The Mystery of Merrill's Asset Reduction

Alea is as eagle-eyed as ever:

The most important and most ignored sentence in Merrill's press release:

The sale will reduce Merrill Lynch's risk-weighted assets by approximately $29 billion.

How is this possible? After all, last month the CDOs being sold were valued at $11.1 billion, and indeed "Merrill Lynch's aggregate U.S. super senior ABS CDO long exposure" has been reduced by precisely that amount.

A CDO is not, in itself, a leveraged instrument: you can't lose more than you invested in it. If you're carrying a CDO on your books at $11.1 billion, there's no way that CDO can be responsible for $29 billion in losses, since its value can't fall below zero.

But there's more: since Merrill is lending Lone Star $5 billion to buy the CDOs, and the loan is non-recourse to Lone Star itself, the maximum that Lone Star can lose is its equity investment of $1.7 billion. If the CDOs went to zero tomorrow, then Merrill would suffer a further $5 billion of losses, and that contingent exposure should appear on the bank's balance sheet somewhere, as the risk associated with the loan to Loan Star.

Maybe that's where the $29 billion number comes from? Up until the sale, Merrill owned the CDOs outright, and if the CDOs all performed then those assets would have more or less their face value of $30.6 billion. Merrill took mark-to-model losses of $19.5 billion on those assets, but it still owned them, and kept them on its balance sheet under risk-weighted assets, valued at face.

Now, however, the CDOs have now been moved off Merrill's balance sheet, and Lone Star gets all of the upside (minus the interest it's paying on its $5 billion loan, of course). The $1.6 billion difference between $30.6 billion and $29 billion would then be Merrill's risk-weighted exposure to the final $5 billion of possible losses: after all, the $5 billion loan to Lone Star is surely now a Merrill Lynch risk-weighted asset of some description.

Let's say that Merrill considered the CDOs to be $30.6 billion of risk-weighted assets before the sale, and $0 afterwards. Then risk-weighted assets will have fallen from $30.6 billion in CDOs, to $5 billion in loans to Lone Star: a drop of $25.6 billion, not $29 billion.

So the only way I can see that Merrill could have got to $29 billion would be by risk-weighting the loan to Lone Star at just 32%: the $5 billion loan, once risk-weighted, would be counted as entailing risk of just $1.6 billion.

But even that doesn't make sense. Before the sale, Merrill would have been counting 100% of the CDO assets as risk assets; after the sale, Merrill would be determining that $3.4 billion of the assets are essentially risk-free. It seems to me there's some kind of weird remarking going on; if anybody could help me out, I'd be much obliged.

Oh, and one other thing from the press release: there was some commentary yesterday saying that the sold-off CDOs were mostly of 2005-and-older vintage. That's not true: it's Merrill's remaining $8.8 billion of CDO exposure which is the older stuff.