I very much enjoyed reading this paper

I very much enjoyed reading this paper. The basic message that I took away was that by and large different banks rate a given borrower in pretty much the same way. While we do see differences, on the whole at least to my eye these differences are generally small. I therefore read the paper as basically good news for the New Basel Capital Accord. If we want to worry about any of the results in this paper we might be concerned that when differences exist they tend to be for the poorer quality credits where a difference of one ratings grade has a bigger effect on capital. But all in all there is not as much to worry about here as I thought there might have been. I have no substantive criticisms of the paper. The dataset is innovative. Mark is well aware of its limitations and these are discussed nicely in the paper. And the interpretation of the results is more than reasonable. Therefore, in my comments, what I would like to do is to focus on two issues that mean that while Mark's paper should give us considerable comfort, we should not yet sleep perfectly at night! The first issue is the existence of evidence that could be seen as conflicting. And the second is the question of how ratings systems can be verified by supervisory authorities. Conflicting evidence I would like to show two pieces of evidence that could be seen as conflicting. The first is from Jordan, Peek and Rosengren (2002). These authors track the ratings of a portfolio of syndicated loans in the US and then calculate what the capital requirement would have been under the proposed Foundation IRB approach (using the January 2001 proposals). They do this is two ways. The first assumes that a bank holding this portfolio of loans uses external credit ratings (from Standard and Poor's)