Bank Failure Prediction with Logistic Regression

Taha Zaghdoudi
2013 International Journal of Economics and Financial Issues   unpublished
In recent years the economic and financial world is shaken by a wave of financial crisis and resulted in violent bank fairly huge losses. Several authors have focused on the study of the crises in order to develop an early warning model. It is in the same path that our work takes its inspiration. Indeed, we have tried to develop a predictive model of Tunisian bank failures with the contribution of the binary logistic regression method. The specificity of our prediction model is that it takes
more » ... o account microeconomic indicators of bank failures. The results obtained using our provisional model show that a bank's ability to repay its debt, the coefficient of banking operations, bank profitability per employee and leverage financial ratio has a negative impact on the probability of failure. 1. Introduction The two last decades are marked by notable banking and financial crises by their extent as well as their exorbitant financial costs. In fact, many developing countries witnessed serious disturbances to their banking systems manifesting the companies' insolvency, the disturbances of exchange flow, which went even to the bankruptcy of big companies. Today, banking and financial crises show a sharp persistence, since most countries affected by the 2008 crisis could not find a way out of it yet. Moreover, the current crisis requires a bigger attention than that paid to its predecessors since it bears a worldwide and banking character. This crisis touched mainly the banking systems having difficulties, knowing that they represent the heart of the economic activity and its financing. From now on, every malfunctioning of banking system will change the behaviour of economic agents; create a feeling of distrust of investors and depositors towards credit establishments, which results into serious disturbances to real economy. The redundancy of these crises gives birth to a feeling of fear towards the installation of a cyclic and chronic phenomenon which exhausts all remedies without getting out of them. From now on, the establishment of an advanced alert model of banking and financial crises becomes more than necessary to better avert eventual financial jolts and seek even to avoid them. Up to here, the majority of empirical writings about crises advanced detection are of macroeconomic order. A study of macroeconomic data is then used to develop an alert system which is able to detect several financial crises in advance. Moreover, this empirical step presents limits since it can't detect the banking weaknesses of microeconomic nature beforehand. Actually, we think that the integration of a microeconomic approach in the construction of a banking crises precocious alert model could enlarge its detection power. This article is organised in the following way. In the section to come, we present a literature review on the microeconomic indicators of banking weaknesses. Section 3 methodology. In section 4, a data presentation and a selection of the model variables. We present our results of the estimation in section 5. Section 6 the conclusion of our work.