Does fintech lending expansion disturb financial system stability? Evidence from Indonesia

Abstract

Literature suggests two contradictory views regarding the impacts of fintech lending on banks. The competition-instability proponents believe that fintech lending expansion erodes bank market and threatens banks as traditional intermediaries, thereby intensifying bank risk-taking and potentially disturbing financial stability. In contrast, the competition-stability proponents believe fintech lending reduces asymmetric information in the credit market, thus reducing bank risk-taking and increasing bank resilience to a systematic shock. This paper aims to examine the impacts of fintech lending expansion on bank-risk taking, i.e., credit channeling activity and bank risk. This study utilizes the OLS, random effects, fixed effects, and two-step GMM regressions to test the conjectures. Consistent with the competition-stability hypothesis, our evidence indicates that shadow banking increases as banks actively seek channels to diversify their risks but are less likely to use fintech lending as a credit channel. This paper also corroborates the notion that fintech lending expansion encourages banks to diversify their risks. Pertaining to the relationship between fintech lending and bank risk, our results suggest that fintech lending expansion encourages banks to work more efficiently to improve their credit quality rather than to intensify bank risk-taking. These findings also indicate that synergy between fintech lending and banks would increase bank credit quality. This paper also examines the reasonable credit limits for fintech lending firms based on MSMEs’ characteristics. Employing the threshold regression, we find that IDR5 billion is the maximum credit provision to induce the profitability of MSMEs whereas IDR6 billion is the maximum credit provision to minimize the default risk of MSMEs.

1. Introduction

Fintech refers to innovative technology-based financial services or products [1]. The literature notes that fintech has at least two roles in the economic system. On the one hand, fintech plays a pivotal role in increasing financial inclusion [[2], [3], [4]], thus being considered a determinant of inclusive growth [5]. On the other hand, fintech also substantially impacts the financial services industry [6,7]. In this regard, advances in financial technology are considered significantly changing the landscape of financial services industry by fostering the emergence of fintech lending [8].

However, how fintech lending affects financial system stability remains inconclusive. The literature suggests two points of view regarding the effects of fintech lending expansion on financial system stability. Based on the competition-fragility view, fintech lending is deemed a head-to-head competitor with banks since they conduct funding and lending activities through online platforms, thus threatening banks as traditional intermediaries [9,10]. Such circumstances escalate bank asset risks as well as degrade bank profitability, hence potentially intensifying bank risk-taking behavior and disrupting the financial system stability [[11], [12], [13], [14], [15]]. Fintech lending as a shadow bank might also be utilized as a regulatory arbitrage as it is less regulated than banks, making the financial system more vulnerable to systemic risks [12,16,17].

Conversely, the competition-stability view argues that a concentrated market potentially triggers moral hazard and financial sector instability [18]. Greater competition in the banking system encourages banks to diversify their risks, thereby increasing bank resilience to systemic shocks [19]. In the fintech-bank relationship context, advancement in information technology allows fintech to help banks develop their businesses and promote their innovation [[20], [21], [22]]. Fintech also helps banks mitigate asymmetric information in the credit market so as to reduce their transaction costs, risk-taking behavior, and total risks [[23], [24], [25], [26], [27]].

In Indonesia, a shadow bank refers to a non-bank institution that receives, manages, and distributes funds to the public. Shadow banks provide various services, including multi-finance financing, factoring, mortgage loans, hedge funds, private equity funds, investment banks, insurance firms, payment systems, etc. In contrast to banks, shadow banks are not strictly regulated by the financial services regulators. One of the shadow banks is fintech lending or peer-to-peer lending. According to the Indonesian Financial Services Authority (OJK) Regulation No.77/POJK.01/2016, fintech lending refers to direct, technology-based credit services between creditors and debtors using the rupiah currency. In other words, fintech lending is a shadow bank that utilizes information technology, and serves as an intermediary between creditors and debtors in the credit market.

Due to uncertainty regarding the fintech lending-bank relationship, regulators are concerned that fintech lending expansion might increase competition in the financial markets, especially in the lending-and-borrowing activities, which could intensify bank risk-taking and ultimately disturb the financial system stability. The regulators are also anxious that banks might employ fintech lending as credit channels that can increase their risks as fintech loans and investments are offered via the internet, which make them less regulated and could be used as a regulatory arbitrage. To maintain the financial system stability, the OJK has set IDR 2 billion per customer as the maximum amount of credit disbursed by a fintech lending firm. This paper purports to examine the impacts of fintech lending expansion on bank-risk taking, i.e., credit channeling activity and bank risk.

Subsequently, micro, small, and medium enterprises (MSMEs) could use banks and fintech lending as sources of funding. Financing through banks requires a vast array of requirements, such as collateral and business licenses, and needs to provide detailed financial statements and bank transactions for certain years to get through a rigorous process. However, MSMEs are faced with difficulties accessing this type of financing due to the lack of collateral and being less professionally managed, thereby unable to present complete financial reports and often considered high-risk profiles by banks [[28], [29], [30]]. Unlike banks, fintech lending does not require stringent requirements from its users, making it more convenient for MSMEs.

In addition, as fintech lending tends to provide credits to lower credit-rating market segments [31,32], fintech lending provides higher opportunities for MSMEs to obtain loans to develop their businesses [33]. In this sense, the OJK noted that credits distributed by the fintech lending industry to MSMEs had experienced a significant increase from IDR16.07 billion in January 2021 to IDR45.73 billion in July 2022. However, existing credit-limit regulations might hinder fintech lending in fulfilling MSMEs’ credit needs, which would indirectly increase their costs of capital. The regulators should set credit limits for fintech lending by considering the default risk of MSMEs to maintain robust risk management in the financial sector. Hence, this paper also examines reasonable fintech lending credit limits based on the MSMEs' characteristics to maximize their growth while at the same time minimizing their default risk.

This paper contributes to the literature in several ways. First, incumbent literature has studied how fintech development affects bank risk [34] and bank risk-taking [24,26], as well as how regulatory differences and technological advantages contribute to the growth of shadow banks [12]. Although some researchers have indicated that fintech has played a role in increasing shadow banks [12,16], empirical evidence on the direct relationship between fintech lending expansion and bank channeling activity is still scarce. To fill this gap, this paper aims to examine whether fintech lending expansion increases bank channeling activity using monthly data from 38 regions in Indonesia from January 2019 to August 2022.

Employing OLS, random effects, fixed effects, and two-step GMM regressions, our evidence suggests that bank channeling increases as banks actively seek channels to diversify their risks [19]. However, our results also indicate that banks are not likely to use fintech lending as credit channels. In addition, we find a positive and significant interaction effect of fintech lending expansion and bank lending activity on credit channeling activity, verifying that fintech lending expansion encourages banks to diversify their risks. Our evidence substantiates the competition-stability view, where fintech lending expansion motivates banks to diversify and ultimately increase bank resilience to systematic shocks.

