In this paper we investigate whether earnings management provides useful explanatory variables for bankruptcy prediction. We study a sample of US active and inactive public firms during the period 1998-2014. To test our hypothesis, we include several earnings management proxies in Ohlson’s bankruptcy prediction model. To avoid serial correlation and evaluate the predictive ability at specific times, we perform several nonlinear rolling models with 1-year estimation window. Our findings show that bankrupt firms engage in aggressive income-increasing earnings management in the year spreceding the bankruptcy and in the year of the bankruptcy. Our results also show earnings management measures increase the predictive ability of bankruptcyprediction models. Our results also show that real earnings management provide more significant explanatory variables for bankruptcy prediction than accrual earnings management. Our findings also show that the inclusion of both accrual earnings management and real earnings management significantly increases thepredictive ability of Altman’s Z-Score. Our results are robust to several robustness checks.

Our paper can contribute both to bankruptcy prediction and to the earnings management literature. Our study shows how to correct accounting-based bankruptcy prediction for the degree of managerial discretion incorporated in the accounting numbers, overcoming a limitation highlighted by prior studies. We also show that earnings management research can provide useful explanatory variables in bankruptcy prediction. Finally, we identify which earnings management tools are the best predictors of bankruptcy. The results of our study can be of interest for practitioners, banks, investors and other financial institutions interested in credit scoring and credit risk appraisal.