Liquidity creation and bank risk-taking: Evidence from a transition market

Abstract

This study contributes to the banking literature by examining the effect of bank liquidity creation on bank risk-taking behaviors in Vietnam - a transition economy. Our data sample comprises 367 observations of 33 Vietnamese commercial banks from 2009 to 2020. We employ the Bias-corrected Least-Squares with Dummy Variables (LSDVC) estimation, which performs better than other dynamic estimators in small and unbalanced panel samples. In this research, bank risk primarily represents non-performing loans (NPLs). Empirical results show that bank liquidity creation significantly reduces NPLs. Otherwise, bank funding diversification significantly increases NPLs in Vietnamese commercial banks. Our findings are robust to alternative measurements of liquidity creation and bank risk. Additionally, we show the moderating role of bank scale in the effects of liquidity creation and funding diversification on bank risk-taking in Vietnamese banks. Our paper is the first research investigating the influence of liquidity creation on bank risk-taking in the specified situation of a transition economy. Besides, it provides empirical evidence to fill the existing research gap. Further, this study provides a list of implications for bank managers and policymakers to manage credit risks and improve the stability of the Vietnamese banking system.

1. Introduction

In the 1980s and the early 1990s, Asian countries undertook a shift toward market-based economies [1]. At that time, the transition economies triggered the implementation of economic liberalization policies. The primary goal of financial liberalization was to turn over government-owned banking systems to private ownership. Due to participation in the World Trade Organization (WTO) and their freedom to conduct business in the banking industry, foreign-owned banks have grown in number since April 2007 [2]. Additionally, this growth increased the efficiency and competitiveness of domestic financial institutions. However, competition among banks encourages risk-taking behaviors that are taking place, such as subprime lending [3] or using more significant leverage to increase profitability [4]. Vietnam's economy is also in this typical flow. State-owned banks in Vietnam reserve their primary duties for facilitating monetary policies, notwithstanding the Vietnamese government's empowerment of the economic change process. In the integrated context, the Basel II convention has been implemented in Vietnam since 2010 to increase lending and establish the capital adequacy ratio (CAR), supporting the stability of the global banking sector. Basel II is a vital and unavoidable trend for Vietnamese commercial banks as they integrate into the global financial system.

[5] suppose bank risk-taking behavior is associated with policies that increase risk through any channels. Credit risk is one of the most obvious risks to manage in the banking industry among the proxy indicators of bank risk-taking activities [6]. After the most recent global financial crisis (GFC), much research has been done on the fast-rising number of non-performing loans [7,8]. Banks must weigh the benefits of loosening credit restrictions against their ability to remain solvent [9]. From July 15, 2021, to November 30, 2021, sixteen Vietnamese commercial banks are thought to have accrued over VND 18,095 billion in reduced interest, according to the State Bank of Vietnam's (SBV) report in 2021. Still, the profit goals motivate banks to create more liquidity. That might result in higher bank risk.

In financial liberalization, liquidity creation and bank diversification have been crucial matters in the banking industry. A process generated by banks that add funds to pay their debt obligations and extend credits to borrowers is called "bank liquidity creation.” This is often accomplished through various operations, including deposit-taking, lending, and investing, thereby effectively creating economic liquidity. Overall, two considerable activities could create liquidity in banks. Firstly, banks finance their illiquid assets with liquid liabilities to generate customer liquidity [10]. Secondly, banks could perform off-balance sheet activities, such as loan commitments, to create customer liquidity [11,12]. In other words, banks create liquidity by performing off-balance sheet activities. Liquidity creation (hereafter LC) is one of the essential functions of the banking system to promote economic development. Banks with highly capitalized assets tend to transfer their liquidity into non-traditional activities [13]. As a result, banks diversify their income sources toward fee-based income instead of depending on interest-based income, leading to reduced capital requirements [14]. [11] shows that engaging in more non-traditional activities reduces banks' liquidity.

A diversified portfolio for banks is considered value-enhancing and does not increase risk exposure [15]. By engaging in non-traditional segments, banks gain more customer information, better monitor loans, and reduce systemic risk. Conversely [16,17], find that agency problems and a potential loss of focus associated with income diversification into non-interest businesses might result in declining loan quality. Accordingly, bank managers cannot cover all the non-interest investments because of a more convoluting organization and product structure. Another strand of research on diversification has focused on bank funding diversification [18]. His study indicates that banks with diversified funding sources tend to join in more risky activities to gain higher profitability. Besides, his finding suggests that activity diversification is suitable for Vietnamese banks because the banking sector still lacks management and experience capacities compared to other markets [19]. [18] has not demonstrated the nexus between funding diversification and non-performing loans (hereafter NPLs). Thus, it is worth investigating whether bank liquidity creation (LC) and bank funding diversification (BFD) affect credit risks measured by NPLs in Vietnamese banks covering 2009 to 2020.

After the 2008 GFC, Vietnam encountered many economic consequences, such as high inflation and economic slowdown. Therefore, to avoid the adverse impact of the economic recession, the SBV issued policies to stimulate demand in 2008. These policies encourage commercial banks to provide credit to domestic enterprises.1 Studies have been undertaken to determine factors influencing Vietnamese bank risk in place of NPLs [4,7,20]. The Vietnamese market has unique characteristics compared to mature banking systems (US, European countries) or emerging markets (China, MENA). The domination of a few large government-owned banks leaves others with a much smaller market share [17]. [21] report that government projects might only be financed through state-owned rather than private banks. Local banks encounter rising competition from foreign banks derived from the WTO agreements. Consequently, banks are encouraged to lower their lending criteria and take risks at higher levels to attract customers and gain additional profits to meet the earnings objective of management. Under fierce competition, controlling NPLs and funding diversification is necessary to investigate the operations of the banking system. Further, the Vietnamese economy and financial market have witnessed significant structural changes during this period. The SBV conducted a comprehensive restructuring process and aggressive actions such as M&A activities to address the high levels of NPLs from 2012 to 2016. According to Credit Suisse's publication, the adjusted data raised NPLs to 12% at the end of 2012 and decreased to 10.1% at the end of 2016, according to the Economist Intelligence Unit. Hence, our study expects to discover the exciting effects of liquidity creation and funding diversification on bank risk-taking in the post-GFC period (2009–2020), where NPLs represent bank risk-taking.