Second, the literature shows that fintech increases bank risk by increasing transaction costs [35] and reducing bank profitability [12]. However, the literature lacks empirical evidence on a direct relationship between fintech lending expansion and bank credit quality. This paper attempts to fill the gap by examining the relationship between fintech lending expansion and bank credit risk as a proxy for bank credit quality. Consistent with the competition-stability view and several prior studies [e.g., 13, 24, 26] our results document that fintech lending expansion pushes banks to work more efficiently to improve their credit quality rather than intensify bank risk-taking behavior. Our findings also indicate that a synergy between fintech lending and banks will increase bank credit quality, shown by the negative and significant interaction effect of fintech lending expansion and fintech lending activity on credit risk.

Third, this paper also attempts to examine reasonable fintech credit limits by considering the characteristics of MSMEs to maximize their growth while maintaining their default risk. We employ the dynamic threshold regression and consider debtors' characteristics, i.e., profitability and leverage, as response variables to estimate reasonable credit limits for fintech lending firms. Our evidence shows that the reasonable credit limits for fintech firms are between IDR5 billion and IDR6 billion in order to maintain MSMEs' default risk. However, our estimation reports that lending above IDR5 billion is not likely to increase MSMEs’ profitability. The findings shed some light that the regulators should pay attention to MSMEs' characteristics, such as profitability and leverage, in determining fintech credit limits to foster MSMEs' growth while at the same time maintaining the financial system stability.

The remainder of the paper is organized as follows. Section 2 provides brief literature review and hypotheses development. Section 3 describes data and methods. In Section 4, we present empirical results and discuss the findings. Section 5 concludes this paper.

2. Literature review and hypotheses development

2.1. Fintech lending expansion and bank risk taking

The emergence of fintech is believed to be able to dramatically change the financial sector, including generating new business models, applications, processes, and products [8,10,20,36]. One of the developments in financial services business process and products is fintech lending [37]. Fintech infrastructure allows fintech lending firms to establish online lending and investment platforms, making them head-to-head competitors to banks [9,38,39]. In contrast to banks, which are subject to stringent credit standards, fintech lending offers credit without sturdy credit standards since online lending features are not strictly regulated [9,14,40]. Such conditions increase competition in the financial services industry as fintech lending operates similarly to the banks, i.e., lending and borrowing [11,12]. Literature suggests two points of view on how competition in the financial services industry affects financial system stability, i.e., competition-instability view v. competition-stability view.

Competition-instability view argues that competition in the financial sector might intensify bank risk-taking behavior to maintain its franchise value, rendering failure and instability to the financial systems [41]. According to Ref. [42], fintech lending firms operate more efficiently to provide financial services at lower prices, thereby increasing competition in the credit market and eroding bank's market share. For instance, fintech growth in the residential mortgage market has penetrated bank's market share and induced 30% growth of shadow banks in the U.S [12]. Meanwhile, fintech lending firms indirectly operate in the deposit market as they also develop investment platforms in their business process. Accordingly, fintech lending expansion has increased competition in the deposit market as well [9]. In some previous studies, fintech lending expansion is found to reduce bank profitability and bank growth as well as to increase the cost of debt, thus increasing risks on bank's asset side [15,35,42]. Such circumstances weaken the position of banks in the financial sector, and may elevate bank risk-taking behavior, which leads to financial system instability [15,38,[43], [44], [45]].

In a more competitive environment, banks earn lower rents, which reduce their incentives for monitoring [46,47]. In addition, the pressure to generate profits in a more competitive atmosphere will encourage banks to choose riskier portfolios [44]. Hence, banks are actively collaborating with and using fintech lending as credit channels to increase their profits [14,16]. However, credit channeling through fintech lending also has the potential to worsen financial system stability since fintech lending relies on fee-based incomes from loan contracts and often targets debtors with lower credit ratings (or high-risk debtors) [3,14,31]. In this regard, bank's loan loss provision will increase if the placement of funds through fintech lending collapses. In other words, the proponents of harmful consequences of fintech lending argue that competition in the financial sector might intensify bank risk-taking behavior to maintain bank ‘franchise value,’ which leads to financial system failure and instability.

On the contrary, the competition-stability view argues that competition in the financial services industry does not necessarily engender excessive bank risk-taking or financial fragility [18]. With respect to studies on competition in the financial sector, a concentrated market allows a bank to charge higher interest rates, so borrowers need to invest in riskier projects to repay their borrowings, which ultimately increases default probability on the bank's asset side, increases bank failure probability, and causes financial sector instability [18,48]. In addition, a highly concentrated financial industry might also trigger moral hazards that intensify risk-taking behavior by large banks due to the ‘too-big-to-fail’ policy [49]. In this sense, greater concentration in the banking sector enlarges bank market power and risk exposure [50], where banking market power is associated with higher insolvency and thus intensifies ‘too-big-to-fail’ benefits [51,52]. Meanwhile, a competitive banking system is found to be less likely to experience systemic crises [53] since greater competition in the banking system encourages banks to diversify their risks [19].

In the context of fintech impacts on banks, fintech lending expansion does not necessarily increase bank risk and/or bank risk-taking behavior as fintech lending is gradually penetrating underserved market segments by traditional financial institutions [32,54]. Fintech infrastructure might also help banks innovate as well as reduce traditional bank information asymmetry and the frictional costs of bank transactions, to improve bank risk management and bank performance as well as foster financial system stability [22,27,55]. Within this context, a recent study finds a significantly negative relationship between fintech expansion and bank risk-taking behavior [24]. In examining the effects of fintech lending expansion on bank risk-taking in China from 2011 to 2020 [26], suggest that forming a synergy between fintech lending and banks would reduce bank risk-taking, where fintech lending could help banks effectively measure customer creditworthiness.

Another research provides evidence that fintech innovation increases bank's capital adequacy and management efficiency, even though it reduces large state-owned commercial banks' profitability and asset quality [24] since large banks tend to have lower capital, less stable funding, and are more likely to perform market-based activities [56]. Besides, large banks are also more complex to supervise, given their level of complexity and their political ability to capture their supervisors [57]. As a concentrated market is dominated by few large banks [58], increased competition will discipline banks to work more efficiently [24]. Overall, supporters of the positive consequences of fintech lending suggest that increased competition in the financial sector disciplines banks to work more efficiently and encourages banks to diversify, thus fostering the financial system stability.

In relation to the impacts of fintech lending expansion on competition in the financial markets, fintech lending firms tend to pursue market segments not served by banks as fintech lending is more flexible to provide credits to lower credit-rating borrowers [31,32]. This evidence indicates that fintech lending and banks are not involved in head-to-head competition in the financial services industry, suggesting fintech lending expansion does not directly affect bank risk. However, since banks often perform credit channeling through fintech lending [12,16], this action might elevate bank risk as a consequence of the fact that fintech lending is more likely to target lower credit-rating borrowers. Taking the competition-stability perspective, we expect a positive relationship between bank lending and credit channeling activity, presuming that banks perform credit channeling to diversify their risks [19]. Under such circumstances, banks might not use fintech lending as their credit channels since fintech lending borrowers are deemed high risk by banks [32], while banks have more incentives to maintain the trust of investors by maintaining the quality of their lending [59]. Hence, we expect a negative or insignificant relationship between fintech lending expansion and credit channeling.

