No Guarantee, Still Gains? Rethinking the Relation Between Loans Without Personal Guarantees and Bank Profitability

Abstract

Japanese financial authorities have been promoting lending without personal guarantees, with the aim of enhancing the stability of the banking sector. However, banks’ willingness to provide loans without personal guarantees varies. Furthermore, no prior study has directly examined the impact of such loans on bank profitability. Since Japanese regional banks recently began disclosing the proportion of the concerned loans, this study aims to clarify their impacts on bank profitability, utilizing 352 bank/year observations from fiscal years 2019 to 2023. The results indicate that the statistical significance of the impacts of the proportion of loans without personal guarantees on bank profitability is not robust, and that the impacts are economically insignificant. This study contributes to the literature by providing the first empirical evidence on the impact of personal guarantees on bank profitability. It also offers practical insights to financial authorities, demonstrating that loans without personal guarantees do not significantly affect the profitability of regional banks, at least during the examined sample period.

1. Introduction

In Japan, it is common practice for banks to require personal guarantees for loans to small and medium enterprises (SMEs). Figure 1 illustrates the percentage of companies that met the relevant conditions for loans from their main banks, based on the “Basic Survey on SMEs”1. Between 2004 and 2010, the percentage of companies with collateral-backed loans declined from 50% to 40%. Conversely, nearly 80% of respondent companies continued to provide personal guarantees.

Meanwhile, Japanese banks have faced increasing pressure to expand lending due to declining profitability, partly driven by the Bank of Japan’s monetary easing measures, including prolonged low interest rates. Financial authorities in Japan have promoted cash-flow-based lending that does not rely on collateral or personal guarantees. In its Financial Monitoring Policy published in September 2013, the Japan Financial Services Agency (JFSA) stated that it “will try to identify obstacles behind banks’ cautious approaches for taking credit risks without collaterals and guarantees” (Japan Financial Services Agency, 2013b). Since then, the JFSA has sought to strengthen the capacity of regional banks, particularly their ability to evaluate the profitability and risks of local businesses with the aim of revitalizing regional economies and stabilizing bank profitability.

In December 2013, the Japanese Bankers Association and the Japan Chamber of Commerce and Industry issued the “Guidelines on Business Owners’ Personal Guarantee,” establishing voluntary rules to facilitate lending without personal guarantees (Japan Study Group on Guidelines for Personal Guarantee Provided by Business Owners, 2013). Despite these efforts, the JFSA observed substantial variation in the efforts and effectiveness of business evaluations conducted across banks. In 2016, official benchmarks were established to evaluate financial intermediation functions using objective evaluation criteria (Japan Financial Services Agency, 2016). Furthermore, in 2019, the JFSA encouraged financial institutions to actively disclose selected indicators, such as “the percentage of new loans without personal guarantees,” to promote transparency (Japan Financial Services Agency, 2019). Consequently, most regional banks began disclosing these indicators as key performance indicators (KPIs).

Prior studies suggest that personal guarantees function similarly to collateral by helping mitigate problems arising from information asymmetry. However, they are generally less effective in transferring risk from borrower to lender (Cadot, 2013; Ono et al., 2012; Ono & Uesugi, 2009). Prior studies have reported that non-performing loan (NPL) ratio, bank size, cost-to-income ratio, equity-to-assets ratio, and non-interest income determine bank profitability (Athanasoglou et al., 2008; Dietrich & Wanzenried, 2011; Kumar et al., 2020, 2022; Liu & Wilson, 2010). While much of the existing literature focuses on the loan contract level, few studies have examined the broader impact on bank profitability. This study fills that gap by analyzing the effect of loans without personal guarantees on bank profitability. This study employs three types of bank profitability measures: return on assets (ROA), return on equity (ROE), and return on loans. This is in light of the Japanese financial authorities’ continued efforts to promote cash-flow-based lending as a means to enhance both financial intermediation and the long-term profitability of banks.

This study addresses two main research questions. The first is whether loans without personal guarantees significantly affect bank profitability. The second concerns the role of bank management quality. Specifically, it examines whether the positive impact of such loans on profitability is more pronounced in banks with high-quality management, which are better equipped to conduct effective screening and monitoring.

To investigate these questions, the study employs two-way fixed effects models that control for both bank-specific and year-specific effects, using bank-clustered robust standard errors. Additionally, two-step Generalized Method of Moments (GMM) estimators are applied to address potential endogeneity issues. These methodological approaches enhance the robustness and credibility of the findings.

The empirical findings indicate that the statistical significance of the effects of the concerned loans on bank profitability is not robust. The findings also show that their average effects are less than the standard deviations of the profitability, indicating that economically insignificant. Although the effects vary slightly based on non-performing loan ratios, the marginal effects remain limited. These findings suggest that some banks may be hesitant to expand such lending due to potential credit risks, as the JFSA identified differences in financial intermediation efforts across banks (Japan Financial Services Agency, 2016).

