Моделирование вероятности банкротства предприятий реального сектора экономики

Вопросы статистики. 2018. Т. 25. № 12. С. 12–27.

Polyakova M. V., Еремеева И. С., Polyakov K.

This research focuses on forecasting situations of impending financial crisis (with potential bankruptcy) in real economy enterprises. By the term potential bankruptcy is understood an enterprise’s financial situation where sum of short-term and long-term liabilities exceed its assets, which leads to a negative net equity. That, as a rule, is explained by accumulated uncovered losses (negative retained earnings). The consequences of such a situation could be very severe. The difference between asset value and borrowed funds approximately equals to net asset value - a key indicator of financial performance of a joint stock company. The decrease of net asset value below the authorized capital for several years leads either to its reduction, which happens rarely, or to a bankruptcy. Furthermore, under certain conditions, an enterprise must publish a notice concerning its net asset reduction and any creditor is entitled to demand early performance of the respective obligations, which, if the amount of debts is large, can become a serious problem and result in a real bankruptcy.
A model of the statistical relationship between a set of financial indicators and a probability of bankruptcy is constructed in the article. We are focused on identifying of significant indicators and the functional form of their values binding with the probability. To solve these problems we use a logit regression-based class of models for panel data and an algorithm for an automatic model specification. It reduces the influence of human factor on determining its type and links it with attributes of the accumulated data. This research is based on 2012-2016 financial records of 463 Russian enterprises from the SPARK-Interfax database. The results of simulation allowed us to determine a set of indicators that significantly affect forecasting quality of potential bankruptcy possibility, as well as the nature of this effect.

Keywords: панельные данные, panel data, Logit Regression, fractional polynomials, обобщенные полиномы, potential bankruptcy simulation, MFP algorithm, моделирование вероятности банкротства, логит-регрессия, алгоритм MFP