Subsequently, regarding the relationship between fintech lending expansion and bank risk-taking behavior, a string of studies indicates that fintech lending expansion weakens bank performance, including bank profitability [15,42], and increases the cost of debt of banks [35]. However, another study reports that fintech lending expansion disciplines banks to work more efficiently [24], suggesting that banks would escalate their business efficiency as competition increases in the financial markets. Adopting a competition-stability view, we conjecture that fintech lending expansion increases bank efficiency by maintaining credit quality. Accordingly, we expect a negative relationship between bank lending activity and credit risk. Furthermore, fintech borrowers tend to use fintech lending due to its convenience, even though they are subject to relatively higher interest rates than those charged by banks [12]. Such conditions indicate that prospective borrowers with higher credit ratings are more likely to borrow from banks to obtain lower interest rates such that fintech lending expansion does not erode bank's market share. Moreover, banks find it easier to assess prospective borrowers to improve their credit quality since less-strong borrowers are more likely to obtain funding from fintech lending. Therefore, we expect a negative or insignificant relationship between fintech lending expansion and bank credit risk.

Hypothesis 1: Bank lending activity is positively related with bank credit channeling

Hypothesis 2: Fintech lending expansion is negatively related or is uncorrelated with bank credit channeling

Hypothesis 3: Bank lending activity is negatively related with bank credit risk

Hypothesis 4: Fintech lending expansion is negatively related or is unrelated with bank credit risk

2.2. Fintech lending limits and MSMEs characteristics

Literature has documented that MSMEs emerge as an economic survival action to support family incomes due to a lack of employment opportunities [[60], [61], [62]]. MSMEs often help the poor by increasing their incomes, and thrive in countries with low employment availability, especially in developing countries [63]. Hence, MSMEs are considered notable actors in the poverty alleviation effort in many countries, including Pakistan, Nigeria, and most Southeast Asian countries [64,65]. However, MSMEs often experience financial distress, especially in accessing credit [28,29]. [30] postulate that MSMEs do not have sufficient assets to be collaterals. Moreover, banks deem MSME borrowers high-risk clients (Rahman et al., 2017). As a result, micro-enterprises often generate less profit and are unsustainable in the long run [62].

Meanwhile, fintech lending infrastructure can reduce information asymmetry [27], which allows fintech lending to provide online credit contracts without stringent credit standards such as those imposed on bank's borrowers [9,66,67], thus alleviating credit rationing [66] and providing higher opportunities for MSMEs to obtain financing to develop their businesses [33]. Fintech lending credit distribution is limited in several countries, however, as regulators are still uncertain as to the impacts of fintech lending on the financial systems, especially banking. As a matter of fact, fintech lending credit is limited to merely IDR2 billion in Indonesia. On the one hand, this policy can minimize potential impacts that might arise in the banking industry due to increased competition in the financial services industry and shadow banking activities. On the other hand, this regulation might hinder MSMEs' growth. Regulators need to pay attention to the creditworthiness of fintech lending users in determining fintech lending credit limits. This is useful for minimizing the effects of fintech lending credit on banks while simultaneously maximizing the growth of MSMEs if banks conduct loan channeling (shadow banking).

Creditworthiness refers to the assessment of a debtor's ability to fulfill his or her credit obligations on the due date without affecting the vitality of the borrower [68]. In other words, creditworthiness predicts whether a borrower will default (unable to pay his or her debt). Based on [29], financial profile is the most valid predictor in forecasting the creditworthiness of a prospective borrower. Literature suggests five accurate characteristics in describing the profile of a prospective debtor, i.e., profitability, liquidity, leverage, coverage, and activity [29,69,70]. Indeed, profitability, liquidity, and leverage are the most significant characteristics to distinguish healthy and unhealthy firms [69].

Profitability is mulled one of the most pivotal financial aspects in predicting a debtor's creditworthiness as it shows how efficiently a firm generates profit to repay its borrowings [68]. [71] finds a negative relationship between profitability and default risk (or bankruptcy risk), which suggests that the higher a firm's profitability, the lower the firm's default risk. Meanwhile, liquidity is also a crucial aspect for firms to survive [72]. Existing studies have documented that liquidity affects firm performance [69,73]. The literature suggests a trade-off in the liquidity-profitability relationship. Generally, higher liquidity denotes a stronger financial position [74]. Higher liquidity enables a company to maintain a satisfactory level of working capital, so the firm can cover its liability with a reasonable safety margin [75]. Moreover, high liquidity also guarantees the company to make very profitable investments that demand immediate payments [76]. In contrast, some authors, such as [77], show that the bigger the current assets of a firm, the lower its profitability and risk of working capital. In other words, a more liquid firm will generate lower profitability but bear a lower working capital risk than do firms with a less liquid working capital.

[78] explains that in reflecting a firm's profitability, liquidity and financial position are also affected by leverage. Financial leverage refers to the level of funding for a firm's operations funded through external resources [79]. [80] suggest that leveraged firms have additional capital available to finance their operations and expansion compared to non-leveraged businesses that hinge solely on equity. Furthermore, debt could also boost up firm performance since debt allows debtholders to help control management behavior [81]. [82] show evidence that indebtedness positively affects firm total factor productivity. However, excessive debt might cause the firm not capable of paying off its short-term obligations (e.g., debt maturing in one year), which may force the firm to liquidate its assets at discount. Accordingly, the downside of employing too much debt is the increasing bankruptcy risk [83].

With respect to the relationship between leverage, liquidity, and profitability [84], explain that debt increases firm liquidity, then liquidity escalates the firm's power in negotiating higher discounts on cash purchases and allows the firm to expand its capacity, which then increase its profitability. Nevertheless, excess liquidity might also trigger overinvestment problems [85,86], e.g., investing in projects with a negative NPV [87,88]. [89] explain that excess liquidity can exacerbate the overinvestment problems as financial resources are considered a ‘free’ resource when capital is easy to obtain with a low discount rate.

The aforementioned findings are in line with several studies that show that funding and firm performance have an inverted U-shaped relationship [84,90]. Those previous findings indicate that internal and external financing composition affects firm performance. In the context of fintech lending credit limits, higher limits allow fintech lending to provide more credits to MSMEs. On the one hand, increasing credits helps MSMEs improve their profitability by increasing liquidity, strengthening their financial position, and enabling them to invest in profitable projects [74,76,84]. On the other hand, excessive credits will increase the likelihood of default of MSMEs as they bear the burden of excessive interest due to higher debt levels [71,72,83]. Hence, we expect a positive relationship between fintech lending credit limits and MSMEs’ profitability up to an optimal point, then a negative or insignificant relation upon passing beyond the optimal point.

Meanwhile, increasing credits can leverage benefits to MSMEs, e.g., debtholder control, more capital to expand their capacity, and induce firms' power in negotiating higher discounts on cash purchases, thereby improving their performances [80,81,84]. However, excess leverage can also trigger overinvestment problems as MSMEs may perceive financial resources as a ‘free’ resource, and hence increases the likelihood of bankruptcy of MSMEs [86,89]. We conjecture a negative or insignificant relationship between fintech lending credit limits and MSMEs’ leverage (or default risk) up to an optimal limit, then a positive relation after passing beyond the optimal limit.