This study contributes to literature in two ways. First, it extends prior research by examining the aggregate impact of loans without personal guarantees on bank profitability, rather than focusing solely on individual loan contracts. This study is the first to empirically examine the impact of such loans on bank profitability in the Japanese context. Second, this study offers practical insights for policymakers. Despite over a decade of promotion by Japanese financial authorities, such loans do not appear to enhance bank profitability. While reducing personal guarantees may support SMEs’ succession, their effectiveness as a KPI for bank performance remains questionable.

The remainder of this study is organized as follows. Section 2 outlines the recent institutional background in Japan. Section 3 reviews previous studies and presents the research questions. Section 4 describes the research design. Section 5 reports empirical results, and Section 6 examines the robustness of the findings with respect to endogeneity. Section 7 concludes.

2. Institutional Background

This section outlines key initiatives introduced by financial authorities to promote cash-flow-based lending and encourage Japanese financial institutions, particularly regional banks, to offer loans without personal guarantees from business owners.

2.1. Types of Commercial Banks and Activities of Regional Banks

Traditionally, Japanese commercial banks have three core services: deposits, lending, and exchange. Under Japanese law, including the Banking Act, financial institutions that manage deposits are classified as “deposit-taking financial institutions,” and are subject to strict supervision and regulation. This category includes ordinary banks and cooperative financial institutions, such as Shinkin banks and credit cooperatives.2 Ordinary banks, as defined under the Banking Act, include city banks, regional banks (including regional banks II3), foreign banks, and other types of banks.

This study focuses on regional banks, given their significant role in domestic lending to SMEs, particularly in local cities.

Table 1. Trends in domestic loans and deposits by type of financial institutions.

Panel A: Loans, average amounts outstanding
	Major banks		Regional banks		Shinkin-banks
	Amounts	Share (%)	Amounts	Share (%)	Amounts	Share (%)
May 2000	291	53.4	185	33.9	69	12.7
May 2005	206	46.9	172	39.2	61	13.9
May 2010	203	44.2	193	42.1	63	13.6
May 2015	203	41.8	220	45.2	63	13.0
May 2020	232	41.2	259	46.2	71	12.6
May 2025	253	39.9	303	47.8	78	12.3
Panel B: Deposits, average amounts outstanding
	City banks		Regional banks		Shinkin-banks
	Amounts	Share (%)	Amounts	Share (%)	Amounts	Share (%)
May 2000	235	41.2	234	40.9	102	17.9
May 2005	256	42.2	243	40.0	108	17.8
May 2010	275	41.5	269	40.7	118	17.8
May 2015	317	41.4	316	41.3	132	17.3
May 2020	416	45.3	356	38.7	147	16.0
May 2025	484	45.7	414	39.0	162	15.3

Note: The table reports the amounts and shares of domestic loans and deposits by financial institution type. Amounts are expressed in trillions of yen. Data are sourced from “Principal Figures of Financial Institutions” available on the website of the Bank of Japan. In the reports, the bank categories differ slightly between loans and deposits. Major banks in Panel A include city banks in Panel B and trust banks.

shows lending and deposit trends by bank type.4 The total amount of loans extended by regional banks rose from 185 trillion yen in 2000 to 303 trillion yen in 2025, with their share of total lending increasing from 33.9% to 47.8%. Although their share of total deposits declined marginally, from 40.9% to 39.0%, the absolute amount of deposits increased from 234 trillion yen in 2000 to 414 trillion yen in 2025.

Table 2. The number of financial institutions.

	City Banks	Regional Banks	Shinkin Banks
FY 1995	11 (+0.0%)	129 (+0.0%)	416 (+0.0%)
FY 2000	9 (−18.2%)	121 (−6.2%)	372 (−10.6%)
FY 2005	6 (−45.5%)	111 (−14.0%)	292 (−29.8%)
FY 2010	6 (−45.5%)	105 (−18.6%)	271 (−34.9%)
FY 2015	5 (−54.5%)	105 (−18.6%)	265 (−36.3%)
FY 2020	5 (−54.5%)	100 (−22.5%)	254 (−38.9%)
January 2025	5 (−54.5%)	97 (−24.8%)	254 (−38.9%)

Note: The table reports the number of financial institutions for each fiscal year. Percentage in parentheses indicates the change from FY 1995. Data up to FY 2020 are sourced from “the trends in the number of insured financial institutions”; data as of January 2025 are obtained from “the number of insured financial institutions” available on the website of the Deposit Insurance Corporation Japan (in Japanese).

illustrates the trends in the number of insured financial institutions under Japanese Deposit Insurance Act.5 In the wake of active mergers and acquisitions in the banking sector, the number of city banks and Shinkin banks has declined by approximately 54.5% and 38.9%, respectively, over the past 30 years. In contrast, the number of regional banks decreased by only about 24.8%. This relatively modest decline suggests that regional banks have maintained a stable presence by building strong relationships with high-quality local clients in their respective regions.

