Competition and bank financial stability: evidence from an emerging economy

Abstract

This article aims to empirically examine the role of competition and concentration on Vietnamese bank stability efficiency in Vietnam using data from 25 commercial banks from 2007 to 2020 using the Tobit regression method is utilized. The results suggest that banks exhibit greater stability in a highly competitive environment. This connection between competition and the efficiency of Vietnamese banks remains significant even amidst financial crises. Furthermore, we employ various alternative risk indicators and exclude state-owned banks from our sample to reassess the nexus between competition and bank stability efficiency, with results that remain steadfastly robust. This study represents a pioneering endeavor to investigate the impact of competition on bolstering banks’ stability through the utilization of the stochastic frontier approach. Our study findings suggest that the implementation of expansion policies that promote competition can improve the efficiency of bank stability in developing markets.

Keywords:

• Competition
• bank risk-taking
• stability efficiency
• Lerner index
• Vietnam

JEL Classification:

• G21
• G32
• G34

Reviewing Editor:

• David McMillan, University of Stirling, United Kingdom of Great Britain and Northern Ireland

Subjects:

• Finance
• Banking
• Financial Management
• Risk Management

1. Introduction

It is widely recognized that a robust financial system is essential for the advancement and growth of economies. A key factor influencing global shifts was the aim to boost the banking sector’s competitive edge (Delis, 2012). Yet, the notion that competition benefits all sectors universally might not be as firmly rooted in the banking industry as it is in other areas (Claessens & Laeven, 2004). Despite thorough investigations, consensus on how competition affects bank stability remains elusive. The academic field presents two diverging views. Initially proposed by Keeley (1990), the competition-fragility or competition-instability viewpoint suggests that increased competition in the banking sector reduces banks’ market power, cuts into their profit margins, and decreases the value of their franchises, pushing them toward higher risk-taking. Conversely, the competition-stability argument, significantly advanced by Boyd and DE Nicoló (2005), maintains that competition actually promotes stability within banks.

In the past two decades, the landscape of competition within the commercial banking sector has undergone significant transformation. Factors, such as the liberalization, deregulation and privatization of the financial market have played pivotal roles in shaping these changes. These developments have not only intensified competition within the banking sector but have also brought into play external competitors, primarily non-banking financial institutions and market-based financing options. The thrust of financial sector reforms has been toward broadening the scope of various segments of the financial market, including the insurance industry, mutual funds, equity and debt markets and non-bank financial entities. These changes have collectively contributed to heightened competition across the board (Prakash et al., 2022; Turk Ariss, 2010). Furthermore, empirical evidence from different nations presents a varied set of outcomes, indicating a lack of conclusive data on how competition influences the stability of banks. This ambiguity underscores the need for further exploration into this compelling area. Consequently, this has drawn the attention of researchers, policymakers and regulatory bodies toward uncovering the dynamics that ensure bank stability amidst growing competitive pressures.

Vietnam has been chosen as the object for this analysis for several reasons. The country’s economy has seen impressive performance, establishing itself as a key player within the Association of Southeast Asian Nations (ASEAN). It has achieved an average annual growth rate of 6.02% from 2007 to 2020. Given the relatively nascent state of its financial system, the banking sector has been instrumental in driving the country’s notable economic expansion. Additionally, the entry of Vietnam into the World Trade Organization (WTO) in 2007 marked the beginning of increased competitive dynamics among banks. In recent years, financial technology entities and payment service companies have gained prominence, positioning themselves as formidable competitors against traditional banks. This evolving landscape, as highlighted by Danisman (2018), has intensified the competitive pressures on established commercial banks. Thus, it becomes crucial to examine the impact of competition on the efficiency and stability of banks in Vietnam.

This study breaks new ground by investigating the link between bank competitiveness and stability in a growing economy, enriching academic discussions with fresh insights. It revisits the connection between competitiveness and stability, employing both traditional and efficiency-adjusted Lerner indices for market power, and moving beyond conventional risk metrics like nonperforming loans (NPLs) ratio or Z-scores. Acknowledging the limitations highlighted by Fan et al. (2019) and Tan (2016), and the critique of Z-score’s predictive accuracy by Tabak et al. (2012), this research employs a stochastic frontier approach to assess bank stability and re-examine competitiveness and risk in Vietnam’s banking sector. This contribution is pivotal, addressing research gaps and offering significant implications for banking management and regulation.

Our research assesses a group of 25 commercial banks in Vietnam spanning from 2007 to 2020 to investigate the effects of competition on bank risk-taking. The results suggest that banks exhibit greater stability in environments with higher levels of competition and concentration. Furthermore, our findings demonstrate that this relationship remains valid even during times of financial crisis. Additionally, we utilize various alternative measures of risk while excluding state-owned banks from our analysis to re-evaluate the connection between competition and bank stability efficiency; these results remain strong and reliable.

The remaining sections of this article are organized as follows: Section 2 provides a succinct overview of the current literature. In Section 3, we explicate our methodology and data sources. Empirical findings are presented in Section 4, with concluding remarks offered in Section 5.

2. Literature review

2.1. The link between competition and bank stability

The correlation between competition and bank stability can be mainly categorized into two divergent viewpoints.

2.1.1. Competition-fragility

The competition-fragility perspective asserts as heightened levels of competition are associated with the instability of banking sectors. According to the research conducted by Keeley (1990), the lack of regulation and elimination of interest rate ceilings within the banking sector of the United States during the 1970s and 1980s served as a catalyst for banks to partake in ventures that carried inherent risks. Consequently, this led to the occurrence of numerous instances of bank failures. This viewpoint is founded upon the correlation between a bank’s franchise value and its propensity for engaging in risky activities. According to Northcott (2004), banks modify the level of risk in their investments in reaction to the degree of competition they encounter. As the level of competition intensifies, the profitability and franchise value of banks diminish. This phenomenon offers bank managers with gambling incentives, as it reduces their potential losses in the event of a decline in the value of their franchise. Consequently, bank managers may opt for investments with higher levels of risk, engage in reckless decisions regarding lending, and exhibit a more lenient approach toward choosing investments and requirements for lending (Li, 2019; Northcott, 2004; Saif-Alyousfi et al., 2020). According to Soedarmono et al. (2011), it is widely believed that franchise value serves as an effective mechanism for ensuring self-discipline among bank managers. Bank managers are motivated to make cautious investment decisions, maintain higher capital buffers, and enhance loan monitoring and screening activities due to the elevated opportunity cost of bankruptcy that is linked to a greater franchise value (Akins et al., 2016; Fu et al., 2014; Rakshit & Bardhan, 2019). The competition-fragility view posits that banks, in response to heightened competition, may relax their lending standards and allocate insufficient resources to the evaluation and supervision of borrowers. This behavior is driven by the banks’ desire to retain their current customer base and attract new customers. However, these actions can ultimately result in an elevated risk of bank insolvency (Danisman & Demirel, 2019; Sinha & Sharma, 2016). Furthermore, Petersen and Rajan (1995) noted that the competition-fragility paradigms exacerbated by the lack of relationship lending incentives. Relationship lending involves building and maintaining long-term, mutually beneficial customer relationships. Despite the high cost, cultivating relationships has been shown to benefit banks financially. This is because valuable and exclusive borrower information improves screening and monitoring (Northcott, 2004). Due to increased competition, banks may have less incentive to cultivate customer relationships. This is because businesses cannot make as much money from these customers (Beck et al., 2013). Financial institutions struggle to recover the costs of nurturing relationships with borrowers in a competitive banking industry because borrowers tend to switch funding sources. Reduced bank oversight and assessment of borrowers can increase bank default risk and suboptimal capital allocation (Northcott, 2004; Turk Ariss, 2010).

