PREDICTION OF CORPORATE BANKRUPTCY IN ROMANIA THROUGH THE USE OF LOGISTIC REGRESSION

The purpose of this paper is to test weather data from synthetic financial statements publically available in Romania can be employed within a logistic regression model to accurately predict the corporate bankruptcy probability over the economic crisis period. The population initially subjected to our study included all the 26,980 companies from the Timis County that submitted financial statements for 2007 to the public authorities. As this population proved very heterogeneous, we focused on a more homogeneous group, composed of 4,327 companies. The target population was chosen by employing both economical and statistical criterions. The data from the synthetic financial statements was used to build 12 ratios, 5 of which have proven to be significant within a logistic model for predicting corporate bankruptcy. It was clear that other sources of information could help improve the accuracy of the predictions, but these sources are not easily accessible to most of the stakeholders. As the synthetic financial statements are publically available, evaluating their utility in the prediction of corporate bankruptcy was one of the main objectives of this research. The proposed model offers an in-sample overall 70.3% accuracy in the prediction of the bankruptcy event over a 5 - year period, with an out of sample overall accuracy of 67.6%. Under these circumstances, the model is considered to be of immediate practical utility, as it can represent a tool for performing a fast estimation of the bankruptcy probability of a company that fits the profile of the target population. As theoretical contributions, the research proves that the companies that filed for bankruptcy during the crisis period showed signs of weaknesses before the beginning of the crisis. Financial ratios that show relevance in the prediction of corporate bankruptcy at local level have been identified and their correlation with the bankruptcy probability has been evaluated. The model is expected to maintain its accuracy with minimal or no additional calibration for companies from the entire Romanian economy that fit the profile of the target population.