Use this URL to cite or link to this record in EThOS:
Title: The impact of financial and non-financial measures on banks' financial strength ratings : the case of the Middle East
Author: Abdallah, W. M.
Awarding Body: University of Salford
Current Institution: University of Salford
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Access from Institution:
The relationship between bank performance measures, namely financial and nonfinancial, and financial strength ratings (FSRs) has created an interesting area of research for many years. This thesis examines econometric qualities including explanatory, discriminatory and predictive powers. The main aims of this thesis are as follows: (1) to identify the main bank performance measures associated with high-FSRs versus low-FSRs; (2) to determine the bank performance measures that can discriminate banks associated with high-FSRs versus low-FSRs; and (3) to compare the predictive capabilities of conventional techniques versus machine-learning techniques in predicting banks’ FSR group memberships in the Middle East. The analysis is performed in three stages: (1) the analysis identifies the association between banks’ FSRs and performance measures by applying a multinomial logit technique; (2) the analysis uses the outcome of the first stage as an input to discriminate high-FSRs from low-FSRs using discriminant analysis; and (3) machine-learning techniques (i.e., CHAID, CART and multilayer perceptron neural networks) and conventional techniques (i.e., discriminant analysis and logistic regression) are used to predict banks’ FSR group memberships. Various performance evaluation criteria (i.e., average correct classification rate, misclassification cost and gains charts) are used to evaluate the predictive capabilities of various modeling techniques. The data set covers the Middle Eastern countries’ commercial banks from 2001 to 2009. Results from the first stage indicate that high-FSR banks in the Middle East are well capitalised, and profitability is associated with the highest relative explanatory power. Second stage results show that three financial variables (i.e., loan loss provision to total loans ratio, asset utilisation ratio and equity to net loans ratio) contribute greatly to the model’s discriminatory power. On the other hand, results for nonfinancial variables reveal that bank size and sovereign rating are the most important to the model’s discriminatory power. The results indicate that financial variables outperform nonfinancial variables in terms of overall discriminatory power. Finally, in the last stage, results show that the predictive capability of CHAID outperforms other machine-learning techniques (i.e., CART and multilayer perceptron neural networks). Regarding conventional techniques, the predictive capability of discriminant analysis is superior to logistic regression. In terms of comparing various predictive techniques, results of the performance evaluation criteria reveal that machine-learning techniques outperform conventional techniques in predicting banks’ FSR group memberships.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available