Hypothesis 5: Credit limits are positively correlated with profitability up to an optimal limit, and then negatively or insignificantly related after passing the optimal level

Hypothesis 6: Credit limits are negatively or insignificantly correlated with leverage (default risk) up to an optimal limit, and then positively related after passing beyond the optimal limit

3. Methodology

This paper examines the impacts of fintech lending expansion on bank risk-taking. We employ bank credit channeling and nonperforming loans (NPL) as response variables in examining the relationship between fintech lending expansion and bank risk-taking. Bank credit channeling refers to a bank's placement of funds in nonbank institutions [91,92]. Meanwhile, nonperforming loans (NPL) are defaulted loans due to borrowers not making payments at the scheduled times. Literature highlights that banks earn lower rents and less profit in a competitive market environment, reducing their incentives for monitoring and stimulating them to establish riskier portfolios, such as by utilizing credit channels and/or lowering their credit standards [14,15,44,46,47]. Therefore, stiffer competition in the financial sector results in increased credit channeling and nonperforming loans [93,94].

Prior studies argue that fintech lending expansion intensifies competition in the financial sector and potentially erodes bank market share [9]. Through their online lending and investment platforms, not only does fintech lending pressurize banks as traditional intermediaries but it also erodes bank lending markets [10,11]. Consequently, fintech lending expansion might increase bank risk-taking due to higher competition in the financial services industry [9,14,40]. In this paper, we use fintech lending expansion as the first explanatory variable. In contrast to previous studies, we define fintech lending expansion as growth in the volumes of credit fintech lending in a particular area. We believe that the growth in credits disbursed by fintech lending could more clearly demonstrate the fintech lending expansion. Furthermore, we employ bank lending activity as the second explanatory variable since bank lending activity through lending rates and profit-taking policy affect default rates and the uses of bank credit channels [95].

Subsequently, the literature also documents that bank-specific characteristics and macroeconomic conditions influence bank risk-taking behavior. Accordingly, we control bank-specific characteristics and macroeconomic factors. Bank-specific characteristics, such as size, the level of liquidity, profitability, reserves level, and efficiency are found to influence the risk behavior of banks in responding to changes in the financial markets [24,26,96]. Following prior research, bank-specific characteristics are measured as follows: size by the natural logarithm of total assets, profitability by return on assets (ROA), liquidity by current ratio, reserves level by loan-to-deposit ratio (LDR), and efficiency by the ratio of net income to operating expenses. Meanwhile, macroeconomic factors, such as market competition, economic growth, and inflation also affect bank risk-taking behavior [97]. We measure market competition as the Herfindahl–Hirschman index, economic growth as the increasing value of gross domestic product (GDP) in each period, and inflation as the rate of increase in prices over a given period, as suggested by previous studies. Table 1 summarizes all operational variables employed in examining the fintech lending and bank relationship.

Table 1.

Variable measurements and descriptive statistics.

Variable	Code	Measurement	Mean	Std Dev	Min	Max
Credit Channeling	CC	CCt – CCt-1	37.5	401.161	−649	832
Bank NPLs	NPL	NPLt – NPLt-1	26.685	474.801	−3605	6916
Fintech Lending Expansion	FLE	TFLt – TFLt-1	331.492	798.69	−406	7113
Bank Lending Activity	BLA	TBLt – TBLt-1	622.076	8478.766	−62109	133,723
Bank Size	Size	Log (TA)	10.964	1.3582	8.456	15.185
Bank Liquidity	CR	CA/CL	1.091	.333	.364	2.011
Profitability	ROA	NI/TA	.007	.003	.001	0.013
Reserves Level	LDR	TBL/TD	.833	.0462	.782	.933
Bank Efficiency	BE	NI/OC	.835	.028	.773	.888
Market Concentration	HHI		.942	.046	.714	.993
Gross Domestic Product	GDP	GDPt – GDPt-1	868.73	9755.133	−168,806	112,823
Inflation Rate	INF	Inflation Ratet	.002	.004	−.018	.023

This table presents variable measurements and descriptive statistics of operational variables. CC is bank's placement of funds in nonbank institutions. NPL is bank's nonperforming loans. TFL is total credits disbursed by fintech firms. TBL is total credits disbursed by banks. TA is bank's total assets. CA is bank's current assets. CL is bank's current liabilities. NI is bank's net income. TD is total deposits. OC is operating costs. HHI is the Herfindahl–Hirschman index with the maximum value of 1.

To inspect the effect of fintech lending expansion on bank credit channeling, we use the following regression model:

$${CC}_{it}=\alpha +{\beta }_1{FLE}_{it}+{\beta }_2{BLA}_{it}+{\beta }_3{Size}_{it}+{\beta }_4{LDR}_{it}+{\beta }_5{ROA}_{it}+{\beta }_6{CR}_{it}+{\beta }_7{BE}_{it}+{\beta }_8{HHI}_{it}+{\beta }_9{GDP}_{it}+{\beta }_{10}{\mathit{INF}}_{it}+{\epsilon }_{it}$$ (1)

where CC_it is the change in total credits channeled by banks through nonbank institutions in period t for region i; FLE_it is the growth of total credits disbursed by fintech lending in period t for region i; BLA_it is the growth of total credits disbursed by banks in period t for region i; Size_it is the natural logarithm of bank total assets in period t for region i; LDR_it is total credits disbursed divided by total third-party funds in period t for region i; ROA_it is bank's return on assets in period t for region i; CR_it is bank's current ratio in period t for region i; Be_it is bank efficiency in period t for region i; HHI_it is the regional Herfindahl index that gauges market competition in period t for region i; GDP_it is the growth of gross domestic product in period t for region i, and INF_it is inflation rate in period t for region i.

We employ the regression model below to examine the relationship between fintech lending expansion and bank's NPLs.

$${NPL}_{it}=\alpha +{\beta }_1{FLE}_{it}+{\beta }_2{BLA}_{it}+{\beta }_3{Size}_{it}+{\beta }_4{LDR}_{it}+{\beta }_5{ROA}_{it}+{\beta }_6{CR}_{it}+{\beta }_7{BE}_{it}+{\beta }_8{HHI}_{it}+{\beta }_9{GDP}_{it}+{\beta }_{10}{\mathit{INF}}_{it}+{\epsilon }_{it}$$ (2)

where NPL_it is bank risk, measured as the change in total nonperforming loans in period t for region i.

Table 1 shows measurements and descriptive statistics of variables applied. Our data sample consists of 34 regions in Indonesia from January 2019 to August 2022. We exclude a sample that has less than three years of fintech and bank data, as suggested by Refs. [91,92]. The final data comprise 1023 region-month observations. The bank and fintech lending data were mainly collected from the Financial Services Authority's (OJK) data release. Meanwhile, gross domestic product (GDP), human development index (HDI), and inflation rate were obtained from Bank Indonesia's database. Subsequently, we test the correlations among independent variables to confirm that the operational models are free from the multicollinearity problem. Panel A of Table 2 shows the correlation matrix among the independent variables. The correlation matrix shows that only one relationship between variables is close to the threshold value of 0.85, i.e., FLE and Size with a value of 0.703. We further test the collinearity among variables, as shown in Panel B of Table 2. All VIF values are below the threshold of 5 [98], suggesting that the regression models are free from the multicollinearity problem.

