fell, they wanted to reduce their debt cost and entered vanilla swaps with Bankers Trust in 1991.

Then they got sucked into increasingly complex products. About 29 contracts total. Ratio swaps, periodic floors, Treasury-linked swaps, knock-out options, corridor swaps. Each one more exotic than the last.

The worst part was not the complexity. It was the dishonesty. BT had internal models for valuing Gibson’s positions. They knew the real numbers. But they consistently understated the losses to Gibson. A taped conversation between BT employees is damning: “We told him $8.1 million when the real number was 14. So now if the real number is 16, we’ll tell him that it is 11. You know, just slowly chip away at the differential.”

Gibson lost over $20 million, almost a year’s profits. They sued BT and settled out of court.

Barings: The Rogue Trader

The Barings story is simple. Nick Leeson, a 28-year-old trader at Barings Futures in Singapore, had control over both the trading desk and back office operations. He was policing his own activities. That is a terrible idea.

In late 1994, Leeson sold straddles on the Nikkei 225 with strikes around 18500-20000. Almost delta neutral. As long as the index stayed stable and volatility stayed low, he profited. Then on January 17, 1995, the Kobe earthquake hit. The Nikkei started falling.

Instead of cutting losses, Leeson doubled down. He started buying index futures, believing that the sheer size of his trades could reverse the market decline. They could not. The Nikkei fell to 17400. As it kept falling, he kept buying. Margin calls became massive.

On February 23, 1995, Leeson fled Singapore. He was eventually arrested in Frankfurt. Losses totaled $1.3 billion. Barings, a 200-year-old bank, went bust. ING bought the corpse for one pound.

Leeson was sentenced to six and a half years. While in prison, he developed colon cancer, his wife divorced him, and they made a movie about it starring Ewan McGregor.

LTCM: When Nobel Laureates Blow Up

Long-Term Capital Management was a hedge fund founded by John Meriwether (ex-Salomon bond arbitrage) with Nobel laureates Myron Scholes and Robert Merton as partners. Star power on steroids.

Edward Thorp, who started one of the first hedge funds in 1969, was invited to invest. He declined: “I didn’t want to have anything to do with it because I knew these guys were just dice rollers.”

LTCM leveraged $4.8 billion into $100 billion in positions. Their notional swap position reached $1.25 trillion, which was 5% of the entire market. They were excused collateral on many deals. And when they failed, the Federal Reserve organized a bailout. Good connections matter.

Their strategies all had the same underlying bet: the world was going to be calm, stable, and converging. They sold German bonds and bought other European bonds (expecting convergence). They went long Brazilian and Argentine bonds while shorting US Treasuries (expecting credit spreads to narrow). They bought Russian GKOs and sold Japanese bonds. They bet on the German yield curve flattening. They bought long-dated swaption straddles and sold short-dated ones.

Then Russia defaulted on August 17, 1998. Markets panicked. Investors fled to quality. Every single one of LTCM’s trades went wrong simultaneously. They were not diversified at all. They had one massive bet on market stability, expressed in five different ways.

On August 21, 1998, they lost $550 million in a single day. They had estimated daily swings should be around $45 million. Their VaR models did not account for extreme moves or liquidity problems.

The liquidity issue was critical. Many banks knew LTCM’s positions and were copying them. When everyone tries to sell the same thing at once, there are no buyers. The concept of “value” becomes meaningless. You just wait until the market loosens up. If you can afford to wait.

In September 1998, the New York Fed organized a $3.6 billion bailout by 14 banks, who took a 90% stake in LTCM. The partners lost 90% of their investment.

Wilmott estimates that LTCM was overbetting to about twice the Kelly level, which is the point where expected growth rate becomes negative.

The Common Thread

All these stories share the same themes: overconfidence, lack of understanding of risks, speculation at inappropriate times, and excessive leverage.

Wilmott ends the chapter with a quote from Robert Merton in 1993, five years before his own fund blew up: “The mathematics of the models are precise, but the models are not, being only approximations to the complex, real world. The practitioner should therefore apply the models only tentatively, assessing their limitations carefully in each application.”

As Wilmott says: Doh!

Previous: CrashMetrics: Preparing Your Portfolio for the Worst

Next: Financial Modeling: A Warning About Models in Practice