A second view is that because the methodologies used for rating CDOs are complex, arbitrary, and opaque, they create opportunities for parties to create a ratings “arbitrage” opportunity without adding any actual value. It is difficult to test this view, too, although there are reasons to find it persuasive. Essentially, the argument is that once the rating agencies fix a given set of formulas and variables for rating CDOs, financial market participants will be able to find a set of fixed income assets that, when run through the relevant models, generate a CDO whose tranches are more valuable than the underlying assets. Such a result might be due to errors in rating the assets themselves (that is, the assets are cheap relative to their ratings), errors in calculating the relationship between those assets and the tranche payouts (that is, the correlation and expected payout of the assets appear to be higher and therefore support higher ratings of tranches), or errors in rating the individual CDO tranches (that is, the tranches receive a higher rating than they deserve, given the ratings of the underlying assets). These arguments are complex and subtle…
That is from a very interesting paper by Frank Partnoy. The paper is not always easy reading but so far it is the best piece on its topic I have found. This was another good section:
If the mathematical models have serious limitations, how could they support a $5 trillion market? Some experts have suggested that CDO structurers manipulate models and the underlying portfolio in order to generate the most attractive ratings profile for a CDO. For example, parties included the bonds of General Motors and Ford in CDOs before they were downgraded because they were cheap relative to their (then high) ratings.67 The primary reason that the downgrades of those companies had an unexpectedly large market impact was that they were held by so many CDOs.
Thus, with respect to structured finance, credit rating agencies have been functioning more like “gate openers” rather than gatekeepers. The agencies are engaged in a business, the rating of CDOs, which is radically different from the core business of other gatekeepers. No other gatekeeper has created a dysfunctional multi-trillion dollar market, built on its own errors and limitations.
There is also a good discussion of how the ratings agencies have claimed First Amendment protection for their activities, more or less successfully. p.96 offers some good policy conclusions.