The subject of the research of this PhD thesis is a critical analysis of the application of absolute and relative indicators of financial analysis in the function of developing a bankruptcy prediction model for the enterprises from processing and trade industries in the Republic of Serbia, as well as a comparative analysis of the results of its application in relation to the results of the application of selected traditional and contemporary bankruptcy prediction models for enterprises in the mentioned industries. A special attention was dedicated to the analysis of the impact of the industry on the power of the enterprises’ bankruptcy prediction when using contemporary bankruptcy prediction models. The main goal of the PhD thesis is to critically examine the advantages in anticipating the bankruptcy of a developed new model predicting bankruptcy of enterprises, based on the indicators of financial analysis with the application of logistic regression, in relation to selected traditiona...l and contemporary models for predicting the bankruptcy of the enterprises from processing and trade industries in the Republic of Serbia.
The sample consists of 204 enterprises from processing and trade industries in the Republic of Serbia, and the time horizon of observation includes the period from 2011 to 2017. The starting point of the research was the analysis of the financial performances of enterprises through 56 absolute and relative indicators, from which 6 relevant indicators were selected for their contribution to the development of a highly powerful predictive model.
As the main result of the research is developed and proposed new model, with the help of using logistic regression, for bankruptcy predicting of enterprises from processing and trade industries, suitable for use in the Republic of Serbia. The proposed model has a higher accuracy of predictions than traditional models developed for efficient markets, such as Altman, Ohlson, and the Zmijevsky models. The contemporary model developed by the application of neural networks has lower predictive accuracy regarding bankruptcy compared to the created model, while the model generated by using decision trees has higher predicting accuracy in comparison to the proposed model created by logistic regression. Within the dissertation is emphasized the difference in the effects of applying the bankruptcy prediction model of enterprises in the Republic of Serbia, developed by the application of logistic regression, when applying on enterprises form different industries. The bankruptcy prediction model developed by using neural networks has higher predictive power if applied to data from individual industry (only processing, or only trade industry), than if applied to data from both observed industries (processing and
trade industry together). However, the decision trees model shows equal accuracy in bankruptcy prediction when applied to data from individual industry as well as when applied to data from both observed industries.
