Прогнозирование банкротства: эконометрическая модель для российских страховщиков
This work is devoted to creating a model which could predict bankruptcy of Russian insurance companies. The aim of the study is to build a model based on panel data; its final version should have a good predictive power. Said topic is relevant because the number of revoked licenses has changed a lot over the past few years — this situation may influence both insurance organizations and the population in a negative way. The paper reflects the main characteristics of bankruptcy as well as analyzes the bankruptcy prediction models which have been made by various authors since the 20th century. In the practical part of the study, an econometric analysis of the collected data was carried out and a logit model was built. The model’s predictive power was tested on a sample of insurers. In addition, a random forest algorithm and a binary classification tree algorithm were used. As a result, it was discovered that the volume of insurance premiums to net profit ratio, which could be calculated only for insurers, and financial stability coefficients influence insurance companies’ bankruptcy the most. Further research can be expanded by including new, more sophisticated methods, such as neural networks or boosting.