One of the fundamental breakdowns of this credit crisis was in the ratings agencies (technically, the Nationally Recognized Statistical Rating Organizations - NRSROs).
This article over on the Financial Times gives a nice overview of the growth of Moody's and the NRSROs, especially in the structured finance arena. You should check it out, twice if you're not really a follower of the market.
The big takeaway from this whole crisis, on the NRSRO side, is that the appropriate compensation model for the NRSROs is for their customers to be the buyers of bonds and other credit products. Additionally, they should probably be private versus public organizations, much like SC Johnson. Many firms are public that have no good reason to be. I am all for public markets, but when a firm serves such a large public role, being publicly traded is probably a handicap as opposed to a benefit. Adding a touch of the Google internal stock market is probably a good idea for these private firms, but their shares do not need to trade on public markets. I really don't get this desire for firms to be public, except if they are struggling. Maybe all those LBO targets are on to something? (About being private, anyway. It sure isn't their business processes.)
As for the point of credit buyers paying the ratings firms, this is ideal because they are the ones who ultimately will benefit from the research. The current structure invites severe moral hazard. The FT articles mentions that the complexity of rated products forced growth in the size of Moody's in order to rate these large entities and the esoteric products coming into the credit market. Well, maybe scale was required, but the flip side of the subscription model is that if a corporation will not allow the NRSRO to rate its product -- if the corporation willfully decreases transparency into its credit quality -- then a suitable discount must be applied in the market to compensate for the lack of information. This is the ONLY model that makes sense. Without the information, the opacity should force the market to discount the bonds, commercial paper, convertibles, or structured product being sold (or underwritten) by the corporation. That's not fair -- its just plain good sense and good business!
The final concern becomes the models themselves. As has been widely noted, the NY Times covered this recently. The models are just mathematical equations. They have no position on this subject; they are merely tools, and like any tool, they can be misused or misunderstood by those wielding them. Human frailty (and that damn good sense) come into play here again. The Times piece looks at how quickly humans were willing to set aside their reason in the pursuit of compensation, or assume that the real world was accurately reflected by the models.
The models were not then, and are not now, the problem. The problem is the gatekeepers wielding the models. Once a flaw - a programming bug, or worse, a design flaw - was found, anything rated with that software should have been re-rated. The decision to NOT re-rate those issues should be considered an act of negligence, possibly willful and criminal negligence. Models will evolve if their creators update them. Secondly, if any product has little to no history, then you cannot reasonably rate it based on historical performance of any other product.
So if you're out there in the credit market in any way, you owe it to yourself to do SOME credit analysis on any product you might buy. Whether we're talking bond funds, or direct investment in credit products, you have to understand what you're buying on some level. Credit analysis is more difficult than equity analysis. Such is life.
So to wrap this (very late) missive up, the ratings agencies can serve a useful function. However, their role of being public watchdogs, a la the press, must be honored. Otherwise, that role is better subsumed within a unified regulator (whenever the SEC and CFTC merge), and let's face it - we don't want that. This is something that is probably better suited to government functionally than most other tasks it takes on, but it still should be a market function. Government could provide credit research and publish it, for free even, but would that research be very good? Could it be subverted? A market system for credit research will go a longer way to promoting the transparency and lack of conflict that investors deserve. Adding that value should come with a price.