2.2. Policy Initiatives from Financial Authorities in Japan Since 2013

Regional banks have traditionally sustained their profitability by offering core banking services to local businesses and individuals.6 Conversely, local cities have been experiencing a marked decline in working-age population, a trend expected to reduce the number of employees and business establishments, thereby shrinking the regional market for banking services.

Furthermore, Japanese regional banks have been facing declining lending profitability driven by narrowing interest margins under prolonged low interest rates. In April 2013, Bank of Japan introduced the “Quantitative and Qualitative Monetary Easing” at its Monetary Policy Meeting, including the adoption of monetary base control, followed by a series of additional monetary easing measures.7 These policies are aimed at stimulating domestic demand and achieving the 2% target inflation rate; however, banks often struggle with declining profitability driven by low interest rates.

Given a low interest-rate environment, how can Japanese regional banks sustain profitability? The JFSA encourages the promotion of cash-flow-based lending, supported by proper credit assessments of businesses, to better meet potential funding needs. In September 2013, the JFSA published its annual policy report titled “New Financial Monitoring Policy.” For the first time, it integrated insights from both on-site inspections and off-site supervision of financial institutions (Japan Financial Services Agency, 2013b). In this report, the JFSA emphasized the importance of banks’ loan screening skills and expressed a strong commitment to engaging in discussions with banks on key issues. Specifically, these issues include banks’ internal factors that hinder appropriate risk-taking without over-reliance on collateral and guarantees, as well as the effective promotion of cash-flow-based lending. The JFSA further emphasized that this initiative helps strengthen banks’ own management foundations by fostering the sustainable growth of borrowers (Japan Financial Services Agency, 2025), which can be expected to improve bank profitability.

2.3. Guidelines for Personal Guarantee Provided by Business Owners

In the same year, a study group on the provision of personal guarantees for SMEs was established (Japan Financial Services Agency, 2013a). This study examines the role of personal guarantees in supporting SMEs. The central issue is exploring alternative lending practices that reduce reliance on personal guarantees and collateral. In May 2013, the group published its final report, noting that personal guarantees in lending can be beneficial, as they strengthen governance through managerial discipline, facilitate financing for SMEs, and reduce capital costs. The report also highlights the potential drawbacks of personal guarantees, including reduced borrower motivation to disclose timely and appropriate financial and management information to banks, and diminished incentives for conducting judicious screening of loans. The report proposes specific measures to address these issues and facilitate loan contracts more effortlessly, without requiring personal guarantees.

To implement the measures outlined in the final report, the Japanese Bankers Association and the Japan Chamber of Commerce and Industry established a study group in August 2013 (Japan Study Group on Guidelines for Personal Guarantee Provided by Business Owners, 2013). The group published the “Guidelines for Personal Guarantee Provided by Business Owners” in December 2013, establishing voluntary rules between financial institutions and business owners to facilitate lending without personal guarantees. The guidelines contributed to a significant increase in loans without personal guarantees (Uesugi et al., 2018).

2.4. Benchmark Indicators on Financial Intermediation

In 2016, the JFSA introduced the “Benchmark Indicators on Financial Intermediation” to objectively evaluate the role of banks in financial intermediation (Japan Financial Services Agency, 2016). Through inspections and supervision, the JFSA has sought to understand how banks promote smooth financial intermediation, including cash-flow-based lending. However, the JFSA found substantial variation in “the extent to which banks implemented financial intermediation practices.” During interviews, SMEs expressed strong concerns that banks continued to rely excessively on collateral and guarantees, with little change in their lending approach. Therefore, the JFSA recognized the need for an objective evaluation of banking practices and established benchmarks.

The JFSA proposed two types of benchmark indicators: common and optional (Japan Financial Services Agency, 2016). Common benchmarks are designed to assess banks’ efforts in promoting cash-flow-based lending. For instance, these benchmarks assess the number and amount of such loans relative to total lending. Meanwhile, optional benchmarks enable banks to conduct more appropriate and detailed self-assessments, considering differences in their business characteristics and management strategies. For instance, one of the optional benchmarks is the proportion of loans without personal guarantees of local SMEs, which constitute the bank’s main clientele. The JFSA also encourages banks to develop and disclose customized benchmarks to support more accurate self-assessment.

2.5. Data Availability of Two KPIs in Japan

Regional banks have published both common and optional benchmarks since FY 2017. However, as disclosure was not legally mandated and lacked a standard method, banks adopted various methods of disclosure. Some banks posted data on designated sections of their specified websites, while others included it in their digital publications. This inconsistency made it difficult to access and compare data across banks.