A lot of empirical evidence supports the competition fragility perspective. Fungáčová and Weill (2013) found that increased competition in the Russian banking sector between 2001 and 2007 increased the risk of a banking crisis. Bank concentration was not considered in the researchers’ analysis. Kasman and Kasman (2015) strongly supported the competition-fragility hypothesis. A panel dataset of 28 Turkish commercial banks from 2002 to 2012 was used. A comprehensive analysis of bank-level variation was not done. Sinha and Sharma (2016) used the H-statistic to assess competition-fragility support in a 2000–2015 sample of Indian scheduled commercial bank. Halim et al. (2023) utilized the Lerner index and HHI to explore the relationship between competition, ownership structure, and economic growth of banks with credit risk and financial stability in the Middle East and North African (MENA) economies. The results reveal that lower levels of bank competition are associated with lower risk-taking by banks and improved financial stability. Srivastava et al. (2023) conducted a study utilizing the Lerner and Boone indices to examine the impact of competition on the stability of Indian banks. The findings indicate that banks are less stable in environments that are more competitive and concentrated. Bakhouche (2024) investigates the link between competition and stability within Tunisia’s banking industry, spanning the years 2005–2020, to determine the influence of cost efficiency on this dynamic. The findings indicate that competition leads to a decrease in stability, aligning with the competition-fragility hypothesis.

H1a:

There is a positive relationship between Competition and Vietnamese bank risk-taking.

2.1.2. Competition-stability

The competition-stability perspective asserts that a higher level of competition results in a creation of more stable banks. This perspective is grounded in the moral hazard and adverse selection issues that may arise when borrowers possess excessive autonomy in their investment choices. In their study, Boyd and DE Nicoló (2005) discovered a negative correlation between heightened competition and bank lending rates, indicating that as competition intensifies and lending rates tend to decrease. When borrowers are able to secure loans at lower interest rates, they are less motivated to undertake ventures that involve higher levels of risk, as they are more likely to possess the means to repay their debts. Consequently, banks encounter diminished levels of borrower default risk and non-performing loans.

According to Singla and Singh (2019), competition serves as a crucial mechanism for external monitoring. The authors contend that competition serves as a crucial governance mechanism in the oversight and regulation of bank managers. A competitive market diminishes the level of managerial discretion and provides bank managers with incentives to enhance performance, despite the heightened risk of bankruptcy. Consequently, this helps to alleviate agency problems. According to the findings of Schaeck and Cihak (2008), there is empirical support for the notion that competition serves as a catalyst for bank managers to pursue efficiency enhancements. Amidu and Wolfe (2013) emphasize that competition plays a significant role in motivating bank managers to engage in diversification strategies, both within and across fee-based income-generating activities, while also assuming diversified risks.

Thanh et al. (2024) conducted a study on the dynamics between bank competition and stability, specifically within the evolving economic landscape of Vietnam, covering the years 2007–2018. The results indicate a positive correlation, showing that increased competition is associated with enhanced stability among Vietnamese banks. Azmi et al. (2024) show that competition has a varied impact on bank stability in dual banking economies, affecting Islamic banks positively but showing no effect on conventional banks.

The competition-stability perspective is supported by a significant corpus of empirical data. The findings of Guo and Chai (2024), Hussain and Bashir (2020) and Li and Peng (2024) study indicate a positive relationship between increased competition and improved stability of Chinese banks. Jeon and Lim (2013) conducted a study wherein they employed two separate samples of depository institutions in Korea for their research. The initial sample comprised of mutual savings banks, encompassing the timeframe spanning from 2003 to 2011. The second sample consisted of commercial banks, covering the time period from 1999 to 2011. The main aim of their inquiry was to examine and contrast the discrepancies between these two classifications of establishments. The observed disparities encompassed a range of factors, such as the structure of ownership, the characteristics of loans, the framework of governance and the treatment by regulatory authorities. The researchers made a discovery that indicates heightened competition has a positive impact on the stability of mutual savings banks, but conversely has a negative impact on the stability of commercial banks. The study conducted by Soedarmono et al. (2013) employed data from commercial banks in 11 Asian countries. The research findings suggest that an increase in competition between 1994 and 2009 led to improved stability within the banking sector. Furthermore, it has been found that an escalation in market power is positively associated with increased volatility in profitability and a heightened level of risk of insolvency. In addition, it has been found that the moderating impact of competition was less evident in situations where banks obtained subsidies as a result of their ‘too big too fail’ designation from the lender of last resort. Akins et al. (2016) conducted a study that investigated the relationship between competition and stability in the commercial banking sector of the United States. The utilization of the Herfindahl index was employed to evaluate competition. Nevertheless, we argue that this measure inadequately accounts for the competition faced by commercial banks from alternative financial intermediaries. In this context, it may have been more appropriate to consider an alternative metric, such as the Lerner index, which directly examines a bank’s conduct, as a potentially more suitable measure for evaluating competition. The study conducted by Akins et al. demonstrates that increased competition has a moderating impact on the probability of bank failure and also affects the distribution of resources toward investments with lower levels of risk. The findings of Maji and Hazarika (2018) study indicate that there were no statistically significant results when examining the impact of competition on credit risk, as measured by the H-statistic. The absence of statistical significance in the findings was attributed by the researchers to the underlying assumption of long-run equilibrium in the H-statistic. Numerous empirical studies have substantiated the competition-stability perspective, as evidenced by the works of Beck et al. (2006), Schaeck and Cihak (2008), Soedarmono et al. (2011), Anginer et al. (2014), Noman et al. (2018) and Li (2019). However, the current body of literature has not yet examined the diversity within banks in any of the aforementioned studies.

