Credit Risk and Corporate Finance

This dissertation consists of three parts in the area of empirical credit risk and corporate finance. The first part evaluates two alternative options for dealing with a high number of potential predictors of corporate default. Credit risk modeling typically involves some form of forward or backward selection of variables. Bayesian model averaging has been proposed to deal with resulting biases and overconfident standard errors as well as to improve upon predictive ability. Using bootstrap analysis, it is shown that Bayesian model averaging indeed yields better out-of-sample predictions. A transformation of the explanatory variables largely eliminates this advantage, however. The success of Bayesian model averaging therefore seems to stem mainly from its role as partial remedy against model misspecification. The second part studies whether reducing a debt overhang improves incentives and thus performance. It provides empirical evidence supporting this argument made in the development economics and corporate finance literature, using a sample of distressed and highly overleveraged Austrian ski hotels undergoing debt restructurings. The vast majority of the hotels experienced substantial debt forgiveness, resulting in significant reductions in leverage of about 23% on average. The reductions in leverage, in turn, caused statistically and economically significant improvements in operating performance of about 28% on average. Changes in leverage are instrumented with the level of snow prior to the debt restructuring. The effect of snow is both statistically and economically significant: a one-standard deviation increase in snow is associated with a reduction in leverage of about 23% on average.