Applying the method of [10], many papers focus on the liquidity creation of Vietnamese commercial banks and provide exciting findings. Still, these studies leave a gap in investigating the effect of bank liquidity creation on NPLs in the specific context of Vietnam. The nexus between monetary policy and liquidity creation is explored in the papers of [22,23]. They find more liquidity amid easing monetary policy differences in liquidity creation among different bank sizes [11]. provides precious evidence of the effects of LC on bank capital and bank activities, respectively [24]. emphasize the role of market power in creating liquidity in Vietnamese banks. Accordingly, monetary policy transmission through the bank liquidity channel is weakened as market power increases. To the best of our knowledge, there is no research on the effect of bank liquidity creation (LC) on the credit risk of the Vietnamese banking system measured by NLPs. In addition, instead of predicting that fund-diversified banks take higher risks due to participating in high-risk activities [18], has yet to explore the effect of bank funding diversification on NPLs. From a liquidity creation perspective, our research strives to fill gaps in existing studies on the Vietnamese banking system.

This paper uses the Bias-corrected Least-Squares with Dummy Variables (LSDVC) estimator proposed by Ref. [25] to analyze a data sample of 33 commercial banks covering the period of 2009–2020. LSDVC estimates remarkably well in our research owing to the small number of 33 Vietnamese banks and the 12-year period [23,26,27]. According to Refs. [7,11,18], we set up dynamic models with a pool of variables: Liquidity creation (LC); Bank funding diversity (BFD); Non-performing loans (NPL); Bank size (SIZE); Return on assets (ROA); Bank capital structure (CAP); Loan growth (LG); Inflation rate (INF); GDP growth rate (GDP). Estimated results show that bank liquidity creation negatively while bank funding diversity positively affects NPLs. Our findings are consistent with [6] that banks could boost economic growth by creating liquidity, thereby improving borrowers' repayment ability. Said another way, banks could diminish NPLs by increasing liquidity. On the contrary, maintaining diversified funding sources motivates banks' managers to participate in risky credit activities. Our findings are consistent with [18] that highly diversified banks tend to take higher risks. Finally, empirical results are robust to alternative proxies of risk-taking behaviors such as Z-score and RROE [17,18] and alternative liquidity creation proxies following [10].

This research contributes empirical evidence of a transition market to the banking literature. Firstly, a significant adverse effect of bank liquidity creation on NPLs is shown in Vietnamese commercial banks. Interestingly, our findings are robust to Ref. [6] paper on MENA markets and consistent with the "economic-enhancing" view, which implies that bank liquidity creation gives borrowers a higher repayment capacity due to excellent repayment ability, thereby enhancing economic development. Likewise, funding sources strongly affect commercial banks' profitability and lending volume. Secondly, a positive association between BFD and NPLs, suggests that BFD mitigates credit quality. This result supports [18] finding that banks with diversified funding tend to undertake more bank risks (Z-score). Moreover, we show that large banks have more advantages of liquidity creation than small banks. Thus, liquidity creation significantly alleviates NPLs in large banks compared to small banks.

Upon empirical results from this research, policymakers and bank managers could apply appropriate strategies to manage credit risks and stabilize the financial systems. Empirical evidence demonstrates that funding diversification enhances bank credit risks while creating more liquidity reduces bank credit risks. Thus, regulatory agencies and bank managers could use a combination of LC and BFD to balance credit risks in Vietnamese commercial banks. Moreover, scale advantages help large Vietnamese banks utilize liquidity creation better to alleviate NPLs significantly. From the case of a transition economy, our findings provide insights for bank managers and regulators to enact appropriate policies related to liquidity creation, bank funding diversification, and NPLs to mitigate excessive risk in Vietnamese banking activities after the 2008 GFC.

This paper is designed into five parts. Part 2 synthesizes the literature review about bank risk-taking behavior, liquidity creation, funding diversification, and their relationship. Empirical models and methodology are introduced in Part 3. Estimated results and robust tests are discussed in Part 4. Part 5 gives conclusions and recommendations.

2. Literature review

2.1. Risk-taking behavior

[5] mention that banking activities involve institutions transforming liquid deposits into illiquid loans [28]. suggest that excessive liquidity induces risk-taking behavior on the part of bank managers. On the other hand, banks' excessive risk-taking is universally due to the inadequate control of moral hazards and played a central role in the events leading up to the financial crisis of 2007 [29]. This research focuses on the three hypotheses to broadly explain the factors affecting credit risks measured by NPL levels. These are the "asymmetric information" hypothesis, the "moral hazard" hypothesis, and the "adverse selection" hypothesis. The "asymmetric information" hypothesis occurs when the amount of information between two parties differs and affects transaction decisions. This disparity is owing to information asymmetry, the accuracy of the information, and the ability to convert the information. The information asymmetry affects decision-making due to the lack of knowledge [30]. Reducing asymmetric information worsens NPLs in banks [15].

The "moral hazard" and "adverse selection" hypotheses are developed from the asymmetric information hypothesis [3]. indicate two types of moral hazards. One of them points out that to maximize profit, bank managers might set lower standards for loan monitoring to increase their lending activities, thus increasing NPLs. The remaining type focuses on the risky investing portfolio of banks, which implies that bank managers might face more credit risks through inefficiently evaluating the assets. Likewise [31], report a negative association between the lagged solvency ratio and NPLs, supporting the moral hazard hypothesis. However, in the study of [8], no evidence is found, implying that the moral hazard hypothesis exists among Greece banks. The adverse selection hypothesis refers to a biased decision made due to a lack of information, which causes a deterioration in the overall credit quality due to good-financial borrowers being displaced by subprime borrowers. It eventually causes an increase in NPLs, a decline in profitability, and capital erosion [32].

2.2. Liquidity creation and non-performing loans

The 2008 GFC showed that financial institutions need to strengthen their liquidity risk management to avoid harming depositors, maintain high-performance standards, and guarantee financial stability [33]. state that increased liquidity creation reduces liquidity risk, which might have a connection with credit risk. Likewise [34], reveals three fundamental liquidity management theories used to prevent and tackle liquidity shortages: commercial loan theory, shiftability theory of liquidity, and income anticipation theory.

Firstly, the commercial loan theory elaborated by Ref. [35] warns banks from processing long-term loans and restricts their earning assets to short-term self-liquidating productive loans and real bills. The theory implies that providing loans in the short term is better to minimize the possibility of defaults that will impact the bank's performance. Bank liquidity could be guaranteed if the bank's productive assets consisting of short-term credit are disbursed in business activities running normally. However, banks will ignore the considerable returns on long-term loans and investments, slowing the economy's growth [36]. If the bank concerned will provide more extended credit, the data source should be taken from bank capital and long-term sources of funds [37].

Secondly [38], develops the shiftability theory of liquidity, which assumes that banks could hold credit instruments such as commercial papers and treasury bills to protect the liquidity level in case of massive deposit withdrawals. Interbank transfers can easily maintain liquidity when a bank is short of ready money. This also suggests that banks need not be highly liquid if they have good connections with financial institutions that could finance them when they face a liquidity shortage. Even so, the theory leaves the major drawback is that in the case of an acute crisis, the selling process from all banks will shock the market [36].