H1b:

There is a negative relationship between Competition and Vietnamese bank risk-taking.

3. Data and methodology

3.1. Data

The sample utilized in this study comprises 25 Vietnamese commercial banks, spanning the time period from 2007 to 2020. These banks were selected to represent a significant portion, specifically over 80%, of the total assets within the Vietnamese banking industry. Bank-specific data is primarily sourced from annual financial reports, encompassing balance sheets and income statements, which are accessible and can be obtained from the websites of the banks and the database (Le et al., 2022)¹. The collection of macroeconomic variables is sourced from the World Bank. Furthermore, it is worth noting that the focus of this analysis is solely on commercial banks, as they are the primary and most dynamic players in the Vietnamese market. It is important to acknowledge that foreign bank affiliates and joint-venture banks face certain limitations in their operations within the Vietnam market. As a result of multiple bank mergers during the designated study period, our ultimate data sample consists of an unbalanced panel comprising 350 observations.

4. Variable definitions

4.1. Dependent variable – stability efficiency

Numerous empirical studies employ the Z-score as a risk/stability indicator in the banking industry, and it is computed by adding a bank’s return on assets and equity to total assets ratio risk (Mirzaei et al., 2013). The Z-score measures the distance from insolvency and is formulated as: (1) $$Z-\mathit{\text{scor}}{e}_{i,t}=\frac{{\mathit{\text{ROA}}}_{i,t}+E_{i,t}/{\mathit{\text{TA}}}_{i,t}}{\sigma {\mathit{\text{ROA}}}_{i,t}}$$(1) where ${\mathit{\text{ROA}}}_{i,t}$ presents return on assets ratio; the ratio $E_{i,t}/{\mathit{\text{TA}}}_{i,t}$ denotes the proportion of equity over total assets; $\sigma {\mathit{\text{ROA}}}_{i,t}$ is the standard deviation of return on assets and is calculated using the three-year rolling window.

However, research by Tan (2016) and Fan et al. (2019) contends that Z-score may not fully reflect the potential stability that each bank might attain. Additionally, the deviation between the bank’s current stability and its maximum stability given economic and regulatory conditions should be taken into account. Therefore, the term ‘stability inefficiency’ is developed. The extent of stability inefficiency indicates how much a certain bank deviates from the optimal Z-score. In this study, the Vietnamese banks stability inefficiency is estimated on the basis of a translog cost function with one output (total assets), three input prices (price of deposits, price of labor and price of physical capital). The two fixed netputs (fixed assets and total equity), and technical change² are included to control for observable heterogeneity. This study specifies both these inputs and outputs of banks based on the intermediation theory and banks are viewed as financial intermediaries that convert deposits into loans and other earning assets (Sealey & Lindley, 1977). All variables in the translog cost function are defined in Table 1 below.

Table 1. Variables definition for estimating bank stability efficiency.

Variable	Definition
$Z-\mathit{\text{scor}}{e}_{i,t}$	The index is obtained from Equation (1)(1) $$Z-\mathit{\text{scor}}{e}_{i,t}=\frac{{\mathit{\text{ROA}}}_{i,t}+E_{i,t}/{\mathit{\text{TA}}}_{i,t}}{\sigma {\mathit{\text{ROA}}}_{i,t}}$$(1)
$Q_{\mathit{\text{it}}}$ (output)	Total assets
$W_1$ (Input price of deposits)	The ratio of interest expenses to total deposits
$W_2$ (Input price of labour)	The proportion of personnel expenses/number of employees
$W_3$ (Input price of physical capital)	The proportion of other operating cost to fixed assets
$Z_1$m (Fixed net put 1)	Fixed assets
$Z_2$m (Fixed net put 2)	Total equity
$\mathit{\text{Trend}}$	Technical change
${\epsilon }_i$	Error term

Source: Author’s summary.

Consequently, the specification of the translog cost function is as follows: (2) $$\mathrm{\text{Ln}}\left(\frac{{Z-\mathit{\text{score}}}_{\mathit{\text{it}}}}{W_{3,\mathit{\text{it}}}}\right)={\alpha }_0+{\alpha }_1{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+\frac{1}{2}{\alpha }_2{\left({\mathit{\text{lnQ}}}_{\mathit{\text{it}}}\right)}^2+{\sum }_{m=1}^2{\beta }_n\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\sigma }_m{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+\frac{1}{2}\mathrm{ }{\sum }_{m=1}^2{\sum }_{j=1}^2{\gamma }_{\mathit{\text{mj}}}\mathrm{l}\mathrm{n}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)\mathit{\text{ln}}\left(\frac{W_{\mathit{\text{jit}}}}{W_{3,\mathit{\text{it}}}}\right)+\frac{1}{2}\mathrm{ }{\sum }_{m=1}^2{\sum }_{j=1}^2{\pi }_{\mathit{\text{mj}}}\mathit{\text{ln}}{Z}_{\mathit{\text{mit}}}\mathit{\text{ln}}{Z}_{\mathit{\text{jit}}}+{\sum }_{m=1}^2{\delta }_m\mathit{\text{ln}}{Q}_{\mathit{\text{it}}}\mathit{\text{ln}}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\epsilon }_m{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\sum }_{m=1}^2{\sum }_{j=1}^2{\theta }_{\mathit{\text{mj}}}\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right){\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\phi }_1\mathit{\text{Trend}}+\frac{1}{2}{\mathrm{\phi }}_2{\left(\mathit{\text{Trend}}\right)}^2+{\phi }_3\mathit{\text{Trend}}{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+{\sum }_{m=1}^2{\mu }_m\mathit{\text{Trend}}\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\vartheta }_m\mathit{\text{Trend}}{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\epsilon }_i$$(2)

The error term ${\epsilon }_i$ consists of the one-sided time-varying inefficiency component (u_i) and the two-sided random error term (v_i), which captures the time-invariant heterogeneity as opposed to inefficiency. The inefficiency term (u_i) is independently and identically distributed with a non-negative truncated normal distribution and is obtained using the formula proposed by Jondrow et al. (1982). Employing the two-step approach (Coelli et al., 2005), we estimate the bank stability efficiency scores by extracting them from the error term, specifically ${\mathit{\text{ZW}}}_{i }=E\left[\mathrm{\text{exp}}\hspace{0.17em}\left(-u_i\right)\right]$. A higher stability efficiency score implies the lower level of bank risk and vice versa. α, β, σ, γ, π, δ, ω, θ, φ, μ, and τ are the estimated parameters, respectively.