Table 2.

Correlation matrix and multicollinearity test.

Panel A. Correlation Matrix among Independent Variables
	FLE	BLA	Size	CR	ROA	LDR	BE	HHI	GDP	INF
FLE	1.000
BLA	0.341	1.000
Size	0.703	0.132	1.000
CR	0.198	0.035	0.484	1.000
ROA	0.012	0.014	0.019	0.050	1.000
LDR	−0.126	−0.035	−0.039	−0.079	−0.330	1.000
BE	−0.131	−0.047	−0.031	−0.032	−0.023	0.355	1.000
HHI	−0.537	−0.061	−0.418	−0.276	−0.125	0.388	0.448	1.000
GDP	0.060	−0.026	0.083	0.020	0.083	0.055	−0.107	−0.063	1.000
INF	0.002	−0.023	−0.008	−0.001	0.061	−0.186	−0.452	−0.216	0.029	1.000

Panel B. Collinearity Diagnostics
Variable	VIF	SQRT VIF	Tolerance	R-Squared
FLE	2.97	1.72	0.337	0.662
BLA	1.21	1.10	0.829	0.170
Size	2.76	1.66	0.362	0.637
CR	1.46	1.21	0.686	0.313
ROA	1.09	1.05	0.915	0.084
LDR	1.37	1.17	0.732	0.267
BE	1.68	1.30	0.594	0.405
HHI	2.05	1.43	0.487	0.512
GDP	1.05	1.02	0.955	0.044
INF	1.28	1.13	0.782	0.217
Mean VIF	1.69

This table exhibits the correlation matrix and multicollinearity test of the models. Panel A presents the correlation matrix among independent variables, where the rule of thumb suggests that a correlation coefficient between independent variables be below 0.85. Panel B presents the collinearity diagnostics test, where literature suggests that a VIF value of each variable be below 5.

Moreover, this paper also examines fintech lending credit limits based on the characteristics of MSMEs as fintech lending users. To find appropriate fintech lending credit limits based on MSME characteristics, we employ two response variables, i.e., profitability and leverage, as suggested by Ref. [29]. Following [29], we measure profitability as EBIT/Sales, and leverage as total liabilities/total assets. As an explanatory variable, we use loan demand, which is defined as the amount of credits needed by MSMEs to develop their businesses, following [99]. Loan demand is measured using a survey item. In addition, we control for business age (AGE) and business owner education level (EDUC) in testing reasonable fintech lending limits, as suggested by Ref. [100].

We utilize the dynamic threshold regression to find proper fintech credit limits based on the characteristics of MSMEs (fintech lending clients), which is denoted as follows:

$$y_{it}={\mu }_i+{\beta }_1^{\prime }{x}_{it}I\left(q_{it}\le \gamma \right)+{\beta }_2^{\prime }{x}_{it}I\left(q_{it}>\gamma \right)+e_{it}$$ (3)

where y_it is the main variable of the stochastic scalar of dependent variable; x_it is the vector $k_1\times 1$ in the time variation of independent variables; I is an indicator function; $q_{it}$ is a transition variable or threshold variable; $\gamma$ is the threshold parameter; ${\epsilon }_{it}={\alpha }_i+v_{it}\text{,}$ is the error component in the regression that is the sum of two variables: (1) an individual component of the unobserved fixed effects and (2) a special random disturbance with an average of 0.

Sample to examine the fintech lending credit limits are fintech lending users. Since we desire to evaluate credit limits that fulfill the needs of MSMEs, we exclude borrowers who have assets above IDR50 billion. The Indonesian regulators define MSME as a business whose assets are below IDR50 billion. We also exclude borrowers who accessed lending for consumptive purposes, such as buying vehicles, mobile phones, etc. Therefore, our sample set are MSMEs with assets below IDR50 billion that borrowed from fintech lending for productive purposes. The MSME data were collected from a survey. We received 412 raw data out of 500 questionnaires that we distributed in the survey, and got 344 clean data after removing respondents with prerequisites above. Table 3 presents the descriptive statistics of respondents.

Table 3.

Descriptive statistics of fintech lending users.

Variable	Symbol	Obs	Mean	Std Dev	Min	Max
Profitability	PROF	344	.2755713	.0203,732	.2192739	.3310239
Leverage	LEV	344	.2611277	.0677,322	.1197591	.4559602
Loan Demand	LD	344	5.94e+09	1.03e+09	3.10e+09	9.44e+09
Business Age	AGE	344	8.781977	.6671122	7	11
Owners Education Level	EDUC	344	3.31245	1.745302	1	4
Unbanked	UB	344	.8023256	.3988253	0	1

4. Results and findings

4.1. Impacts of fintech lending expansion on bank credit channeling activity

This section presents and discusses the testing results of the relationship between fintech lending expansion and bank credit channeling activity by employing four econometric approaches: (1) Pooled Ordinary Least Squares (OLS); (2) Random Effects Model (REM); (3) Fixed Effects Model (FEM); and (4) Two-Step Generalized Method of Moments (Two-Step GMM). However, an unreported test of the GMM shows that model (1) cannot satisfy the Arellano-Bond condition, where AR (2) produces a significant z-value. Accordingly, we do not show the GMM results of model (1). Another unreported test shows evidence that the pooled OLS does not pass the Breusch-Pagan test, indicating that the OLS results contain the heteroskedasticity bias. In this regard, we use heteroskedasticity-robust standard errors in all regression models. Table 4 reports the testing results of model (1) employing OLS, REM, and FEM.

Table 4.

Testing results of model (1).

Panel A. Testing Results
	(1) OLS	(2) REM	(3) FEM	(4) OLS	(5) REM	(6) FEM
FLE	−.0007(−0.55)	−.0007(−0.56)	−.0049** (−2.23)	−.0041** (−2.53)	−.0041** (−2.56)	−.0206*** (−6.31)
BLA	.0028* (1.93)	.0028* (1.78)	.0030** (1.97)	.0020 (1.30)	.0020 (1.25)	−.0005 (−0.04)
FLExBLA				2.68e-09*** (2.86)	2.68e-09*** (2.57)	1.34e-08*** (6.45)
Size	−6.6101 (−0.41)	−6.6101 (−0.42)	886.7573** (2.35)	−10.5323 (−0.65)	−10.5323 (−0.67)	830.5437** (2.25)
CR	−62.0001 (−1.34)	−62.0001 (−1.35)	−1883.056*** (−6.56)	−77.1262* (−1.67)	−77.1262* (−1.67)	−2093.391*** (−7.40)
ROA	−19646.32*** (−5.61)	−19646.32*** (−5.06)	−21322.81*** (−5.64)	−19884.25*** (−5.67)	−19884.25*** (−5.13)	−21854.87*** (−5.90)
LDR	−8237.918*** (−22.73)	−8237.918*** (−22.93)	−6144.344*** (−14.33)	−8092.189*** (−22.06)	−8092.189*** (−23.26)	−5549.203*** (−12.91)
BIE	−4768.187*** (−9.31)	−4768.187*** (−8.39)	1468.536* (1.85)	−4362.423*** (−8.11)	−4362.423*** (−7.41)	3326.749*** (4.01)
HHI	−1453.286*** (−3.87)	−1453.286*** (−3.74)	−11454.84*** (−11.05)	−2383.801*** (−4.67)	−2383.801*** (−4.49)	−15735.99*** (−12.97)
GDP	−.0028*** (−2.71)	−.0028*** (−2.11)	−.0031** (−2.51)	−.0027*** (−2.89)	−.0027*** (−2.05)	−.0031** (−2.53)
INF	6484.689*** (2.64)	6484.689*** (2.26)	5132.888* (1.88)	6447.039*** (2.61)	6447.039*** (2.25)	5037.951* (1.89)
Const	18586.31*** (40.62)	18586.31*** (38.37)	13290.54*** (2.96)	19076.99*** (39.28)	19076.99*** (36.73)	16182.07*** (3.66)
No Obs	990	990	990	990	990	990
R-Squared	0.5767	0.5767	0.6362	0.5795	0.5795	0.6515
Prob > F/t	0.0000	0.0000	0.0000	0.000	0.000	0.0000