Thirdly, the income anticipation theory developed by Ref. [39] suggests that banks could reduce the risk associated with long-term loans by making monthly principal and interest payments. If the borrower's expected income is to repay these loans in periodic and regular premiums, the bank's managers could plan its liquidity based on the predicted earnings of the borrowers, enabling the bank to grant medium- and long-term loans in addition to short-term loans. This will allow the bank to provide high liquidity when the cash inflows are regular and predictable.

Existing literature on the effect of liquidity creation on NPLs is rare. Some papers examine the effect of NPLs (or credit risk) on liquidity creation [11,12,40]. Other studies consider liquidity creation as a determinant of market liquidity [9] or systemic risk [41]. The causal link between economic growth and the ability to repay the debts of borrowers is shown by many studies [7,8,31,42]. [7] report that a growing economy positively impacts enterprises' production, business, and import-export activities. Thus, a growing economy also increases the loan repayment capability of enterprises. Conversely, a slow-moving economy could diminish bank asset quality [43] and affect borrowers' repayment ability [8]. [44] show that higher LC leads to more bank failures in developed countries. This could be attributed to the excessive liquidity risk these banks take as a proportion of their total assets [6]. indicates that LC negatively affects NPLs in MENA banks. The author explains that higher liquidity increases the economic situation and benefits borrowers' capacity to repay their debt. Hence, debt repayment directly raises loan quality, leading to decreasing NPLs. This research also argues that banks might effectively monitor and gain more experience in credit-screening processes when they increase their LC. With time, bank managers could screen out reputable borrowers who will repay their debts in maturity. As a result, they have more competence in managing credit risk, as measured by the NPLs [40]. showed an insignificant relationship between bank liquidity creation and NPLs.

[6] points out two opposite views to examine the connection between bank liquidity creation and NPLs. The first view shows that bank liquidity creation positively impacts NPLs. According to the Klein-Monti model and its extensions, banks also pursue considered profit-maximizing goals [45,46]. Banks widen their spread between loans and deposit rates to gain their targets. However, banks might face a loan portfolio default risk, which increases liquidity risk since it lowers the amount of cash inflow [6]. Similarly [33], argues that banks increase their liquidity creation when they take higher risks through business loans and commitments. During this process, banks take higher liquidity risks by rendering themselves illiquid. Besides, banks impede their selection of "good" borrowers due to ineffective screening processes, causing an increase in NPLs. Furthermore, according to the financial intermediation theory by Refs. [47,48], banks' asset and liability sides are interrelated. Risky assets and excessive fund withdrawals trigger several bank runs. Banks face more liquidity risk through their transformation function. By transferring liabilities to assets, banks hold illiquid assets while providing liquidity to borrowers and fulfilling liquid claims to depositors.

Thus, an increase in LC causes more risks and illiquid assets related to a rise in NPLs. Moreover, bank managers might be encouraged to set low standards to generate more loans and create financial liquidity [28]. The Trade-Off theory (TOT) acknowledges the potential for risk-shifting behavior. This refers to the potential for banks to engage in riskier lending strategies to maximize their earnings. If banks take on excessive risk, it could result in a higher proportion of loans transferring to NPLs if borrowers are unable to repay their debts. Hence, creating liquidity might encourage banks to take unwarranted risks, increasing NPLs.

In the second view [6], points out the negative influence of LC on NPLs in banks. The author states that banks promote economic growth by providing liquidity to corporations and individuals. By creating liquidity, banks diminish the cost of external funds to businesses, thus improving the economy's growth [49]. report that Russian banks enhance economic growth by creating more liquidity. In this paper, we propose hypothesis 1 to test the impact of bank liquidity creation on NPLs in the specific context of the Vietnamese market. Examining hypothesis 1 enriches the current literature on the banking industry from a transition market perspective.

Hypothesis 1

Liquidity creation (LC) has a negative effect on non-performing loans.

2.3. Bank funding diversity and non-performing loans

Bank is one of the financial institution types, and its capital structure is more distinctive than non-financial companies [50,51]. Bank's liabilities originate from customers' deposits, issued bonds, central bank debts, and interbank debts. Therefore, deposits play a crucial role in the context of a bank's bankruptcy cost. The higher the number of deposits, the greater the potential bankruptcy costs. When a bank faces more financial distress, the repayment of customer deposits becomes a significant concern. The TOT maintains the assumptions of market efficiency and symmetric information and extends determining an optimal capital structure through the addition of various factors, such as taxes, expenses associated with financial distress, and agency costs. For instance, if a bank decides to take on more risks to boost profitability, it might need higher capital levels to mitigate potential losses. However, having too much equity stipulates free cash flow and limits the bank's ability to generate returns for its shareholders, thus causing agency cost problems [52]. On the other hand, if banks prioritize debt financing without considering the associated risks and costs, leading to an increase in the NPL volume and conflicts of interest between managers and bondholders [53]. Managers might perceive debts as raising bankruptcy costs owing to bank failures in long-term lending.

[54] document that having strong funding sources could help banks survive during the crisis to the data of the US and European banks in the pre-crisis and afterward. By diversifying their funding sources, banks could reduce dependence on a single type of external financing, enhancing their overall financial stability. On the other hand [55], argues that increased bank asset liquidity belittles stability for the banking system during financial crises, not regular periods. Contrary to previously published studies [18], confirms that Vietnamese commercial banks with highly diversified funds tend to have higher risk-taking behaviors. Firstly, increased diversification of funding draws a parallel with a low liquidity risk or high funding liquidity. Secondly, funding diversity could involve a mix of funding instruments with varying maturities and characteristics. Banks might have trouble fulfilling their financing fund obligations if they do not appropriately manage the maturities of their assets and liabilities. Furthermore, banks might fail to adequately monitor and control the associated funding risks, such as interest rate or foreign currency risks. Such problems arise if a bank has longer-term loans or investments covered by short-term deposits, exerting pressure on liquidity and possibly causing an increase in NPLs. Along the same lines [28], reports that stable funding motivates banks to take more risks. When banks attract extensive customer deposits, they reduce funding liquidity risks, so managers engage in more risk activities [28]. [56] state that risk-averse managers in banks tend to chase aggressive lending strategies when they have a copious source of liquidity to gain more compensation. Thus, we propose the second hypothesis to examine the effect of funding diversification on NPLs, based on inferences from Ref. [18] study in Vietnam.

Hypothesis 2

Bank funding diversification (BFD) positively affects non-performing loans.

