Smooth estimation of rating transitions by nearest neighbors

Abstract

Many types of duration data suffer both from left-truncation and right-censoring. We show how these deficiencies can be overcome at the same time when estimating the hazard rate nonparametrically by kernel smoothing with the nearest neighbor method. We infer the uniform consistency of the estimate from the Hoeffding inequality, applied to a generalized empirical distribution function. Finally, we apply our estimator to rating transitions of corporate loans in Germany. JEL numbers; C14, C13, C24, C41, G24, G12