فراتحليل مقايسه مدل‌هاي پيشبيني ورشكستگي و نقش متغيرهاي بازاري

The Meta-Analysis of the comparison of bankruptcy prediction models an‎d the role of market variables

مديريت

In this study, the aim of the meta-analytic comparison of the bankruptcy prediction power of models that have been implemented in previous empirical studies was to compare the prediction accuracy of the neural network approach with logistic regression, statistical analysis, decision tree, etc. based on the characteristics of the models, such as including the type of country, the type of eva‎luation window, the variable selec‎tion method, the end of the estimation period, the data interval until bankruptcy, an‎d the credit rating of the publication. In this study, using the meta-analytic approach, the contradiction in the studies was identified an‎d concluded that: 1- Which of the different models predicts bankruptcy better? 2- What is the relationship between increasing the number of market variables an‎d increasing the accuracy of bankruptcy prediction models? In this study, the meta-analysis approach was implemented in seven stages: determining the subject an‎d statistical population, collecting studies, screening studies that are applicable for meta-analysis, collecting information related to the results of each study an‎d examining its conditions, determining the effect size, calculating the joint effect size, an‎d assessing the homogeneity an‎d heterogeneity in the effect sizes. The statistical population of this study included all experimental tests conducted on bankruptcy prediction that were published in past scientific reports between 1990 an‎d 2022. 30 articles were finally reviewed an‎d 573 effect sizes were calculated an‎d used to test 2 hypotheses using CMA software. According to the results of this study, there is a significant difference between the prediction accuracy of neural network methods, regression, statistical analysis, decision tree, an‎d other methods; That is, it can be claimed that the neural network has a higher prediction accuracy than regression, an‎d according to the results, the decision tree has the highest prediction accuracy. On the other han‎d, with the increase in the number of market variables in bankruptcy prediction models, the accuracy of the model has increased; therefore, it seems that the market characteristics of companies can predict bankruptcy better than their accounting characteristics. Perhaps because the market an‎d capital market investors are likely to notice the bankruptcy of the company in a timely an‎d appropriate manner, an‎d this is reflected in the market behavior.

5