2.4. Moderating role of bank size in the effects of liquidity creation and bank funding diversification on non-performing loans

The influence of liquidity creation created by different banks' sizes on these banks' stability might diverge [44,57]. In the presence of economies of scale, larger banks benefit from lower costs and higher profitability [1,17]. These banks often have better investment and diversification opportunities, which could facilitate higher LC and thus decrease bank risks [34]. states that the development of bank size with higher LC could lead to a decrease in NPLs and an increase in asset value, which contribute to the beneficial impact of LC on bank stability. According to the study of [58], the signal theory implies that more reputable banks tend to signal their financial stability and dependability to the market by providing liquidity or maintaining high historical amounts of liquidity. Firstly, due to their stable and open business practices, these banks will be less vulnerable to liquidity risk with the steady increase of depositors. Secondly, by offering stable and ample liquidity, these banks establish a strong reputation that enables them to draw in more qualified customers with sound financial statements than their rivals, which lowers the cost of monitoring loans and the number of NPLs. Thirdly, this reputation signaling encourages more excellent LC as firms strive to maintain their favorable image and easy access to capital.

Under the view of the shiftability theory of liquidity, larger banks tend to have more diversified funding decisions. These banks often have more government support and easy access to interbank or capital markets [34]. finds that the increase in the bank scale leads to low liquidity levels. Similarly, rapid scale upgrades and expansions exceed a bank's actual capability and lead to a decline in asset quality and an increase in NPLs [1]. The results also support the too big to fail (TBTF) theory which implies that large banks consider themselves too important to fail, so they need not increase their liquidity level. The fall in the stability of large banks due to increased LC also provides evidence of the moral hazard issues of excessive risk-taking associated with their perceived TBTF position [44]. [1] revealed an exciting result: Vietnamese non-listed banks outperformed listed banks from 2008 to 2016. Listed banks with more considerable and more diversifying funds might also incur higher agency costs associated with relatively higher information asymmetry between managers and shareholders, resulting in operating inefficiency. In terms of a transition market, we expect the impact of liquidity creation and funding diversification could be considerably controlled by bank size.

Hypothesis 3

The relationships between bank risk and liquidity creation and bank funding diversification are moderated by bank scale.

3. Data and methodology

3.1. Sample

To measure LC and BFD variables, we manually collect data from the annual reports of each bank because the detailed bank-level information is not accessible from public databases in the Vietnam market. Our unbalanced panel data includes information from audited financial statements for 33 domestic commercial banks between 2009 and 2020. The list of 33 banks includes 31 commercial banks (www.sbv.gov.vn) and two merged banks (Ocean Bank and CBbank). We also exclude foreign banks in Vietnam, which might not reflect the operation of banking in Vietnam. Additionally, the research period lasts from 2009 to 2020 because we could collect sufficient data for this study within this span. The global banking industry has recovered since 2009, and its recovery ended in 2020 due to the COVID-19 pandemic. Thus, the impact of external shocks on the data during this period was minimal. All financial variables are winsorized at the 1% and 99% levels to mitigate outlier issues [10]. We also remove observations without sufficient information to calculate the required variables [59]. Our final sample includes 367 annual observations from 33 Vietnamese commercial banks. The annual Inflation and GDP growth rates are collected from the World Bank database in the period 2009–2020.

3.2. Empirical models

According to Refs. [6,18,23], we construct the dynamic model (1), which investigates the influence of liquidity creation (LC1, LC2) and funding diversification (BFD) on NPLs. Firstly, we use the Generalized Method of Moments (GMM) estimation to address the endogeneity and heteroskedasticity problems owing to the inherent existence of these issues in dynamic models. GMM estimations are not valid if the instruments outnumber the groups which are not appreciated to the rule of thumb suggested by Ref. [60]. This scenario might happen if applying the GMM method for the small and unbalanced panel sample; thus, using the LSDVC (Bias-corrected Least-Squares with Dummy Variables) technique developed by Refs. [61,62] and extended by Ref. [25] could overcome this matter [23,26,27]. With our sample, LSDVC estimation is expected to be more suitable than GMM estimations regarding bias and root mean squared error [25]. Additionally, according to Ref. [63], the persistence of the financial variables—the main reason for serial correlation—is resolved by LSDVC estimation. Additionally, LSDVC estimation avoids unobserved heterogeneity and endogeneity caused by the dependent variable's fractional nature [64]. Many recent studies have used the LSDVC method in various research aspects of the banking industry, for example [65], on credit risk in banking systems [66]; researched Islamic banks [26]; in capital, funding liquidity, and lending application.

In this study, we prioritize using the LSDVC method initialized by AB [25,62]. The form of the dynamic model (1) is designed as below:

$${NPL}_{i,t}=\alpha +{\beta }_1{NPL}_{i,t-1}+{\beta }_2{LC}_{i,t}+{\beta }_3{BFD}_{i,t}+{\beta }_4\sum {\left(Bank-SpecificFactors\right)}_{i,t}+{\beta }_5\sum {\left(Macroeconomic\right)}_t+u_{i,t}$$ (1)

where ${NPL}_{i,t}$ is the dependent variable; ${NPL}_{i,t-1}$ is the lag value of NPL. NPL is used as a key dependent variable in this research because of three reasons. Firstly, liquidity creation is positively associated with illiquid assets [67]. The most usual form of bank illiquid assets is non-performing loans (NPLs). Secondly, bank funding diversification motivates Vietnamese banks to participate in high-risk activities. A shortcoming of the paper [18] is that it has yet to investigate the nexus between BFD and NPLs. Thirdly, the Vietnamese government has not yet enacted a bank bankruptcy law but has already issued corporate bankruptcy law in 2014. Hence, NPL is indeed an ideal proxy for bank risk. According to Refs. [17,18], we use two alternative variables of bank risk-taking to ensure the robustness of our main results (Z-score and RROE variables). Liquidity creation (LC) and Bank funding diversity (BFD) are prominent independent variables, where LC1 and LC2 are two proxies of liquidity creation (LC). Other control variables are Bank-specific factors (SIZE, ROA, CAP, and LG) and Macroeconomic variables (GDP and INF); $u_{i,t}$ is the error term for the bank (i) in the year (t). Table 1 reports detailed variables definitions. Most importantly, liquidity creation (LC) is an essential function of banks to provide economic capital. Following [10], we estimate the liquidity creation (LC) variable. The construction of liquidity creation measurement is described explicitly in Table 2. Still, bank equity could affect the ability to create liquidity [10]. To investigate the relationship between liquidity creation (LC) and NPLs and ensure that our finding is robust, we create two liquidity creation variables, including LC1 and LC2. LC1 is measured by excluding equity from LC measurement (with a weight of −1/2), as shown in Table 2, while LC2 is computed by including equity.

Table 1.

Variables definitions.

