The Early Warning of Financial Failure for Iraqi Banks Based on Robust Adaptive Lasso Logistic Regression

Abstract

It is well known that there are many mathematical financial failure models that have been proposedin the financial literature for specific stock markets. Some researchers are not aware these mathematical modelswere constructed to be fitted for that market data, not for other markets. Iraq stock market exchange is one of thesemarkets in which the researchers used imported models such as Kida, Sherrod, Altman, and others to predictfinancial failure. Therefore, the development of a financial failure warning model for banks has become verycrucial for the Iraqi bank sector in the stock market exchange. Unfortunately, there is no clear information aboutthe financial failure of Iraqi banks as a response variable, and the financial indicators contain outliers. Theobjective of this paper is to propose an algorithm to know the performance of efficient and inefficient banks basedon their indicators during specific time periods. The output of this algorithm will be considered as responsevariables. Then, a weighted adaptive lasso logistic regression algorithm that has a high breakdown point is used totackle outliers’ problem. Thirteen banks have been chosen as the most traded during the period (2010-2017), andfor each bank (27) financial indicators were collected. Our proposed model is compared with adaptive lassologistic regression by using Deviance, Misclassification, Area Under Curve, Mean Square Errors, and MeanAbsolute Errors. Consequently, the results showed that the weighted Adaptive Lasso Logistic Regression model ismore robust and relevant than others to be a financial model to warn of the failure of the banks in Iraq's stockmarket.