4.2. Independent variables – competition

Researchers use the Lerner index to assess bank market power, a widely accepted measure. Due to its advantages over HHI, concentration ratio, and H-statistics, the Lerner index is chosen. Blair and Sokol (2015) claim that the Lerner index is the most widely used economic metric for market power. The proposed method considers pricing power, market contestability and bank revenue, making it more precise than the market concentration method, H-statistic and concentration ratio. Coccorese (2009) supports the Lerner index’s assessment of banks’ market power, particularly their deviation from monopoly and perfect competition. In addition to addressing small sample bias, the Lerner index can assess market power at a bank-year level (Jeon et al., 2011). This makes it ideal for this study, which uses structural panel data and a small sample size of 25 Vietnamese joint stock commercial banks. The methodology also effectively assesses market power of banks of different ownership structures and sizes, taking market concentration and demand elasticity into account. The conventional Lerner index (CMP) calculation by Berger et al. (2009) is used in the study. Price power, the percentage difference between the output price and the marginal cost (MC), is being measured. Market power is measured from 0 (perfect competition) to 1 (pure monopoly). This index suggests that market power increases with index value. The Lerner index below zero indicates that the output price is lower than the MC, which may be due to suboptimal bank behavior (Fu et al., 2014). It can be expressed algebraically as: (1) $$\text{Conventional\hspace{0.17em}\hspace{0.17em}Lerner\hspace{0.17em}\hspace{0.17em}}{\text{index}}_{\text{it}}\left(\mathit{\text{CMP}}\right)=\frac{P_{\mathit{\text{it}}}-M{C}_{\mathit{\text{it}}}}{P_{\mathit{\text{it}}}}$$(1) where P_it is bank i’s average production price, calculated by dividing total income (including interest and non-interest income) by total assets at time t. Using statistical data to estimate loan and deposit prices is limited. Thus, a single indicator since bank total assets are widely used to measure output (Berg & Kim, 1994; Shaffer, 1993). Due to financial product and service heterogeneity, the output price includes interest and non-interest income. Bank i’s MC of total assets at time t is MC_it. However, it cannot compute directly due to lack of information and no perfect method to determine inputs that reflect all bank activities and financial intermediary roles. A popular intermediary approach that calculates MC from the following translogarithmic cost function is effective (Berger et al., 2009; Sealey & Lindley, 1977; Turk Ariss, 2010): (2) $$\begin{array}{c}{\text{LnTC}}_{\text{it}}={\alpha }_0+{\alpha }_1\mathit{\text{ln}}{Q}_{\mathit{\text{it}}}+\frac{1}{2}{\alpha }_2{\left(\mathit{\text{ln}}{Q}_{\mathit{\text{it}}}\right)}^2+{{\displaystyle \sum }}_{k=1}^3{\beta }_k\mathit{\text{ln}}{W}_{k,\mathit{\text{it}}}\\ +\frac{1}{2}{{\displaystyle \sum }}_{k=1}^3{\gamma }_k\mathit{\text{ln}}{Q}_{\mathit{\text{it}}}\mathit{\text{ln}}{W}_{k,\mathit{\text{it}}}+\frac{1}{2}{{\displaystyle \sum }}_{k=1}^3{{\displaystyle \sum }}_{j=1}^3{\delta }_{\mathit{\text{kj}}}\mathit{\text{ln}}{W}_{j,\mathit{\text{it}}}+{\phi }_1\mathit{\text{Trend}}\\ +\frac{1}{2}{\phi }_2{\left(\mathit{\text{Trend}}\right)}^2+{\phi }_3\mathit{\text{Trendln}}{Q}_{\mathit{\text{it}}}+{{\displaystyle \sum }}_{k=1}^3{\mu }_k\mathit{\text{Trendln}}{W}_{k,\mathit{\text{it}}}+{\epsilon }_i\end{array}$$(2)

In this context, TC represents the aggregate cost incurred by the bank. Q serves as a measure of the bank’s output, specifically its total assets. W refers to the prices of three inputs: labor (W₁), funds (W₂) and fixed capital (W₃). Lastly, ε represents the error term in the model. According to Koetter et al. (2012), the bank’s production function involves the utilization of labor and physical capital as inputs to acquire deposits, which are subsequently utilized to finance loans and other income-generating assets. The calculations provided are as follows. The total cost is determined by the aggregate of interest expenses, personnel expenses and other operating and administrative expenses. The variables W1, W2 and W3 represent the ratios of personnel expenses (or staff costs) to total assets, interest and similar expenses to deposits from customers and other operating and administrative expenses (excluding staff costs, which are already accounted for in W1) to fixed assets, respectively. The inclusion of a time trend in the analysis accounts for the impact of technological advancements, resulting in shifts in the cost function over a period of time. In the event of a technical modification, holding all other factors constant, the average cost is expected to decrease. To estimate the cost function, the utilization of fixed effects is utilized to capture the impact of potential unobserved variables that are specific to each bank (Maudos & Nagore, 2005).

The estimation is typically conducted with the imposition of symmetry restrictions, specifically the condition that α_it = α_ti. The literature acknowledges that the estimated MC comprises the cumulative value of both marginal financial and operating costs, while disregarding the expenses associated with risk (Berger et al., 2009; Fu et al., 2014; Maudos & Nagore, 2005). The MC of the bank is determined by taking the first derivative of the total cost function derived from equation (2)(2) $$\mathrm{\text{Ln}}\left(\frac{{Z-\mathit{\text{score}}}_{\mathit{\text{it}}}}{W_{3,\mathit{\text{it}}}}\right)={\alpha }_0+{\alpha }_1{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+\frac{1}{2}{\alpha }_2{\left({\mathit{\text{lnQ}}}_{\mathit{\text{it}}}\right)}^2+{\sum }_{m=1}^2{\beta }_n\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\sigma }_m{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+\frac{1}{2}\mathrm{ }{\sum }_{m=1}^2{\sum }_{j=1}^2{\gamma }_{\mathit{\text{mj}}}\mathrm{l}\mathrm{n}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)\mathit{\text{ln}}\left(\frac{W_{\mathit{\text{jit}}}}{W_{3,\mathit{\text{it}}}}\right)+\frac{1}{2}\mathrm{ }{\sum }_{m=1}^2{\sum }_{j=1}^2{\pi }_{\mathit{\text{mj}}}\mathit{\text{ln}}{Z}_{\mathit{\text{mit}}}\mathit{\text{ln}}{Z}_{\mathit{\text{jit}}}+{\sum }_{m=1}^2{\delta }_m\mathit{\text{ln}}{Q}_{\mathit{\text{it}}}\mathit{\text{ln}}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\epsilon }_m{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\sum }_{m=1}^2{\sum }_{j=1}^2{\theta }_{\mathit{\text{mj}}}\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right){\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\phi }_1\mathit{\text{Trend}}+\frac{1}{2}{\mathrm{\phi }}_2{\left(\mathit{\text{Trend}}\right)}^2+{\phi }_3\mathit{\text{Trend}}{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+{\sum }_{m=1}^2{\mu }_m\mathit{\text{Trend}}\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\vartheta }_m\mathit{\text{Trend}}{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\epsilon }_i$$(2) mentioned earlier. (3) $${\mathrm{\text{MC}}}_{\mathrm{\text{TAit}}}=\frac{{\partial \mathit{\text{TC}}}_{\mathit{\text{it}}}}{{\partial Q}_{\mathit{\text{it}}}}=\frac{{\mathit{\text{TC}}}_{\mathit{\text{it}}}}{Q_{\mathit{\text{it}}}}\times {[\alpha }_1+{\alpha }_2{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+\frac{1}{2}\mathrm{ }{\sum }_{k=1}^3{\gamma }_k\mathit{\text{ln}}{W}_{k,\mathit{\text{it}}}+\mathrm{ }{\phi }_3\mathit{\text{Trend}}]$$(3)