Name of variable	Symbol	Description	Calculation
A. Dependent variable
Non-performing loans Louzis et al. (2020), Hoang et al. (2020)	NPL	Non-performing loans measure the bank's risk-taking and are calculated by non-performing loans divided by total loans.	$NPL=\left(Group3loans+Group4loans+Group5loans\right)/Totalloans$ Note: according to Decision No.493/2005-QD-NHNN issued by the State Bank of Vietnam, non-performing loans include Group 3, Group 4, and Group 5 loans.
Hesse and Cihák's (2007), Amidu & Wolfe (2013), Vo (2018), Vuong & Nguyen (2020)	Z-score	Z-score is the inverse risk indicator that measures the number of standard deviations a bank's profit must fall before insolvency.	$Z-score=\frac{{ROA}_{it}+{E/A}_{it}}{\sigma ROA}$
	RROE	Risk of Return on ROE	$RROE=\frac{{ROE}_{it}}{\sigma ROE}$
B. Independent variables
Liquidity creation Berger & Bouwman (2009), Dang (2020)	LC	The ability of banks to create liquidity by transferring liquid liabilities into illiquid assets.	LC = [½ × (illiquid asset + liquid liability + illiquid OBS) + 0 × (semiliquid asset + semiliquid liability + semiliquid OBS) – ½ × (liquid asset + illiquid liability + liquid OBS)]/Total asset
Bank funding diversity Vo (2018)	BFD	The funding diversity ratio runs from zero to one, with higher values indicating greater funding diversity.	$BFD=1-\left[{\left(\frac{SE}{Fund}\right)}^2+{\left(\frac{GOV}{Fund}\right)}^2+{\left(\frac{IBD}{Fund}\right)}^2+{\left(\frac{CD}{Fund}\right)}^2+{{\left(\frac{DER}{Fund}\right)}^2+{\left(\frac{LTF}{Fund}\right)}^2+\left(\frac{OTH}{Fund}\right)}^2\right]$
Bank Specific control
Bank size Louzis et al. (2012), Hoang et al. (2020)	SIZE	Asset size, as measured by the natural logarithm of total assets.	$SIZE=\mathit{Log}\left(Totalassets\right)$
Return on asset Dao & Phan (2020), Hoang et al. (2020)	ROA	Return on assets is measured as the efficient use of assets of the bank.	$ROA=Netincome/Totalassets$
Bank capital structure Louzis et al. (2012), Hoang et al. (2020)	CAP	Bank capital (CAP), measured by the ratio of equity to total assets, shows a bank's sufficient capital status and safety and health.	$CAP=ShareholderEquity/Totalassets$
Loan growth Louzis et al. (2012), Nguyen (2014)	LG	Loan growth rate	$\mathit{LG}=\left({Totalloans}_t-{Totalloans}_{t-1}\right)/{Totalloans}_t$
Macroeconomics
Inflation rate Nkusu (2011)	INF	The yearly inflation ratio of Vietnam.	We collect the inflation rate term of Vietnam from the World Bank database
Gross Domestic Product Isik & Bolat (2016), Nkusu (2011)	GDP	The annual GDP ratio of Vietnam.	We collect Vietnam's GDP growth rate from the World Bank database.

Note: Table 1 summarizes the variables employed in this research and the related previous studies aligning with each variable. A dependent variable is the non-performing loans (NPL). The key independent variables are bank liquidity creation (LC) and bank funding diversity (BFD). Other control variables are classified into bank-specific control and macroeconomics. Bank capital structure (CAP), return on assets (ROA), bank size (SIZE), and loan growth (LG) are in group bank-specific control. The annual inflation rate (INF) and the Gross Domestic Product growth rate (GDP) are macroeconomic factors.

Table 2.

The construction of liquidity creation measurement.

Illiquid assets (Weight = 0.5)	Semi-liquid assets (Weight = 0)	Liquid assets (Weight = −0.5)
Corporation loans	Interbank loans	Cash and due from institutions
Other assets	Loans to individuals	Total securities
Liquid liabilities (Weight = 0.5)	Semiliquid liabilities (Weight = 0)	Illiquid liabilities and equity (Weight = −0.5)
Customer deposits	Deposits from institutions	Other liabilities
Trading liability	Other borrowed money	Equity
Illiquid off-balance-sheet guarantee (Weight = 0.5)	Semi-liquid off-balance sheet guarantees (Weight = 0)	Liquid off-balance sheet guarantees (Weight = −0.5)
Loan guarantee commitments	Other off-balance sheet guarantees	All derivatives
Letter of credit commitments

Note: Table 2 shows the construction of liquidity creation measurement. Berger and Bouwman (2009) developed a measure for bank liquidity creation. We also make certain adjustments, as suggested by Dang (2020). Total customer deposits are treated as liquid liabilities with a weight of 0.5, considering no differences in the ease, cost, or time required for customers to obtain liquid funds from Vietnamese banks. The fundraising business could include current accounts, savings, and time deposits.

To examine the moderating effect of bank size on the effects of bank liquidity creation and funding diversification on bank risk-taking, we estimate the dynamic model (2) by AB [25,62].

$${NPL}_{i,t}=\alpha +{\beta }_1{NPL}_{i,t-1}+{\beta }_2{LC}_{i,t}+{\beta }_3{BFD}_{i,t}+{\beta }_4{DUM*LC}_{i,t}+{\beta }_{5}DUM*{BFD}_{i,t}+{\beta }_6\sum {\left(Bank-SpecificFactors\right)}_{i,t}+{\beta }_7\sum {\left(Macroeconomic\right)}_t+u_{i,t}$$ (2)

where DUM is a dummy variable representing large banks, it denotes one if the bank size is above the median and zero otherwise. Interaction variables (DUM*LC and DUM*BFD) are created and added to the dynamic Model (2) to estimate the moderating role of bank scale in the impact of liquidity creation (LC) and bank funding diversification (BFD) on non-performing loans (NPL).

4. Empirical results and discussion

4.1. Descriptive statistics

Table 3 shows the descriptive statistics of variables from 33 Vietnamese commercial banks from 2009 to 2020. The average value of NPLs in Vietnam is 2.26%, which is relatively high compared to other countries in the region. For instance, the average NPLs in Indonesia is 1.73% [68], and China's is 1.78% [12]. The highest NPLs in Vietnam is 8.437%, belonging to the Saigon-Hanoi Joint Stock Commercial Bank in 2012. The bank with the lowest NPLs was Tien Phong Bank in 2010, with only 0.335% of total outstanding loans. The average value of LC1 is 32.19%, higher than the average value of LC2, which is 27.53%. This indicates that Vietnamese commercial banks could generate liquidity over total assets ranging from 27.53% to 32.19%.

Table 3.

Descriptive statistics.

