Dear allI have come across an article by Christian Bluhm on CDO modelling, in which he asserts two things I find difficult to understand.Firstly, he constructs credit spead curves from observed default and rating transition probabilities. He then suspiciously often conditions his statements on the reader's belief in the correctness of the spreads. Well I can't make myself believe in it, since real probabilities should be adjusted by the market price of risk to be used in pricing.Secondly, the author finds it obvious that the sum of single-name CDS spreads received by a CDO originator is larger than the sum of spreads paid on CDO tranches, because of diversification. Now, usually assembling portfolios is not an opportunity for arbitrage, rather, holding single assets is suboptimal.Any comments and remarks will be greatly appreciated. The article can be found on www.defaultrisk.com.Best regards - Dunbar

problem with #1 is that market pricing is risk-neutral, and any basis between the current market pricing and backwards-looking statistics is not worth much... it's the same thing as recognizing that the IR markets may not obey the PEH and still not being able to arb it efficiently.thing about #2 that is not always appreciated is that most of this debt is coming from banks that already price with more diversification than any CDO can provide (not to mention a great freedom from investment and trading parameters generally limited by the CDO indenture). So it's not clear that this pricing theory allows for cost-of-capital arb by itself. A bank's cost of capital is pretty low to start with - so it's also not obvious that securing debt by specific assets (i.e. not junking up the cost of capital with the risk of rogue gamblers on the FX desk) would be an improvement.happened to be perusing graham+dodd '34 in the bookstore this weekend, and found the sentiments in the introduction apropos: Historical risk/return is not the optimizing parameter for investors, and even when asset classes appear through [recent] history to be poor investments in terms of risk/return, there will still be buyers. At the time they were speaking of corporate debt, where upside was limited as it always is with bonds, and yet downside was not all that safe given the collapse of the market bubble. Familiar, no?

Perhaps the idea is that with a downward sloping correlation smile (higher on equity and lower on senior) the originator will indeed lock in a positive carry by selling the complete capital structure. This arbitrage only exists if the originator can underpay to be short portfolio loss risk across the whole cap structure - definitely not the case currently.

downward sloping correlation smile = irrational marketdepends on the copula of course, but it runs contrary to the expected economics... in that case I'd be trading tranches with external leverage