Cantor, Richard, Frank Packer, and Kevin Cole, "Split Ratings and the Pricing of Credit Risk", Journal of Fixed Income, Vol. 7, No. 3, (December 1997), pp. 72-82.
Abstract: Despite the fact that over fifty percent of all corporate bonds have different ratings from Moody's and Standard and Poor's at issuance, most bond pricing models ignore these differences of opinion. Our work compares a number of different methods of accounting for split ratings in estimating bond pricing models. We find that pricing rules that use only the Moody's or Standard and Poor's ratings produce unbiased but highly inefficient forecasts. If models rely instead on simply the higher or lower of the two ratings (but not both), greater bias is introduced with insignificant gains in efficiency. In general, the average rating is the best guide to predicting yields in terms of both bias and forecast prediction. However, the forecasting advantage from using the average rating rather than the lower rating derives almost entirely from the below-investment-grade subsample.