Variable	Mean	Median	Maximum	Minimum	Std.	Obs.
NPL	0.0226	0.0202	0.0844	0.0034	0.0139	367
LC1	0.3219	0.2821	0.6291	−0.0956	0.1518	367
LC2	0.2753	0.2821	0.5571	−0.1026	0.1574	367
BFD	0.4454	0.4484	0.6933	0.1225	0.1273	367
SIZE	14.005	14.011	15.195	12.522	0.5302	367
ROA	0.0089	0.0071	0.0287	0.0001	0.0068	367
CAP	0.0944	0.0826	0.2381	0.0262	0.0433	367
LG	0.2515	0.1938	1.2459	−0.2209	0.2510	367
INF	0.0598	0.0408	0.1868	0.0063	0.0465	367
GDP	0.0596	0.0621	0.0708	0.0291	0.0108	367

Note: Table 3 reports the descriptive statistics of NPL, LC1, LC2, BDF, SIZE, ROA, CAP, LG, INF, and GDP variables. Our study includes unbalanced panel data from 367 observations in 33 Vietnamese commercial banks from 2009 to 2020.

4.2. Pearson correlation matrix

Table 4 reports the Pearson correlation matrix between variables. All values of the correlation pairs (except for LC1 and LC2) are below 0.7. So, the selection of our variables is appropriate. Furthermore, the collinearity test reported that the variance inflation factor (VIF) value between variables ranged from 1.02 to 2.62. This indicates that our sample does not have multicollinearity problems among the independent variables.

Table 4.

Pearson correlation matrix.

	NPL	LC1	LC2	BFD	SIZE	ROA	CAP	LG	INF	GDP	VIF
NPL	1.0000
LC1	−0.0144 (0.7839)	1.0000
LC2	−0.0418 (0.4252)	0.9910 (0.0000)	1.0000								1.97
BFD	0.0347 (0.5078)	−0.6478 (0.0000)	−0.6758 (0.0000)	1.0000							2.62
SIZE	−0.1890 (0.0003)	0.3811 (0.0000)	0.4570 (0.0000)	−0.4701 (0.0000)	1.0000						1.87
ROA	−0.0930 (0.0751)	−0.0794 (0.1288)	−0.1968 (0.0001)	0.3563 (0.0000)	−0.0470 (0.3696)	1.0000					1.43
CAP	0.1521 (0.0035)	−0.2695 (0.0000)	−0.3656 (0.0000)	0.3509 (0.0000)	−0.5390 (0.0000)	0.2856 (0.0000)	1.0000				1.63
LG	−0.0559 (0.2854)	−0.1647 (0.0015)	−0.1981 (0.0001)	0.2775 (0.0000)	−0.2375 (0.0000)	0.3210 (0.0000)	0.1071 (0.0403)	1.0000			1.24
INF	0.2075 (0.0001)	−0.2981 (0.0000)	−0.3177 (0.0000)	0.5181 (0.0000)	−0.2615 (0.0000)	0.2587 (0.0000)	0.2095 (0.0001)	0.0533 (0.3087)	1.0000		1.42
GDP	−0.0744 (0.1551)	0.0024 (0.9634)	0.0137 (0.7930)	0.0264 (0.6136)	−0.0121 (0.8173)	−0.0768 (0.1422)	−0.0558 (0.2868)	0.0011 (0.9825)	−0.0446 (0.3937)	1.0000	1.02
VIF		1.80		2.60	1.75	1.29	1.58	1.10	1.39	1.02

Note: Table 4 discloses the correlation between all variables and Variance inflation factor (VIF) tests for detecting the severity of multicollinearity, including LC1, LC2, BFD, SIZE, ROA, CAP, LG, INF, and GDP.

4.3. Analyzing main results

Estimated results from the dynamic model (1) by using a two-step system GMM and LSDVC estimates are reported in Columns (1) and (2) of Table 5, respectively. According to the GMM estimation results, the regression coefficient of LC1 and LC2 is −0.0107 and −0.0116, which indicates that if liquidity is created by 1%, the NPLs will fall between 1.07% and 1.16%, respectively. For the LSDVC estimation, the liquidity creation makes about a 2.41–5.06% reduction in NPLs. The LSDVC results show a positive association between bank funding diversification and NPLs. Banks with diversified funds will increase NPLs from 7.62% to 9.44%. Nonetheless, GMM estimator results indicate there is no relationship between bank funding diversification and NPLs. The GMM estimated results in Column (1) might be biased because instruments overwhelm groups, and the Sargan tests reveal the invalidity of instrumental variables. Thus, LSDVC findings are more accurate in our research.

Table 5.

Regression dynamic model (1) by using a two-step system GMM and LSDVC estimations.

Independent variables	Two-step system GMM estimation NPL is the dependent variable				LSDVC estimation NPL is the dependent variable
	(1)				(2)
	Coef.	Std Err.	Coef.	Std Err.	Coef.	Std Err.	Coef.	Std Err.
lag.NPL	0.4893***	0.0360	0.4618***	0.0299	0.0655***	0.0001	0.0650***	0.0001
LC1	−0.0107*	0.0060			−0.0507**	0.0201
LC2			−0.0116**	0.0057			−0.0241*	0.0130
BFD	0.0066	0.0080	0.0073	0.0053	0.0762***	0.0130	0.0944***	0.0122
SIZE	0.0016	0.0014	0.0010	0.0018	0.0760***	0.0238	0.0831***	0.0164
ROA	−0.4665***	0.0650	−0.4070***	0.1050	1.3302***	0.3640	1.1927***	0.2797
CAP	0.0683***	0.0147	0.0557**	0.0237	−0.1560***	0.0518	−0.1539***	0.0499
LG	−0.0015	0.0035	0.0009	0.0027	−0.0074*	0.0041	−0.0067	0.0048
INF	0.0318***	0.0080	0.0262***	0.0077	0.1771***	0.0218	0.1677***	0.0251
GDP	−0.0566***	0.0185	−0.0603***	0.0187	−0.1491	0.7325	−0.1054	0.4781
Constant	−0.0112	0.0228	−0.0019	0.0283
Observations	337		337		337		337
Instruments	59		59
Number of groups	33		33		33		33
AR (1) test	0.007		0.006
AR (2) test	0.440		0.492
Sargan test	0.287		0.275

Note: Table 5 summarizes the regression results with the NPL dependent variable by using a two-step system GMM and LSDVC estimations. Column (1) shows the GMM regressions and Column (2) reports estimated results with the LSDVC method. AR (1) and AR (2) tests show tests for correlations at lag (1) and lag (2). The Sargan test confirms the validity of instrumental variables. The LSDVC in this section, initialized by AB, is bootstrapped by 50 iterations for the standard errors (Bun & Kiviet, 2003 and Bruno, 2005). All variable definitions are displayed in Table 1. ***, **, and * indicate significance levels at 1%, 5%, and 10%, respectively.