Panel B. Model Specification Test
		(1) REM – FEM	(2) REM – FEM
Hausman Test	Chi2	137.66	142.22
	Prob > Chi2	0.0000	0.0000

This table presents the testing results of the relationship between fintech lending expansion and credit channeling activity. Panel A shows the testing results of model (1). Columns (1) and (4) display findings using Pooled Ordinary Least Squares. Columns (2) and (5) document results using Random Effects Model. Columns (3) and (6) exhibit the testing results using Fixed Effects Model. Panel B shows a model specification test. The t and z-values are in parentheses. *, **, and *** denote significances at 10%, 5%, and 1% levels, respectively.

As shown by the regression results for credit channeling activity in Table 4, fintech lending expansion is negatively related to credit channeling activity. FEM regression (column 3 in Panel A) produces a parameter of −0.0041, significant at five percent level. The Hausman specification test that we perform, as shown in column (1) in Panel B, suggests that FEM generates more precise results than does REM. These findings indicate that banks are less likely to use fintech lending as their credit channels, thus substantiating our second hypothesis. Fintech lending often targets consumers considered too risky by banks, and relies heavily on fee-based income from online lending transactions. Hence, banks generally consider fintech lending too risky to use as a credit channel since they also have incentives to maintain the trust of lenders [32,59]. Our findings corroborate [113], who reports that Indonesian banks are more likely to use finance and leasing companies as credit channels rather than fintech lending firms.

Meanwhile, the regression results find evidence that bank lending activity positively affects credit channeling activity. FEM regression (column 3 in Panel A) produces a regression coefficient of 0.0030, which is significant at five percent level, thereby supporting our first hypothesis. This finding is robust as all econometric models employed generate similar results. The larger the amount of loans disbursed by banks, the higher is the default risk. Hence, banks need to evaluate and improve their portfolios to diversify risks [19]. We examine further how the pressure of fintech lending expansion affects the credit channeling behavior by incorporating an interaction variable between fintech lending expansion and bank lending activity into the model, as shown in columns (4), (5), and (6) in Panel A. The regression results indicate that the interaction variable positively and significantly affects credit channeling activity. While fintech lending expansion has a negative relation with credit channeling activity, an increase in bank lending activity coupled with fintech lending expansion will increase bank credit channeling activity. This finding is consistent in three econometric methods (columns 4, 5, and 6 in Panel A), supporting [19] who finds that increased competition in the financial sector encourages banks to diversify their risks.

With respect to bank-specific characteristics, our results suggest that larger banks (β = 886.76) and less capitalized banks (β = −1883.06) are more likely to conduct credit channeling activity. Larger banks are typically less capitalized and have less stable funding, thereby more inclined to implement market-based activities to increase profit [56]. These results are consistent with the competition-stability view, which reveals that a concentrated market is more likely to experience a systematic crisis [53] since a concentrated market is dominated by a few large banks [58]. Moreover, we find evidence that less profitable banks (β = −21322.81) and less effective banks (β = 1468.536) tend to undertake a higher level of credit channeling activity. Fintech lending expansion exerts stronger impacts on small banks since smaller banks tend to be less effective and have lower profitability [24]. Small banks are more likely to leverage credit channels in response to more strenuous competition in the financial sector since they are more agile in adjusting their strategies and business models [13]. In addition, our results report that banks with lower loan-to-deposit ratio are more likely to engage in credit channeling activity. According to Ref. [101], regulations on the loan-to-deposit ratio target both sides of the balance sheet, i.e., loans are targeted from the asset side and deposits from the liability side. However, Indonesian banks conduct credit channeling not to avoid regulatory constraints. In this case, Indonesian banks might keep their LDR low due to their concerns over higher interest rates, rupiah depreciation, inflation risk, and new government policies.

Regarding macroeconomic conditions, we find that credit channeling activity is significantly higher in more competitive markets (β = 11454.84). Commensurate with the competition-stability view, this finding indicates that the emergence of fintech lending firms helps control bank credit risk by strengthening competitive channels [24]. Our results also portray that credit channeling activity is higher in lower GDP (β = −0.0031) and higher inflation (β = 5132.888) areas, suggesting that credit channeling encourages economic growth under normal circumstances and accelerates economic recovery during economic crisis [102].

4.2. Impacts of fintech lending expansion on bank NPLs

Following procedures used in testing model (1), we examine the impacts of fintech lending expansion on bank NPLs. Table 5 presents the empirical results of model (2) employing REM, FEM, and GMM. The testing results using REM show that fintech lending expansion is not significantly related with bank NPLs. However, FEM results produce evidence that fintech lending has a negative and significant relation with bank NPLs. On the contrary, a negative and significant relationship between bank lending activity and bank NPLs is found using REM but not in FEM. Hence, we run a two-step GMM to account for potential endogeneity, heteroskedasticity, and autocorrelation issues in the data. The post-estimation test in Panel B in Table 5 shows that the GMM model passes Sargan and Arellano-Bond tests, suggesting that instruments used are valid.

Table 5.

Testing results of model (2).