However, the conventional Lerner index assumes that banks are capable of attaining complete efficiency, meaning they do not employ excessive scarce resources or allocate resources in inadequate proportions (technical and allocative efficiency). If banks do not meet these efficiency criteria, the calculation using the conventional approach may be biased, as banks could take advantage of pricing opportunities that arise from their market power. Hence, the author proceeds to estimate Equation (2)(2) $$\mathrm{\text{Ln}}\left(\frac{{Z-\mathit{\text{score}}}_{\mathit{\text{it}}}}{W_{3,\mathit{\text{it}}}}\right)={\alpha }_0+{\alpha }_1{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+\frac{1}{2}{\alpha }_2{\left({\mathit{\text{lnQ}}}_{\mathit{\text{it}}}\right)}^2+{\sum }_{m=1}^2{\beta }_n\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\sigma }_m{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+\frac{1}{2}\mathrm{ }{\sum }_{m=1}^2{\sum }_{j=1}^2{\gamma }_{\mathit{\text{mj}}}\mathrm{l}\mathrm{n}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)\mathit{\text{ln}}\left(\frac{W_{\mathit{\text{jit}}}}{W_{3,\mathit{\text{it}}}}\right)+\frac{1}{2}\mathrm{ }{\sum }_{m=1}^2{\sum }_{j=1}^2{\pi }_{\mathit{\text{mj}}}\mathit{\text{ln}}{Z}_{\mathit{\text{mit}}}\mathit{\text{ln}}{Z}_{\mathit{\text{jit}}}+{\sum }_{m=1}^2{\delta }_m\mathit{\text{ln}}{Q}_{\mathit{\text{it}}}\mathit{\text{ln}}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\epsilon }_m{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\sum }_{m=1}^2{\sum }_{j=1}^2{\theta }_{\mathit{\text{mj}}}\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right){\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\phi }_1\mathit{\text{Trend}}+\frac{1}{2}{\mathrm{\phi }}_2{\left(\mathit{\text{Trend}}\right)}^2+{\phi }_3\mathit{\text{Trend}}{\mathit{\text{lnQ}}}_{\mathit{\text{it}}}+{\sum }_{m=1}^2{\mu }_m\mathit{\text{Trend}}\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{W_{\mathit{\text{mit}}}}{W_{3,\mathit{\text{it}}}}\right)+{\sum }_{m=1}^2{\vartheta }_m\mathit{\text{Trend}}{\mathit{\text{lnZ}}}_{\mathit{\text{mit}}}+{\epsilon }_i$$(2) employing a stochastic cost frontier methodology that incorporates the potential cost inefficiencies of banks, commonly referred to as the efficiency-adjusted Lerner index (AMP). This approach utilizes maximum likelihood estimation and is widely regarded as more effective in assessing the impact of competition on efficiency. The cost (in)efficiency of a bank can be measured by assessing the disparity between the minimum cost and the actual cost output. In order to employ stochastic frontier analysis (SFA), it is necessary to possess knowledge of the parametric form of the error distribution. This study aligns with the existing literature and incorporates certain assumptions. Specifically, the error term ($v_i$) is considered a two-sided error that accounts for statistical noise. It is assumed to follow a standard normal distribution with zero mean and a certain variance. On the other hand, the random variable u_it represents cost inefficiency and is strictly non-negative. It is assumed to follow a half-normal distribution, which means that its normal distribution is truncated at zero from below. The variance of u_it is also taken into account. In relation to the variable of interest, alternative distributional assumptions such as exponential, truncated normal and gamma distributions have been proposed in previous studies (Bayeh et al., 2021; Koetter et al., 2012; Williams, 2012). The primary distinction between the two methodologies primarily lies in their technical approach to measuring MC. Although there may be technical variations, it is anticipated that these variations will yield comparable outcomes on the dependent variable.

4.3. Explanatory variables

Further, the too-big-to-fail hypothesis illustrates that larger banks are more likely to take greater risks, which ultimately results in a higher level of bank insolvency (Beck et al., 2006). However, larger banks may be more efficient and stable as they are less financially constrained (Karavitis et al. (2021). Thus, we include bank size (${\mathit{\text{SIZE}}}_{\mathit{\text{it}}}$), measured by the natural logarithm of total assets, to account for the effects of bank size on Vietnamese stability efficiency. Bank performance may also have a substantial impact on corporate risk (Kinateder et al., 2021). Therefore, this research further explores the linkage between the performance of Vietnamese banks, calculated by the return on assets (${\mathit{\text{ROA}}}_{\mathit{\text{it}}}$), and their stability efficiency. Moreover, Le (2020) asserts that the higher a bank’s liquidity ratio, the greater its exposure to risk. As a result, the ratio of liquid assets on total assets (${\mathit{\text{LATA}}}_{\mathit{\text{it}}}$) is employed to monitor the liquidity risk on bank stability efficiency. In addition, bank’s solvency may be significantly affected by the higher degree of operating costs than operating profit. Bank capitalisation ${\mathit{\text{CAP}}}_{\mathit{\text{it}}}$ is measured by the ratio of total equity to total assets. The positive association between bank stability efficiency and CAP is suggested by the conventional view, indicating that banks with higher capital and lower leverage tend to exhibit greater efficiency compared to those with lower capital, as stated in the research by Kwan and Eisenbeis (1997).