The following analyses mainly concentrate on the results in Column (2). Negative coefficients of liquidity creation variables support Hypothesis 1, that banks improve economic growth by creating liquidity, thus positively affecting the repayment capability [6,44]. Banks with strong LC capabilities are more likely to have sufficient funds to supply liquidity to firms and households to stimulate economic growth. In times of economic stress, LC could help banks provide their support to borrowers facing temporary financial difficulties. When borrowers have access to credits when needed, they are more likely to meet their financial obligations on time, reducing the probability of loans becoming non-performing. Another explanation is the improvement of the banks' credit-screening processes. Banks gain more experience in evaluating assets and the quality of borrowers through their operation, thus reducing information asymmetry. As a result, banks have more ability to control the NPL levels efficiently. Economically, banks with effective liquidity management could be in better positions to recognize and take proactive steps to impair the possibility of loan defaults. They could allocate resources prudently, regularly monitor borrower's finance, and take proactive measures to mitigate credit risks [6]. shows that LC negatively influences NPLs in both the short and long term in the case of MENA countries. Our results also imply that LC strengthens the bank's stability in Vietnam, which coincides with [44]. Our findings imply that after the 2008 GFC, heightened LC impairs NPLs in the Vietnamese banking system. According to that, LC strengthens the financial system stability and contributes to economic growth in Vietnam.

Likewise, bank funding diversity (BFD) positively impacts NPLs. Banks with more diversification of funding sources motivate managers to undertake more risks, including increasing loan volumes to pursue profit goals [28]. An increase in equity levels allows banks to provide loans without being covered by cash deposits [34]. Another explanation is that risk-averse bank managers tend to pursue riskier strategies when their liquidity risk is low owing to the advantage of diversification [56]; thus NPLs tend to be extended. That is consistent with the TOT's predictions. In addition, our results imply that funding diversification forces managers to spend more effort managing the capital cost from different sources instead of focusing on improving asset quality, including screening and monitoring loans. From a long-standing legacy of a centrally planned economy, lending has historically been government-mandated or often reliant on individual relationships or collateral rather than credit analysis and risk management [1]. These results provide evidence of the agency cost problems, which are conflicts of interest between managers and shareholders. Hence, the more diversified banks' funding sources, the higher credit risks banks will take. Interestingly, both income and funding diversification ascend the risk-taking behavior in Vietnamese commercial banks [[16], [17], [18]]. Positive BFD coefficients support Hypothesis 2 and are consistent with Vo's (2018) findings. Moreover, a positive relationship between BFD and NPL variables fills the research gap in Vietnamese bank literature.

Evidence from Table 5 shows a negative relationship between bank capital (CAP) and NPLs, showing that an increase in bank capital lessens the non-performing loans. When highly capitalized, banks use lower financial leverage, which could reduce the distress risks. Our result aligns with [4,69] papers but is inconsistent with [7,42,43] studies. Table 5 shows a positive relationship between bank size (SIZE) and NPLs. This finding implies that an increase in NPLs at Vietnamese commercial banks is related to their enlarging size [18]. We find a positive nexus between ROA and NPL variables, consistent with [4,31]. Additionally, Table 5 indicates a positive impact of the inflation rate (INF) on NPLs, consisting of [70], and differs from Ref. [71]. Estimated results imply that a higher inflation rate increased NPLs in Vietnamese commercial banks from 2009 to 2020. A high inflation rate weakens the ability of borrowers to repay, so non-performing loans increase significantly. Further, all estimated coefficients remain signs across columns of Table 5, implying that findings with an alternative LC variable (LC2) are consistent with the primary LC variable (LC1).

In Table 6, we employ Z-score and RROE as alternative proxies for bank risk-taking [17,18]. Z-score and RROE indexes measure the accounting distance to default for a given institution. Regression results for the Z-score dependent variable regarding LC1 and LC2 variables are listed in Columns (1) and (2), respectively. Estimated results for the RROE dependent variable regarding LC1 and LC2 variables are listed in Columns (3) and (4), respectively. Most results in Columns (1)–(4) suggest LC mitigates risks in Vietnamese commercial banks. Empirical results in both Table 5, Table 6 strongly confirm our arguments for Hypotheses 1. In short, our main findings are robust to alternative proxies of liquidity creation and bank risks. Unfortunately, the results in Table 6 indicate no relationship between BFD and bank default risks (Z-score and RROE).

Table 6.

Regression dynamic model (1) with alternative variables of bank risk-taking by using LSDVC estimations.

Independent variables	Z-score is the dependent variable				RROE is the dependent variable
	(1)		(2)		(3)		(4)
	Coef.	Std Err.	Coef.	Std Err.	Coef.	Std Err.	Coef.	Std Err.
lag. Z-score	0.5381***	0.0522	0.5558***	0.0549
lag.RROE					0.2013***	0.0244	0.2002***	0.0242
LC1	0.1495**	0.0658			0.4015**	0.1885
LC2			0.1092*	0.0657			0.4617**	0.1836
BFD	0.1476	0.1039	0.1204	0.0756	0.1722	0.2261	0.2166	0.2220
SIZE	0.0921***	0.0291	0.0671**	0.0307	0.2064**	0.0876	0.2017**	0.0871
ROA	−3.0739***	0.8583	−2.3660***	0.8433	111.3960***	4.6900	111.5551***	4.6804
CAP	−0.8366***	0.2581	−0.6246**	0.2444	−6.6684***	0.7621	−6.4900***	0.7579
LG	0.0513	0.0396	0.0431	0.0315	−0.0697	0.0821	−0.0748	0.0818
INF	0.1894	0.1511	0.1413	0.1386	−0.3532	0.3948	−0.3512	0.3928
GDP	−0.0648	0.3312	−0.2319	0.3278	2.4293**	0.9877	2.3535**	0.9827

Note: Table 6 summarizes the regression results from the dynamic model (1) by using the LSDVC. The risk-taking behavior proxies are Z-score and RROE. All variable definitions are displayed in Table 1. ***, **, and * indicate significance levels at 1%, 5%, and 10%, respectively. Columns (1) and (2) report estimated results from the dynamic model (1) with the Z-score dependent variable. The LSDVC in this section, initialized by AB, is bootstrapped by 50 iterations for the standard errors (Bun & Kiviet, 2003 and Bruno, 2005). Columns (3) and (4) report estimated results from the dynamic model (1) with the RROE dependent variable.

4.4. Moderating role of bank size in the effects of liquidity creation and bank funding diversification on non-performing loans

To examine potential differences between large and small banks more closely, we add respectively two interaction variables (DUM*LC and DUM*BFD) to the dynamic model (1). Empirical results in Table 7 demonstrate that the relationship between bank liquidity creation (LC), bank funding diversification (BFD), and NPLs are significantly moderated by bank scale. Columns (1) and (2) respectively show that DUM*LC1 and DUM*LC2 coefficients are negative and significant, while DUM*BFD coefficients are positive and significant.