Panel A. Testing Results
	(1) REM	(2) FEM	(3) GMM	(4) REM	(5) FEM	(6) GMM
FLE	−.0421 (−1.34)	−.3761*** (−7.09)	−.3860*** (−102.32)	−.0074 (−0.22)	−.3268*** (−5.83)	−.3191*** (−51.43)
BLA	−.0067*** (−3.55)	−.0017 (−0.87)	−.0016*** (−22.37)	.0019 (0.58)	0.0055 (1.63)	.0057*** (44.66)
FLExBLA				−2.67e-06*** (−3.04)	−2.36e-06*** (−2.64)	−2.41e-06*** (−104.93)
Size	77.0363*** (4.30)	227.7881 (0.51)	−407.7045** (−2.08)	73.7225*** (4.12)	240.0806 (0.53)	242.7738 (0.92)
CR	−108.8589** (−2.05)	−549.649 (−1.06)	−708.1393*** (−7.18)	−98.2997* (−1.86)	−566.7633* (−1.66)	−1021.983*** (−4.92)
ROA	−17.265.29*** (−3.85)	−19033.35*** (−4.02)	−21482.06*** (−33.42)	−17019.51*** (−3.81)	−18636.87*** (−4.13)	−20237.64*** (−30.15)
LDR	554.8673 (1.40)	857.5975* (1.68)	178.9912 (0.28)	558.6584 (1.41)	813.9727 (1.60)	1380.081*** (8.46)
BIE	−235.819 (−0.36)	1299.322 (1.36)	1799.836*** (6.30)	−535.6768 (−0.81)	915.6105 (0.95)	991.9816*** (4.65)
HHI	268.2303 (0.59)	−3691.386*** (−3.01)	−6608.635*** (−11.29)	669.5514 (1.43)	−3023.122** (−2.42)	−4103.491*** (−4.04)
GDP	0.0039*** (2.56)	.0035** (2.39)	.0036*** (73.25)	.0040*** (2.67)	.0037** (2.52)	.0036*** (53.65)
INF	1879.103 (0.57)	885.1588 (0.27)	2566.116*** (3.42)	1674.924 (0.51)	766.6494 (0.24)	820.421 (0.70)
Const	−1076.8* (−1.93)	69.6661 (0.01)	10042.32*** (3.68)	1674.924** (−2.15)	−336.534 (−0.06)	603.8707 (0.18)
No Obs	990	990	924	990	990	924
R-Squared	0.0671	0.0325	–	0.0758	0.0400	–
Prob > F/t	0.0000	0.0000	0.000	0.000	0.000	0.000

Panel B. Post-Estimation Test
		Without Interaction	With Interaction
Sargan of Overidentifying Restrictions	Chi2	24.4491	22.6097
	Prob > Chi2	1.000	1.000
Arellano-Bond test for AR (1)	z	−1.3883	−1.3886
	Prob > z	0.1650	0.1650
Arellano-Bond test for AR (2)	z	−1.0485	−1.0787
	Prob > z	0.2944	0.2807

This table presents the testing results of the relationship between fintech lending expansion and bank NPLs. Panel A presents the results of model (2). Columns (1) and (4) display the findings using Random Effects Model. Columns (2) and (5) report the results using Fixed Effects Model. Columns (3) and (6) exhibit the results using the two-step GMM. Panel B documents post-estimation tests. The t and z-values are in parentheses. *, **, and *** denote significances at 10%, 5%, and 1% levels, respectively.

As reported by the testing results using the two-step GMM (column 3 in Panel A), we find a negative and significant relationship between bank lending activity and bank credit risk (β = −.0016). Therefore, our third hypothesis is confirmed. This finding substantiates the conjecture that increased competition in the financial sector encourages banks to work more efficiently [13,103]. Meanwhile, we find evidence that fintech lending expansion is negatively and significantly related with bank NPLs (β = −0.3860), thus supporting our fourth hypothesis. This indicates that fintech lending expansion reduces credit risk borne by banks, and helps banks screen lower credit-rating borrowers since lower credit-rating borrowers are more likely to use fintech lending to access credits [12]. This finding also suggests that instead of competing with banks, fintech lending firms tend to complement banks in providing a broader range of services in the financial sector [9,32].

Several recent studies reveal that fintech lending could help banks reduce asymmetric information in the credit market, thereby reducing bank credit risk [24,26]. We examine this notion by employing an interaction variable between fintech lending expansion and bank lending activity on bank NPLs, as shown in columns (4), (5), and (6) in Panel A. The results show that the interaction variable negatively and significantly affects bank NPLs. Greater fintech lending expansion, coupled with higher bank lending activity, tend to lower bank credit risk. This finding posits a very pivotal implication. Although an increase in loan ratio will be followed by an increase in interest income and also a higher credit risk [103], increased competition in the financial sector encourages banks to work more efficiently, especially after the flourishing of fintech lending companies.

With respect to bank-specific characteristics, we find that smaller banks (β = 407.7045) tend to have more NPLs. Larger banks have better staffing as well as a more efficient and standardized credit process, resulting in more sound credit risk management [103]. We also observe that less capitalized (β = −708.1393) and less efficient (β = 1799.836) banks are more likely to experience default risk. This finding corroborates [104], who shows that higher capitalized banks in ASEAN countries are more efficient, thus bearing less credit risk. In addition, our results find proof that less profitable banks tend to have higher NPLs (β = −21482.06). More profitable banks are more vigilant with credit disbursement process so as to manage credit risk in a sturdier fashion [105].

Pertaining to macroeconomic conditions, our regression results show that banks in a more competitive market are more likely to experience credit default risk (β = −6608.635). This supports the bad management hypothesis [106], which states that highly profitable banks have less incentives to engage in high-risk activities, including excessive loan growth. In addition, our empirical findings show that GDP growth (β = 0.0036) and inflation rate (β = 2566.116) exacerbate bank NPLs. Higher inflation rate and GDP growth will trigger policymakers to increase the policy rate to curb inflation. However, increasing interest rates will also encourage banks to spur their lending activities due to excessive deposits on the liabilities side. As increasing lending activity might be followed by a higher default risk [103], a soaring inflation rate and GDP growth may escalate bank NPLs.

Our overall findings confirm the competition-stability view. Fintech lending expansion encourages banks to diversify their risks by utilizing credit channels to expand their portfolios. Fintech lending expansion also motivates banks to work more efficiently in response to increased competition in the financial sector. Moreover, our empirical evidence suggests that fintech lending not only complements banks in providing a broader range of services in the financial sector, but also helps banks screen lower credit-rating borrowers. Such an environment reduces total bank risk, and hedges the financial systems from systemic shocks.

4.3. Fintech lending limits based on MSME characteristics

In the previous section, we have discussed that banks are less likely to utilize fintech lending firms as their credit channels since fintech lending is considered too risky by banks. However, fintech lending limits should be considered and evaluated to prevent disturbances to financial system stability if banks harness fintech lending as a shadow-banking mechanism. This section examines and estimates appropriate fintech lending limits based on the characteristics of MSMEs. The salient purpose of the lending-limit policy is to maximize loans given to MSMEs while minimizing the probability of default. Table 6 reports the results of the threshold regression of MSME characteristics on fintech lending limits.

Table 6.

Threshold regression results.

	(1)	(2)
Intercept (1st regime)	0.4480*** (15.73)	2.029 (1.60)
Intercept (2nd regime)	0.6159*** (9.66)	3.565** (2.37)
Intercept (3rd regime)	0.1095*** (4.24)
Intercept (4th regime)	0.0327 (−0.54)
Education	2.841 (1.38)	2.398 (1.62)
Education2	3.950 (3.44)	−0.301 (−1.62)
Business_age	0.0060* (2.77)	−9.107*** (−4.17)
Business_age2	0.0050 (1.63)	1.960*** (4.96)
Threshold (1)	100,000,000	2,000,000,000
Threshold (2)	200,000,000	6,000,000,000
Threshold (3)	5,000,000,000
R-Square	0.1246	0.138
N	344	344

This table presents the results of the threshold regressions of profitability and risk (leverage and prudent behavior) on credit needs (loan demand). Column (1) shows the results of the profitability threshold regression with credit needs being the threshold variable. Column (2) reports the results of the leverage threshold regression with credit needs being the threshold variable. Column (3) shows the results of debtor's reputation threshold regression with credit needs being the threshold variable. The t and z-values are in parentheses. *, **, and *** denote significances at 10%, 5%, and 1% levels, respectively.

