It gives too much authority to banks to bake their own numbers:
Willem Buiter: Basel II: back to the drawing board?: Two crucial inputs into Pillar 1 (Minimum Capital Requirements) of the proposed Basel II Framework for the International Convergence of Capital Measurement and Capital Standards have, if not gone belly-up, at least been severely compromised by the recent financial markets turmoil. They are the reliance on credit ratings provided by the internationally recognised rating agencies (currently Moody’s, Standard & Poor’s and Fitch) and the crucial role assigned to internal models in everything from stress-testing to marking-to-model illiquid assets.
It is clear that, as regard rating complex structured products, the three internationally recognised rating agencies have done a terrible job.... If rating were merely difficult, you would expect as many over-ratings as under-ratings. What we see instead, is a persistent bias: ratings seem to systematically over-estimate the creditworthiness of the rated instrument or structure. The reason for this must be the distorted incentive structure faced by the rating agencies. They are inherently and deeply conflicted.
First, almost unique in any appraisal process, the appraiser in the rating process is paid by the seller rather than the buyer. Second, the rating agencies provide (remunerated) technical assistance/advice on how to design structures that will attract the best possible rating to the very issuers whose structures they will subsequently rate. Third, rating agencies increasingly provide other financial services and products than ratings (or ratings advice). As with auditors, there is the risk that the rating (audit) service may be subverted in the pursuit of remunerative sales of these other products.
I am not asserting that the rating process of complex financial instrument is unavoidably utterly corrupt and useless... reputational considerations mitigate the conflict of interest.... The recent financial turmoil has led to a demystification of quants and other high-tech builders and maintainers of mathematical-statistical models and algorithms. We have had a powerful reminder of the ‘garbage in – garbage out’ theorem. On many occasions marking to model has turned out to be marking-to-make belief or marking-to-myth....
With so many illiquid, non-traded instruments on their books (and in off-balance-sheet vehicles that may have to be brought on balance sheet again soon), many banks are confronted with the fact that ‘fair value’, when it cannot be measured objectively by a market price, is unlikely to be calculated fairly by techie employees of the bank whose activities are not understood by the bank’s risk managers or top management, and whose pay and prospects depend in a pretty obvious way on the numbers their models crank out....
[B]anks [should] not be allowed to hold on their balance sheets, or to be exposed to through off-balance sheet connections, complex structures whose valuation cannot be verified easily by third parties. This is tough and will be unpopular with the industry, but necessary for financial stability.... [C]omplex products tend to be motivated by regulatory avoidance and tax avoidance considerations, and should be discouraged by regulatory design. True risk trading and risk sharing require simple, transparent instruments, designed for specific contingencies (states of nature), rather like elementary Arrow-Debreu securities. They don’t require convoluted bundles of heterogeneous opaque contingent claims.