Due to the five year gap in 30 year bond issuance, in mid-September the CME revised the deliverable basket for the June 2015 T-bond contract. It deleted the 6.25 of May, 2015 because its delivery value would have been so far below the values of the other bonds in the deliverable set. This would have made the contract more susceptible to a squeeze because only that bond would effectively be available for delivery due to the way the contract works.
It looks like a Treasury futures trader failed to do his or her homework.
The price of 30-year Treasury futures expiring in June traded for less than 145 for about two hours yesterday before shooting up to more than 150. The 7.3 percent surge in their price yesterday, on the first day these particular contracts were traded, was unprecedented for 30-year Treasury futures, according to data compiled by Bloomberg. Volume amounted to 1,639 contracts with a notional value of $164 million.
What sets these futures apart from others is they’re the first ones where the U.S. government’s decision to stop issuing 30-year bonds from 2001 to 2006 must be accounted for when valuing the derivatives. The size and speed of yesterday’s jump indicates the initial traders of the contracts hadn’t factored in the unusual rules governing these particular products, said Craig Pirrong, a finance professor at the University of Houston.
“That is humongous,” said Pirrong, referring to the 7.3 percent jump. “We’re talking about a move you might see over weeks or a month occur in a day.” Pirrong said he suspected it was an algorithmic trader using an outdated model. “I would not want to be that guy,” the professor said.
Here’s a quick and dirty explanation. Multiple bonds are eligible for delivery. Since they have different coupons and maturities, their prices can differ substantially. For instance, the 3.5 percent of Feb 39 sells for about $110, and the 6.25 of 2030 sells for about $146. If both bonds were deliverable at par, no one would ever deliver the 6.25 of 2030, and it would not contribute in any real way to deliverable supply. Therefore, the CME effectively handicaps the delivery race by assigning conversion factors to each bond. The conversion factor is essentially the bond’s price on the delivery date assuming it yields 6 percent to maturity. If all bonds yield 6 percent, their delivery values would be equal, and the deliverable supply would be the total amount of deliverable bonds outstanding. This would make it almost impossible to squeeze the market.
Since bonds differ in duration, and actual yields differ from 6 percent, the conversion factors narrow but do not eliminate disparities in the delivery values of bonds. One bond will be cheapest-to-deliver. Roughly speaking, the CTD bond will be the one with the lowest ratio of price to conversion factor.
That’s where the problem comes in. For the June contract, if the 6.25 of 2030 was eligible to deliver, its converted price would be around $142. Due to the issuance/maturity gap, the converted prices of all the other bonds is substantially higher, ranging between $154 and $159.
This is due a duration effect. When yields are below 6 percent, and they are now way below, at less than 3 percent, low duration bonds become CTD: the prices of low duration bonds rise less (in percentage terms) for a given decline in yields than the prices of high duration bonds, so they become relatively cheaper. The 6.25 of 2030 has a substantially lower duration than the other bonds in the deliverable basket because of its lower maturity (more than 5 years) and higher coupon. So it would have been cheapest to deliver by a huge margin had CME allowed it to remain in the basket. This would have shrunk the deliverable supply to the amount outstanding of that bond, making a squeeze more likely, and more profitable. (And squeezes in Treasuries do occur. They were rife in the mid-to-late-80s, and there was a squeeze of the Ten Year in June of 2005. The 2005 squeeze, which was pretty gross, occurred when there was less than a $1 difference in delivery values between the CTD and the next-cheapest. The squeezer distorted prices by about 15/32s.)
The futures contract prices the CTD bond. So if someone-or someone’s algo-believed that the 6.25 of 2030 was in the deliverable basket, they would have calculated the no-arb price as being around $142. But that bond isn’t in the basket, so the no-arb value of the contract is above $150. Apparently the guy* who didn’t get the memo merrily offered the June future at $142 in the mistaken belief that was near fair value.
After selling quite a few contracts, the memo non-reader wised up, and the price jumped up to over $150, which reflected the real deliverable basket, not the imaginary one.
This price move was “humongous” given that implied vol is around 6 percent. That’s an annualized number, meaning that the move on a single day was more than a one-sigma annual move. I was being very cautious by saying this magnitude move would be expected to occur over weeks or months. But that’s what happens when the reporter catches me in the gym rather than at my computer.
This wasn’t a fat-finger error. This was a fat-head error. It cost somebody a good deal of money, and made some others very happy.
So word up, traders (and programmers): always read the memos from your friendly local exchange.
*Or gal, as Mary Childs pointed out on Twitter.