Column (1) of Table 6 presents the results of the threshold regression of profitability on credit needs. It can be seen that profitability is significantly related to credit needs in regimes one to three, before the relationship becomes insignificant in the fourth regime. These results verify our fifth hypothesis. In general, evidence from the threshold regression of profitability on credit needs depicts that credits disbursed by fintech lending up to IDR5 billion per borrower would increase MSME profitability. However, credits above IDR5billion per borrower will not increase the profitability of MSMEs. This finding is in line with an inverted U-shaped relationship between profitability and capitalization in both MSMEs [84] and large firms [107]. Our results do not find a significant relationship between profitability and the level of education, implying that higher education of borrowers is not always a predictor of profitability. This is similar to Ref. [108], who conclude that the level of education does not always lead to higher profitability. Meanwhile, our results show that younger firms tend to generate higher profit, supporting [109,110].

Column (2) of Table 6 presents the results of the threshold regression of leverage on fintech credit limits. Firm leverage is not significantly related to credit needs in the first regime, but this relationship becomes significantly positively in the second regime, thus supporting our sixth hypothesis. This suggests that loans above IDR6 billion per borrower will result in “overheating” leverage for MSMEs, so it increases the probability of default. In addition, we find that older firms are more inclined to have higher financial leverage. These findings correspond with [111], who reveal that financial leverage affects the intensity of firm performance as firm ages. Overall, our empirical findings from the threshold regressions indicate that a reasonable credit limit for fintech lending firms in Indonesia is between IDR5 billion and IDR6 billion. In this regard, IDR5 billion is the maximum loan channeled to each borrower to induce profitability of the MSME. On the other hand, IDR6 billion refers to the maximum loan disbursed to each borrower without increasing the probability of default. Since a loan amount above IDR5 billion per borrower no longer enhances MSME profitability, IDR5 billion could be deemed the most reasonable fintech lending limit. As MSMEs are financially constrained [30,112], assessing fintech lending limits based on the characteristics of MSMEs allows the fintech lending firms to provide MSMEs with sufficient funding to grow, which in turn will spur economic growth. At the same time, determining credit limits based on MSME characteristics also helps banks monitor and maintain their asset quality, thereby minimizing risks and financial instability if banks utilize fintech lending as their credit channels.

5. Conclusion and future research agenda

5.1. Conclusion

The fintech lending industry has experienced significant development over time. However, the impacts of fintech lending on banks remain inconclusive. The competition-instability proponents believe that fintech lending expansion erodes bank market and threatens banks as traditional intermediaries, thus intensifying bank risk-taking and potentially disturbing financial stability. In contrast, competition-stability enthusiasts believe fintech lending reduces asymmetric information in the credit market, thereby reducing bank risk-taking and increasing bank resilience to systematic shocks. To shed some light on these issues, this paper examines the impacts of fintech lending expansion on bank risk-taking behavior. We employ two approaches: (1) credit channeling activity and (2) nonperforming loans (NPLs) of banks. Credit channeling activity represents credit channels’ utilization intensity, which is proxied by increases in the total amount of funds placed by banks in nonbank financial institutions. Meanwhile, bank NPLs reflect bank effectiveness and the degree of risk-taking measured as an increase in NPLs over the year.

Overall, our main findings substantiate the competition-stability view. With respect to the relationship between fintech lending expansion and credit channeling activity, our results confirm the notion that fintech lending expansion encourages banks to diversify their risks but are less likely to use fintech lending as their credit channels since fintech lending borrowers are considered high-risk by the banks. Regarding the relationship between fintech lending expansion and bank risk, our results indicate that increased competition in the financial sector motivates banks to work more efficiently. In addition, fintech lending helps banks screen borrowers by attracting lower credit-rating borrowers, suggesting that fintech lending tends to complement, rather than compete with, banks in providing a broader range of services in the financial sector.

With regard to bank-specific characteristics, our empirical findings suggest that larger banks with less capitalized and less stable funding are more likely to implement market-based activities to increase profit. Meanwhile, larger with higher capitalized banks tend to have better staffing and credit process, thus working more efficiently, having better control over NPLs, and bearing less credit risk. Furthermore, less profitable and less efficient banks are more inclined to utilize credit channels and have a higher probability of default, suggesting that more profitable and efficient banks are more rigorous with the credit disbursement process.

Regression results regarding macroeconomic conditions report that banks are more likely to utilize credit channels and have higher NPLs in more competitive markets, which suggests that competition controls excessive loan growth by strengthening competitive channels. Moreover, credit channeling activity is found to be higher in lower GDP and higher inflation areas, indicating that credit channels support economic growth under normal conditions while accelerating economic recovery during financial crisis. However, excessive GDP growth and inflation rates could also exacerbate NPLs due to the increased lending activity of banks in response to monetary regulations in reducing money supply.

This study also evaluates and examines appropriate fintech credit limits based on MSME characteristics, i.e., profitability and leverage. We find that loans disbursed by fintech lending up to IDR5 billion per borrower would increase MSMEs' profitability. Nevertheless, credits above IDR5 billion per borrower will not improve the profitability of MSMEs. Meanwhile, loans above IDR6 billion per borrower might engender excessive leverage for MSMEs, which is more likely to cause insolvency. This evidence corroborates the conjecture that the relationship between credit and performance is inverted U-shaped. Our findings from the threshold regressions of MSME characteristics indicate that a proper credit limit for fintech lending firms in Indonesia is between IDR5 billion and IDR6 billion per borrower. Within this perspective, IDR5 billion per borrower is the maximum amount of loans that is able to spur MSMEs’ profitability whereas IDR6 billion per borrower is the maximum lending amount that could prevent the default risk from soaring.

5.2. Future research agenda

This research purports to examine the impacts of fintech lending expansion on bank risk-taking behavior using credit channeling activity and NPLs. However, as the sample used in this study are banks in 34 states in Indonesia, its findings only portray how fintech lending expansion affects bank risk-taking in Indonesian financial services industry. Testing this relationship further in the global context will surely enrich the literature as it can better describe the effects of increased competition due to fintech lending expansion on banks in other parts of the world. In addition, this study is bound by a piece of regulation by the Indonesian Financial Services Authority where the current limit lending by a fintech firm is IDR2 billion per borrower. Such circumstances may not enable us to fully capture the real dynamics of the relation between fintech lending expansion and bank risk-taking behavior.

Author contribution statement

Eddy Junarsin: Conceived and designed the analysis; Analyzed and interpreted the data; Wrote the paper. Rizky Yusviento Pelawi: Conceived and designed the analysis; Analyzed and interpreted the data; Wrote the paper. Jordan Kristanto: Analyzed and interpreted the data; Contributed analysis tools or data; Performed data collection. Isaac Marcelin: Analyzed and interpreted the data; Contributed analysis tools or data. Jeffrey Bastanta Pelawi: Analyzed and interpreted the data; Performed data collection.

Data availability statement

Data will be made available on request.

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.