Table 7.

Moderating role of bank size in the effects of liquidity creation and bank funding diversification on non-performing loans.

Independent variables	NPL is the dependent variable (1)		NPL is the dependent variable (2)
	Coef.	Std Err.	Coef.	Std Err.
lag.NPL	0.1215***	0.0001	0.4415***	0.0469
DUM*LC1	−0.0396***	0.0078
DUM*LC2			−0.0277**	0.0113
DUM*BFD	0.0456***	0.0038	0.0158***	0.0057
LC1	−0.0274*	0.0157
LC2			−0.0131	0.0102
BFD	0.0591**	0.0268	0.0241**	0.0108
ROA	0.4454***	0.1127	0.4257***	0.1493
CAP	−0.5268***	0.0240	−0.0794***	0.0236
LG	−0.0386***	0.0059	0.0017	0.0036
INF	0.2129***	0.0251	0.0735***	0.0158
GDP	−0.8120***	0.1377	−0.0683	0.0459

Note: Table 7 reports the regression results from the dynamic model (2) by using the LSDVC. We use a dummy (DUM) variable that represents large banks. DUM equals one if the bank scale is above the median size and otherwise. Interaction variables (DUM*LC and DUM*BFD) are created to estimate the moderating role of bank scale on the impact of liquidity creation (LC) and bank funding diversification (BFD) on non-performing loans (NPL). The LSDVC initialized by AB is bootstrapped by 50 iterations for the standard errors (Bun & Kiviet, 2003 and Bruno, 2005). ***, **, and * indicate significance levels at 1%, 5%, and 10%, respectively.

The negative coefficients of DUM*LC variables suggest that large Vietnamese banks generate greater liquidity and significantly reduce NPLs compared to small Vietnamese banks. Small banks have fewer opportunities for diversifying their funds, primarily granting credits to create liquidity, exposing them to credit risk (NPLs). Along the same lines, large banks have a professional commercial ability to attract more customers and provide additional loans that reduce the liquidity level [34]. Our results coincide with [34] and the shiftability theory of liquidity. Moreover, since small banks are already impacted due to the cost implications of economies of scale enjoyed by large banks, these banks might relatively increase instability in the liquidity-creating process [44].

Additionally, the positive coefficient of DUM*BFD variables implies that diversifying funds makes large banks take higher credit risks than small banks [17]. demonstrate that large Vietnamese banks have more favorable conditions for diversifying activities. Large and industry-leading banks universally have a higher reputation than small banks, thus having access to more capital and a better customer base. Our findings are partially consistent with the signaling theory. With abundant capital, banks are more comfortable in increasing credit risk to achieve higher profits, especially in the context of Vietnam, where more than 70%2 The income of commercial banks originates from loan interest. An empirical investigation [18] shows that Vietnamese banks tend to take higher risks in the fund diversifying process. Overall, these results in Table 7 supply supportive evidence for the “too big to fail” hypothesis; furthermore, they suggest Vietnamese large banks trade off risk and returns created by funding diversification.

5. Conclusions and implications

Our research gives a deep insight into the influences of bank liquidity creation LC), bank funding diversity (BFD), bank characteristics, and macroeconomic factors on bank risk-taking from the perspective of a transition market – Vietnam. This research analyzes a panel sample of 367 annual observations from 33 Vietnamese commercial banks in the post-financial crisis (2009–2020). We use Bias-corrected Least-Squares with Dummy Variables (LSDVC) estimations which perform better than GMM estimations in the case of small and unbalanced panel samples. To suit the Vietnamese context and fill existing gaps, we use the variable non-performing loans (NPLs) to represent bank risk-taking. Our main findings show that liquidity creation attenuates bank risks. Estimated results are also robust to alternative bank risk and liquidity creation proxies. In contrast, funding diversification motivates Vietnamese banks to take higher credit risks. The adverse link between liquidity creation and NPLs implies that banks provide liquidity to the economy to increase economic growth. Therefore, borrowers have a higher loan repayment capability, resulting in lower levels of non-performing loans, thereby minimizing bank risks. In addition, a negative nexus between BFD and NPL variables suggests that diversifying funds in Vietnamese commercial banks promotes bank managers to be less cautious about credit risks. Finally, the advantage of scale helps large banks create better liquidity and significantly reduce NPLs.

This study enriches bank risk-taking literature in a way to examine the impact of liquidity creation on bank risk-taking in the specified context of a transition economy - Vietnam. Moreover, findings from this paper offer three practical implications for bank managers and policymakers in emerging and transition markets. Firstly, regulators should be vigilant in making policies related to BFD because this diversification might make commercial banks face more credit risks due to excessive LC. Secondly, the LC process significantly reduces NPLs. Thus, both Vietnamese bank managers and regulators should implement reforms to encourage banks to provide more liquidity for the economy. According to the LC process, credit quality increases since borrowers have more opportunities to repay their debts. Thirdly, Basel II compliance in the Vietnamese banking system could result in lower LC, whereas higher LC could result in more bank failures. Thus, the trade-off between the advantages of increased LC and the benefit of financial stability is generated by tighter capital requirements for Vietnamese authorities in strengthening the banking industry. Fourthly, combining LC and BFD is a mixed strategy to help regulate and balance NPLs in Vietnamese commercial banks after the 2008 GFC. Finally, bank managers should improve their credit quality instead of focusing on the volume of loans in a bloom period. In addition, an efficient management system and adequate capital could minimize the economic instability influencing the banking system.

This study has the following limitations. Firstly, our paper only concentrates on commercial banks in the Vietnam market. Coming studies might extend to investigating multi-countries to create in-depth insights on this topic. Secondly, future research should examine the moderating effect of ownership structure on the association between liquidity creation and risk-taking. The presence of state or foreign ownership in banks might change the bank's view of risk-taking. Thirdly, due to focusing on one banking system, empirical results from this paper could not be compared to the ones of other markets, which provide more exciting evidence.

Funding statement

This study is supported by Ton Duc Thang University and Ho Chi Minh University of Banking.

Ethics and statement

This study does not involve animals or humans.

Author contribution statement

Giang Thi Huong Vuong: Analyzed and interpreted the data; Wrote the paper. Phuong Thi Thanh Phan; Cuong Xuan Nguyen; Danh Minh Nguyen: Performed the experiments; Contributed reagents, materials, analysis tools or data. Khoa Dang Duong: Conceived and designed the experiments; Wrote the paper.

Data availability statement

Data will be made available on request.

Additional information

No additional information is available for this paper.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.