A logistic regression approach to long-term bankruptcy prediction: The role of financial and non-financial indicators

The main aim of the research is to examine the possibility of developing logistic regression (LR) model that could reliably predict the bankruptcy of Serbian companies three years in advance based on financial and non-financial variables. This is important both for business owners and external stakeholders. Owners can predict failure on time and define remedial measures and action plans in accordance with that. External stakeholders, on the other hand, can use these models to identify financial risks before deciding to start cooperation with a specific company. The main motive for the research stems from the lack of bankruptcy prediction models in the scientific community of the Republic of Serbia, especially when it comes to long-term prediction. It is necessary to predict bankruptcy early enough to be able to take measures. A prediction one year in advance, which is a common case in the existing literature, may be too late to preserve the business’s future. According to the authors’ findings, no long-term prediction models have been developed for the Serbian market. Existing traditional models are developed for foreign countries, meaning they are not suitable for developing countries like Serbia. The research sample includes 94 companies of all sizes and is balanced: half of the companies are healthy, and the other half are bankrupt. A total of 36 financial and 7 non-financial independent variables are included in the modelling. Financial analysis is done in MS Excel, while statistical analysis (logistic regression) is done in IBM’s SPSS program v. 26. The research results demonstrate that statistical and financial analyses are effective for bankruptcy prediction modelling, considering that the generated model has significant predictive (classification) power of 80%.