One of the big unanswered questions surrounding the fourth-quarter collapse of Merrill Lynch is how, exactly, it contrived to lose $15 billion in December. The general suspicion in the blogosphere is that it was all a function of marking to market loans which had been kept on the books at artificially high levels. But according to the WSJ, a large part of it might simply have been a trading-desk blowup of greater-than-Kerviel proportions:
Behind some of the losses in the quarter are two related trades that Merrill hasn’t discussed publicly in detail.
Broadly, both trades are set up to generate returns from corporate bonds while hedging the exposure to the debt through derivatives using credit-default swaps. Those derivatives provide protection against defaults on the bonds.
Merrill, according to a person familiar with the situation, ran two versions of the trade. One was a plain-vanilla strategy while the other was a more complex version. According to this person, Merrill was one of the biggest traders in the complex trade among U.S. firms. European banks made similar trades.
The idea is that the two sides of the trades — either the plain-vanilla version or the complex bet — are supposed to move in tandem. For both trades, things went awry in the fourth quarter when bank-lending markets froze. That ultimately triggered a sharp drop in bond prices. The value of default insurance rose, but not enough to cover the drop in the bond prices.
Is it even possible to lose billions of dollars on a CDS basis trade? I look forward to hearing more about this, but if the basis gapped out by say 100bp, then you’d need to own $100 billion of bonds just to lose $1 billion on the trade.
Is there some kind of dataset anywhere of the CDS basis on corporate bonds? I’d love to have a look at what happened to the basis in the fourth quarter of last year, to see if there was any large spike which might explain what happened at Merrill. The WSJ report is a tantalizing taster, now I want to get my teeth into this story!
Update: Hemant, in the comments, makes an interesting point about duration: if you’re hedging a five-year bond with a five-year CDS, and the bond has a modified duration of say three years, then a 300bp move in the CDS basis — which can happen — could result, he says, in a loss of 9% of the par value of the bond. Which means that a billion-dollar loss would require the desk to be holding "only" $11 billion in bonds.