"The Kamakura reduced form and hybrid default probability models have reached unprecedented accuracy in version 4.0," said Warren Sherman, Kamakura President and Chief Operating Officer. "The Kamakura Jarrow-Chava model captured 41.70% of total defaults in the top 1% of KRIS default probabilities. The hybrid model captured 40.45% of total defaults in the top 1% of KRIS default probabilities. This accuracy level is far in excess of that possible under modern Merton model technology, which is why many vendors are forced to cap their default probabilities at a low level. Kamakura clients believe that these kinds of ceilings on default probability levels obscure model performance, and we agree with their assessment. Both the Jarrow-Chava and the hybrid model, with ROC accuracy ratios in excess of 95% on the public company universe, outperform modern Merton model technology by more than 10%. The Jarrow-Chava model, for example, explains more than 85% of five year variation in defaults in out of sample tests."