Abstract. The current research targeted to develop a methodology for bankruptcy prediction applicable for Romanian companies. Under the effect of the economic crisis, as well as of the entrance in the European Union, Romania has registered a significant increase of the yearly bankruptcy frequency, which makes the availability of bankruptcy assessment tools more important than ever before. Using the logistic regression, the study proposes a multivariate model based only on financial data easily accessible to all the stakeholders of the company.

The target population consisted of all the companies from the Timis County with annual sales of over 10,000 lei (aprox. 2,200 Euro). The model was developed over a paired sample that included all the companies from the 2010 target population that went bankrupt by the end of 2012. The testing was performed over the entire target population from the period 2007-2010. The study has thus included 53,252 yearly financial statements from the period 2007-2010.

The proposed model will not allow the analyst to conclude with absolute certainty whether the analyzed company will go bankrupt or not within a two year horizon. Instead, it will offer the possibility of placing the companies on risk classes, thus creating the basis for a differentiated behavior of the stakeholder toward a specific company.

Keywords: ratio analysis, financial statements, risk, accuracy, benchmark.