For macroeconomic variables the annual GDP growth rate and the annual inflation rate $(\mathrm{\text{INF}}$) are employed to examine the influences of economic expansion and inflation consequences on bank stability efficiency in Vietnam (Le, 2020; Perry, 1992).

4.4. Model specification

The Tobit regression is suggested as a suitable approach to address the constraint posed by the range of the bank stability efficiency score, which typically falls between 0 and 1 (Kumbhakar & Lovell, 2003). The Tobit model is capable of incorporating the characteristics of the efficiency metrics distribution, leading to the generation of reliable and coherent outcomes (Sufian & Akbar Noor Mohamad Noor, 2009). The Tobit regression, also known as the one-sided censored Tobit model, is employed in this study to address the limited range of efficiency values (ranging from 0 to 1) obtained from the application of the stochastic frontier approach in the initial stage. This approach is supported by the works of Adesina (2019), Eyceyurt Batir et al. (2017), and Shaban and James (2018). Next, proceed to perform a regression analysis incorporating the competition variable in the subsequent stage. Based on the studies of Adesina (2019), Eyceyurt Batir et al. (2017), and Meles et al. (2016), we established the following baseline model. $${\mathit{\text{ZW}}}_{\mathit{\text{it}}}={\alpha }_0+{\beta }_1{\mathit{\text{Competition}}}_{\mathit{\text{it}}}+\gamma {\mathit{\text{Bank characteristics}}}_{\mathit{\text{it}}}+{\theta \mathit{\text{macroeconomic}}}_{\mathit{\text{it}}}+{\epsilon }_{\mathit{\text{it}}}$$ where ${\mathit{\text{ZW}}}_{\mathit{\text{it}}}$ denotes the stability efficiency scores of bank (see, Section 3.2.1); Competition (see, Section 3.2.2); $\mathrm{b}{\mathit{\text{ank charactristics}}}_{\mathit{\text{it}}}$ and ${\mathit{\text{macroeconomic}}}_{\mathit{\text{it}}}$ are the bank-specific features and macroeconomic factors, respectively. The subscript j represents a bank (j = 1,2, …, 25) while t, ε denote the time trend and the error term, respectively.

The descriptive statistics for the variables are presented in Table 2. The mean (standard deviation) of bank stability (ZW) is 0.318 (0.173). Besides, the key independent variable, CMP has a mean value of 0.692, and a standard deviation of 0.05 while AMP is 0.801 (0.038). In addition, the correlation matrix of all variables is displayed in Table 3. These results confirm that there are no multicollinearity concerns in this study.

Table 2. The descriptive statistics of variables.

Variables	Definitions	Obs	Mean	Std. Dev	Min	Max
ZW	The stability efficiency scores	332	0.318	0.173	0.033	0.895
CMP	Conventional Lerner index	339	0.692	0.058	0.267	0.892
AMP	Efficiency-adjusted Lerner index	339	0.801	0.038	0.506	0.930
SIZE	The natural logarithm of total assets	342	32.227	1.313	28.420	34.989
ROA	Returns on assets	341	0.909	0.805	−5.512	5.952
LATA	Liquid assets to total assets	339	0.322	0.124	0.049	0.816
CAP	Total equity to total assets	339	0.094	0.052	0.021	0.462
GDP	The growth rate	350	0.060	0.011	0.029	0.071
INF	The annual inflation rate	350	0.073	0.061	0.006	0.231

Source: Author’s summary.

Table 3. The correlation matrix of variables.

	ZW	CMP	AMP	SIZE	ROA	LATA	CAP	GDP	INF
ZW	1
CMP	−0.256	1
AMP	−0.256	0.994	1
SIZE	−0.191	−0.243	−0.162	1
ROA	−0.050	0.680	0.695	−0.095	1
LATA	0.059	0.331	0.323	−0.277	0.299	1
CAP	0.069	0.372	0.328	−0.649	0.440	0.209	1
GDP	0.031	0.125	0.127	0.008	−0.069	−0.003	−0.039	1
INF	0.258	−0.099	−0.116	−0.374	0.269	0.243	0.383	−0.107	1

Source: Author’s summary.

The table shows the correlation coefficients of all variables employed in the main equation.

5. Empirical results

5.1. Competition and bank risk-taking

Table 4 illustrates the effect of Competition on bank stability efficiency over the period 2007–2020, using the Tobit regression.

Table 4. Baseline regression results of competition on Vietnamese bank stability efficiency.

Variable	ZW	ZW
CMP	−3.354***
	(0.564)
AMP		−5.088***
		(0.886)
SIZE	0.078***	0.065***
	(0.021)	(0.020)
ROA	0.158***	0.159***
	(0.043)	(0.044)
LATA	0.339**	0.344**
	(0.150)	(0.152)
CAP	−0.829	−0.833
	(0.462)	(0.464)
GDP	3.823*	3.87*
	(1.982)	(1.971)
INF	−0.395	−0.334
	(0.346)	(0.345)
Constant	4.731	6.041
	(0.947)	(1.144)
No.of Obs	328	328
Pseudo R2	0.246	0.231

Note: This table presents the results of main equation. Robust standard errors are in parentheses. *, ** and ***Significant at 10, 5 and 1% levels, respectively.

Our analysis focuses on the key variables of interest, specifically the Lerner index. Our findings indicate a statistically significant and positive relationship between the Lerner index and bank stability efficiency. The findings indicate that the competition-stability relationship hypothesis suggests that a rise in competition promotes the financial stability of banks. This finding provides evidence in favor of hypothesis H1B. The findings presented align with the outcomes reported by Jeon and Lim (2013) in Korea, Akins et al. (2016) in the United States, Maji and Hazarika (2018) in India, and Guo and Chai (2024); Hussain and Bashir (2020) Li and Peng (2024) in China, Thanh et al. (2024) in Vietnam. This can be explained away with several reasons. First, building on the research conducted by Singla and Singh (2019) increased competition can mitigate agency issues, resulting in banks having less motivation to engage in overall risk-taking. Following Thanh et al. (2024) the study shows that in Vietnam’s banking industry, dominated by the big four state-owned commercial banks (SOCBs), the principal-agent issue remains persistent and pronounced. Throughout the 2000s, a considerable portion of the lending by SOCBs was directed toward state-owned enterprises, which are notorious for their inefficient use of capital. This led to an increase in non-performing loans originated by these public entities. Therefore, fostering competition and mitigating agency issues within Vietnam’s banking sector, particularly among SOCBs, could enhance the stability of banks. Second, there is a correlation between the increase in competition and the reduction of bank lending rates. In this context, the decrease in borrowing costs is likely to enhance the profitability of the borrower’s investments, thereby increasing their ability to repay the bank (Li & Peng, 2024).

