The Vietnam banking industry and its retail banking: Artificial neural networks and statistical analysis : A thesis submitted in partial fulfilment of the requirements for the Degree of Doctor of Philosophy in Finance at Lincoln University

Abstract

The 2008 Global Financial Crisis (GFC) had adverse effects on Vietnamese banks. After the crisis, the banks faced many unexpected challenges, including recession, stagnation in credit growth rates and an increased number of non-performance loans (NPL). The Vietnamese government released the “Restructuring Financial Institutions 2011-2015” on 1st March 2012. This report detailed steps to enhance the banking industry’s performance and ensure that it met international banking standards. The report recommended merger and acquisition (M&A) deals as one way to achieve its objectives to improve bank competitiveness and efficiency. The restructuring, via M&A deals, is expected to have an even greater impact on market concentration and competition in the Vietnamese banking industry. As Hoang, Phan, and Bandaralage (2016) note, monopolies may occur because of a decrease in the number of banks in the market. Using unbalanced yearly data from 2008 to 2017, this is the first study to use the self-organisation map (SOM) technique to track financial trajectory patterns and categorise Vietnamese banks into super-class groups. The study identifies two super-class groups. Group 1 has only joint-stock banks; group 2 contains commercial state and joint-stock banks. Using the non-structural indicator the Lerner index to capture market power and the Data Enveloped Analysis technique to measure efficiency, this study shows there are significant differences in Lerner scores (which represent bank market power) of the two bank groups. Though there are notable significant (at the 1% level) differences between the two bank groups’ market power, their cost-efficiency scores (which represent bank efficiency) are the same. The differences in the Lerner scores provide evidence of a group of strong banks that is isolated from other banks. This phenomenon is the result of market contraction through M&A activity. This study argues that this strong bank group has the potential to be monopolists and may damage the competitive banking environment.

Using the system generalised method of moments to overcome challenges associated with endogeneity and generating unbiased estimating parameters, this study also shows that bank size and capitalisation ratio have a positive relationship with market power. These findings suggest that large, well-capitalised banks enjoy greater market power because of their dominant position. Their greater capital reserves mean that they can take greater risks and lend to more borrowers. However, the negative relationship between bank size and cost efficiency indicates that big banks are less effective in distributing their costs than small banks, which leads to lower efficiency. This is because successful M&A activity may improve bank size quickly, but negatively impact on bank efficiency because of increased NPL values and operational costs. A higher NPL rate weakens both the efficiency and market power of Vietnamese banks. Bank executives in both groups 1 and 2 must focus on decreasing their NPL rate to ensure their efficiency and maintain their market power. The study also argues that Vietnamese banks gain efficiency during periods of rapid GDP growth or high inflation rate; in sum, banks raise their lending interest rates in response to a high inflation rate. The results also indicate that market power has trends to maintain over time. This is because the one-year lagged market power is positive and significantly impacts on bank market power. This finding indicates that the Vietnam banking industry is characterised by non-transparent information, networking lending relationships and limited regulations around monopolistic practices, which leads to bank monopolies and reduces competition and innovation.

Because of increased numbers of internet users and emerging e-commerce, retail banking is a profitable market for Vietnam domestic banks to ensure long-term development. However, less than 50% of bank executives are satisfied with their current credit internal rating systems; most believe that their credit scoring models (CSMs) are ‘poorly’ designed. This study proposes a more accurate combination method that enables banks to build a more efficient credit scoring model and hence make better credit-granting decisions. The component planes generated by the SOM technique provide a visual illustration of the classification process and thus can assist researchers in understanding the motivation process. The merger between Logit and Multilayer Perceptron outperforms the logit regression alone. This is because the overall classification accuracy is improved to 1.2% using the hybrid technique. Type II errors also decrease significantly, from 34.7% to 15.3% (a reduction of over 50 per cent), using the hybrid technique. This leads to a massive reduction in the misclassification cost (to 56%). Other evaluating discriminatory power criteria, the AUC, Gini and KS coefficients, provide robust evidence that the hybrid technique has greater classification power than logit regression alone.