Bankruptcy Prediction: Static Logit Model versus Discrete Hazard Models Incorporating Macroeconomic Dependencies

Abstract: The purpose of this master thesis is to (i) compare the out-of-sample prediction power of one static logit model and two hazard models, (ii) explore whether incorporating macroeconomic patterns improves forecasting results, and (iii) examine the determinants of corporate failures from the pool of accounting and market driven variables. We perform our study on 102 US listed manufacturing firms that defaulted between 2000 and 2009. The out-of-sample period spans over the crisis period 2007-2009. We find that the static logit outperforms both hazard model specifications in out-of-sample accuracy. Next, we find that incorporating macroeconomic patterns can improve forecasting results of hazard models. The most important firm-specific determinants of bankruptcy are profitability, stock return, short-term solvency, further significant variables are cash holdings, relative market size, and leverage.