For bank-specific features, the significant positive effect of bank size (SIZE) on Vietnamese bank stability suggests that larger banks are more diversified and have a greater degree of risk management skill, making them less dangerous than smaller banks (Tabak et al. (2012). The results also report that ROA is significantly and positively related to bank stability efficiency, implying that Vietnamese bank with higher level of stability efficiency have higher profitability. This conclusion accords with the finding of Fang et al. (2019) who suggested a strong correlation exists between these variables. The ratio of liquidity risk (LATA) is shown to be positively linked to the stability efficiency of Vietnamese banks, indicating that banks which are liquid are more stable. This can be explained by the fact that the high value of liquid assets allows banks to overcome any acute challenges caused by unanticipated cash withdrawals. Özşuca and Akbostancı (2016) and Ghenimi et al. (2017) who confirm that strong, liquid and well-capitalized banks are less risk prone and more stable.

Regarding macroeconomic factors, firstly, a positive effect of GDP on the efficiency of bank stability strengthens the conventional notion that during cyclical upswings in the economy, there is a rise in demand for banks’ services and products, resulting in more bank stability. This is in accordance with (Le, 2020). Lastly, INF corresponds favorably with bank stability efficiency, indicating that banks are able to effectively foresee inflation and adjust interest rates. Consequently, Vietnamese banks may be able to boost their earnings and stability (Le, 2021).

5.2. Robustness check

To further validate our conclusion, we have conducted some additional tests to check the robustness of our main finding.

First, in contrast to the financial systems of developed countries, Vietnam’s banking sector is dominated by four state-owned banks (also called big 4 banks). Consequently, it is essential to determine the sensitivity of our results by eliminating state-owned banks and including only private banks (PRIVATE) in the sample. The results given in Table 5 are qualitatively similar to our baseline findings, demonstrating that the inclusion of the four largest banks did not distort our baseline results.

Table 5. The results of considering the private banks and crisis.

Variable	PRIVATE		CRISIS
	ZW	ZW	ZW	ZW
CMP	−3.023***		−2.047***
	(0.792)		(0.339)
AMP		−4.061***		−2.967***
		(1.172)		(0.527)
SIZE	−0.053**	−0.042*	−0.042***	−0.032**
	(0.026)	(0.025)	(0.013)	(0.012)
ROA	0.147***	0.131**	0.085***	0.078***
	(0.053)	(0.052)	(0.027)	(0.027)
LATA	0.318**	0.315*	0.135	0.136
	(0.160)	(0.162)	(0.091)	(0.092)
CAP	−0.814	−0.839	−0.282	−0.257
	(0.507)	(0.513)	(0.295)	(0.297)
GDP	2.986	2.737	2.381**	2.316**
	(2.196)	(2.192)	(1.102)	(1.108)
INF	−0.117	0.101	−0.070	−0.033
	(0.430)	(0.429)	(0.226)	(0.229)
GFC			0.073	0.035
			(0.047)	(0.141)
Constant	3.726	4.573	2.851	3.470
	(1.163)	(1.417)	(0.559)	(0.658)
No.of Obs	276	276	328	328
Pseudo R2	0.140	0.126	1.157	1.090

Note: This table presents the results of main equation. Robust standard errors are in parentheses. *, ** and *** Significant at 10, 5 and 1% levels, respectively.

Second, following Fu et al. (2015) and Le et al. (2020), and among others, $\mathit{\text{CRISIS}}$, a dummy variable that takes a value of 1 for the period 2007–2009 and 0 otherwise, is included in the model to control for the impact of the recent global financial crisis. The results in Table 5 (CRISIS). The study reveals that the Crisis does not have a statistically significant effect on the bank stability efficiency of Vietnamese banks, aligning with the findings of Nguyen et al. (2014) regarding Vietnamese banks. However, our primary discoveries have been validated.

Third, in accordance with the studies conducted by Berger and Bouwman (2009) and Le (2019), the categorization of large and small banks is based on their total assets, with those above or below the median being classified accordingly. The findings are displayed in Table 6. The results indicate that there is still a correlation between competition and the efficiency of bank stability, which remains applicable to both small and large banks. Furthermore, our primary findings remain consistent.

Table 6. The results of considering the subsamples.

Variable	Small banks		Large banks
	ZW	ZW	ZW	ZW
CMP	−2.001*		−2.972***
	(1.200)		(0.788)
AMP		−2.157*		−5.071***
		(1.247)		(1.259)
SIZE	−0.069	−0.050	−0.220***	−0.209***
	(0.046)	(0.032)	(0.062)	(0.059)
ROA	0.134*	0.107**	−0.029	0.001
	(0.072)	(0.053)	(0.094)	(0.093)
LATA	0.557**	0.451**	−0.473	−0.448
	(0.243)	(0.177)	(0.293)	(0.285)
CAP	−0.759	−0.510	−6.001	−5.711
	(0.599)	(0.430)	(1.854)	(1.796)
GDP	0.305	−0.354	8.653***	8.763***
	(3.797)	(2.605)	(2.986)	(2.914)
INF	0.146	0.025	−1.213	−1.241
	(0.508)	(0.350)	(0.708)	(0.685)
Constant	3.758	3.469	9.636	11.233
	(1.972)	(1.667)	(2.385)	(2.595)
No.of Obs	165	165	163	163
Pseudo R2	0.226	0.235	0.700	0.719

Note: This table presents the results of main equation. Robust standard errors are in parentheses. *, ** and *** Significant at 10, 5 and 1% levels, respectively.

Fourth, to avoid the endogeneity issue that could emerge between the macroeconomic conditions and bank characteristics and stability efficiency (Adesina, 2019). Using the system GMM recommended by Arellano and Bover (1995) potential bias resulting from unobserved heterogeneity and endogeneity is mitigated. We use the system GMM with a one-year lag of all explanatory variables to re-estimate the model. Table 7 presents the outcomes of the application of system GMM. In all of the models, the Hansen test p values are statistically not significant, suggesting that there are no over-identifying constraints. Alternatively, the instruments are valid and all conditions for the moments are met. Furthermore, since the p value for AR1 is statistically significant, the hypothesis that there is no first-order autocorrelation between the first residual differences is rejected. The p values of AR2 in our models are statistically insignificant, so this does not necessarily imply inaccurate estimations (Arellano & Bond, 1991). All of them, the prerequisites are satisfied. Nevertheless, the outcomes displayed in Table 7 support the findings we made previously.

Table 7. The results of GMM, NPL and Z-scores.

