Measures of predictive success for rating functions

Sebastian Ostrowski, Peter Reichling

Abstract

ABSTRACT

This paper aims to develop an adequate measure of predictive success and accuracy of rating functions. First, we show that the common measures of rating accuracy, ie, area under curve and accuracy ratio, lack informative value for single rating classes. A 1991 paper by Selten built up an axiomatic framework for measures of predictive success. We introduce a measure for rating functions that fulfills the axioms proposed by Selten. Furthermore, we carry out an empirical investigation to analyze the predictive power and accuracy of Standard & Poor's and Moody's ratings, and compare the rankings according to the area under curve and our measure.