Variables	ZW	ZW	ZSCORE	ZSCORE	NPL	NPL
πt-1	0.993***	0.992***	0.163***	0.161***	0.323***	0.215***
	(0.006)	(0.006)	(0.018)	(0.018)	(0.099)	(0.062)
CMP	−0.100***		−7.915***		15.235***
	(0.022)		(1.673)		(5.151)
AMP		−0.139***		−11.895***		19.700**
		(0.033)		(2.510)		(7.573)
SIZE	−0.002***	−0.002***	−0.035	−0.038	0.300	−0.009
	(0.001)	(0.001)	(0.047)	(0.051)	(0.268)	(0.253)
ROA	0.005***	0.005***	1.609***	1.610***	−1.094***	−1.010***
	(0.001)	(0.001)	(0.072)	(0.070)	(0.292)	(0.198)
LATA	−0.007	−0.007	−0.372	−0.351	1.951	2.626
	(0.004)	(0.005)	(0.375)	(0.392)	(1.247)	(1.163)
CAP	−0.046	−0.045	−3.093	−3.066	19.252	14.382
	(0.017)	(0.018)	(0.703)	(0.670)	(3.496)	(2.1362)
GDP	0.090***	0.081***	5.154***	4.720**	−31.333***	−24.080***
	(0.014)	(0.014)	(1.873)	(1.852)	(4.044)	(6.420)
INF	−0.030***	−0.029***	−0.017	−0.165	6.902***	5.206***
	(0.007)	(0.008)	(0.513)	(0.502)	(1.154)	(0.924)
Constant	0.132	0.177	6.827	11.021	−18.810	−13.977
	(0.033)	(0.044)	(1.752)	(2.575)	(10.563)	(11.014)
AR1 (p value)	0.044	0.038	0.038	0.039	0.021	0.084
AR2 (p value)	0.108	0.148	0.175	0.169	0.136	0.100
Hansen test (p value)	0.786	0.750	0.716	0.732	0.575	0.740
No.of Obs	299	299	317	317	274	274
No.of groups	25	25	25	25	25	25

Note: This table presents the results of main equation. Robust standard errors are in parentheses. *, ** and *** Significant at 10, 5 and 1% levels, respectively.

Finally, we replace the dependent variable (ZW) by employing the Z-score (ZSCORE) as a direct measure of banking soundness and an inverse measure of risk. We further consider another alternative measure of risk indicator, the nonperforming loan (NPL) ratio. The results shown in Table 7 of models (ZSCORE) and (NPL) indicate that the presence of competition has a significant impact on the bank’s particular risk.

6. Conclusions

This article evaluates the effects of competition on the stability and efficiency of Vietnamese banks, employing the stochastic frontier approach. It uses data from 25 Vietnamese commercial banks for the period 2007–2020. The Tobit regression method is employed to estimate the nexus between competition and the stability efficiency of Vietnamese banks. Initially, the study shows that an increase in competition fosters bank stability efficiency in Vietnam over the analysis period. It further explores the relationship between competition and the stability efficiency of Vietnamese banks under the influence of the financial crisis, concluding that the findings remain valid despite this factor. Additionally, by using a variety of alternative risk indicators and excluding state-owned banks from the sample, the study re-examines the competition and bank stability efficiency nexus, finding the results to be robust. The findings are strongly robust across a variety of tests. The discussion encompasses multiple points. First, drawing on Singla and Singh (2019), it’s noted that heightened competition can alleviate agency problems, reducing banks’ propensity for risk-taking. Thanh et al. (2024) further reveal that in Vietnam’s banking sector, heavily influenced by the four major SOCBs, the principal-agent dilemma is notably significant. A substantial part of SOCBs’ loans during the 2000s targeted state-owned enterprises known for capital mismanagement, leading to an upsurge in non-performing loans. Hence, increasing competition and addressing agency issues, especially within SOCBs, might improve bank stability. Second, Li and Peng (2024) highlight a direct relationship between growing competition and decreased lending rates. Lower borrowing costs could, therefore, boost borrowers’ investment returns, improving their repayment capability to banks.

Our research provides valuable insights into the literature regarding the banking sector and risk management. Notably, our study breaks new ground by exploring the impact of competition on bank stability efficiency through a stochastic frontier approach. The findings of this article provide some policy recommendations for Vietnam’s banking industry. Initially, additional liberalisation measures could be implemented to improve bank stability. For example, the State Bank of Vietnam (SBV) may explore raising the maximum proportion of bank shares that foreign investors can acquire. This adjustment could increase competitiveness and boost the sector’s strength. Furthermore, regulatory agencies may consider lowering entrance barriers for both new local and international banks, thereby increasing competition and variety in the market. Second, policymakers should prioritise strengthening transparency, accountability and internal controls in banking organisations. To achieve this purpose, detailed and enforceable norms outlining the duties and obligations of bank management and boards of directors must be implemented. It also includes encouraging open communication between banks, regulatory authorities and the general public. To summarise, by implementing these policy recommendations, policymakers can work to create a more stable and competitive banking system in Vietnam, one that is prepared to confront future economic challenges and contribute to the country’s long-term growth and development.

Nevertheless, this study is constrained by its dataset, which exclusively comprises Vietnamese banks and is limited to a relatively brief timeframe. Additional research should be conducted to investigate this correlation in other developing nations in order to enhance our comprehension of the prevailing circumstances. Furthermore, it is possible that there is a non-linear correlation between competition and the efficiency of bank stability. Hence, it is imperative for researchers to conduct further investigations in order to ascertain the precise threshold of competition that is necessary to uphold the long-term stability of banks.

Credit authorship contribution statement

Dat T. Nguyen: Conceptualization, Data collection, Writing – original draft, Project administration; Tu Le: Formal analysis, Methodology, Writing – original draft; Son Tran: Visualization, Supervision, Validation, Writing – review & editing.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Funding

This work was supported by University of Economics and Law, Ho Chi Minh City, Vietnam and Vietnam National University, Ho Chi Minh City, Vietnam.

Notes on contributors

Dat T. Nguyen

Dat T. Nguyen is a researcher at the Institute for Development & Research in Banking Technology, University of Economics and Law, Vietnam. His work focuses on econometrics in banking and finance.

Tu DQ. Le

Dr. Tu Le is a researcher at the Institute for Development & Research in Banking Technology, University of Economics and Law, Vietnam.

Son Tran

Assoc. Prof. Son H Tran is currently working for the University of Economics and Law, Vietnam National University, Ho Chi Minh, and the Institute for Development and Research in Banking Technology.

Notes

1 Since the data is sourced from publicly available platforms and is already published, no additional permission is required to use it for research purposes.

2 The technical trend variable ranges from 1 to 14, for the period 2007– 2020, respectively.