In June 2019, the Cabinet in Japan approved the document “Follow up on the Growth Strategy” and defined two KPIs to assess the effectiveness of financial intermediation (Japan Cabinet Secretariat, 2019). The first KPI is the percentage of guarantee requests during business succession, while the second captures the proportion of new loans that do not rely on business owner guarantees (Japan Cabinet Secretariat, 2019). As the document also emphasized the importance of visualizing both KPIs, the JFSA published detailed definitions (Japan Financial Services Agency, 2019). It also expressed its expectation that both major and regional banks would voluntarily disclose trends in these KPIs to the greatest extent possible (Japan Financial Services Agency, 2019). Subsequently, many regional banks began disclosing the two KPIs through their publications rather than as part of common and optional benchmarks. Furthermore, the JFSA began publishing semi-annual data on both KPIs for major and regional banks on its website, starting from the second half of FY 2019. It later began disclosing annual data from FY 2023 (Japan Financial Services Agency, 2024).

At the time of writing, time-series data on both KPIs of regional banks in Japan are publicly available online.8 This study focuses on the second KPI, the percentage of new loans that do not rely on business owner guarantees. Annual data for FY 2023 published by the JFSA were obtained, along with disclosure materials, such as disclosure reports and integrated reports available on the websites of 100 regional banks as of FY 2023. 87.4% (437 of 500 observations) of the second KPI data from FY 2019 to FY 2023 were found to be available. These data have been archived in a data repository to ensure public accessibility and transparency (Yoshinaga & Onishi, 2025).

3. Literature Review and Research Questions

This section reviews prior studies on the role of personal guarantees in lending and identifies key determinants of bank profitability discussed in the literature. Building on these insights, the research questions of this study are formulated.

3.1. The Role of Personal Guarantees

Personal guarantees provided by business owners in loan contracts serve a direct role similar to collateral, transferring a portion of default risk from the lender to the borrower. In the presence of information asymmetry, personal guarantees also play an indirect role in mitigating adverse selection and moral hazards. Creditworthy borrowers offer collateral or personal guarantees to signal their reliability (Bester, 1985; Manove et al., 2001). Both collateral and personal guarantees offer borrowers an incentive to impose financial discipline on management to gain the trust of banks (Boot & Thakor, 1994). Jimenez et al. (2006) utilize Spanish data to demonstrate that reliable borrowers are more likely to offer collateral when lenders cannot directly observe borrowers’ creditworthiness. Ono et al. (2012) examine the impact of collateral on borrower performance after loan contracts, reporting that collateralized loans improve borrowers’ profitability and reduce risk. These findings support the view that collateral helps mitigate both adverse selection and moral hazard.

However, heavy reliance on collateral and personal guarantees tends to reduce banks’ efforts to review loan applications, resulting in inefficient lending practices. Manove et al. (2001) theoretically show that the use of collateral in loan contracts can reduce the screening efforts of banks below the socially optimal level under information asymmetry. This could potentially lead to the financing of unviable projects. Cadot (2013) analyzes bank lending to wine farmers in France. He empirically demonstrates that loans secured by both collateral and bank monitoring are associated with an increased risk of repayment delays, whereas loans involving either collateral or monitoring alone reduce the risk. This suggests that the presence of collateral diminishes banks’ monitoring efforts.

Meanwhile, as personal guarantees are less effective than collateral in transferring risk from the borrower, previous studies have found their impact to be relatively weaker. Ono and Uesugi (2009) report no significant relationship between credit scores and the use of personal guarantees. Similarly, Ono et al. (2012) find that pledging personal guarantees does not significantly enhance borrowers’ profitability. Peltoniemi and Vieru (2013) demonstrate that personal guarantees are more likely to be required in transaction-based loans characterized by high information asymmetry. They also report that such loans carry significantly higher interest rates compared to relationship-based loans when personal guarantees are included. Cadot (2013) finds that personal guarantees have no significant impact on repayment delays. Uesugi et al. (2018) find that borrowers who obtain loans without personal guarantees, such as those with lower debt ratios and higher internal ratings from lenders, tend to be financially sound. Furthermore, they determined that borrowers using such loans after 2011 improved their creditworthiness and reduced default risks. These findings suggest that personal guarantees are not necessarily effective in preventing adverse selection or moral hazard.

3.2. Determinants of Bank Profitability

Previous studies have identified multiple factors influencing bank profitability.9 The first factor is the NPL ratio, which reflects higher credit risk and lower management quality (Louzis et al., 2012; Podpiera & Weill, 2008), thereby reducing the bank’s overall profitability. Therefore, this factor is expected to have a negative impact on bank profitability. Consistent with this expectation, previous studies in Japan find a negative impact of NPLs on bank profitability (Kumar et al., 2022). Similarly, the significantly negative relation between NPLs and bank profitability is reported in South Asian countries (Nisar et al., 2018), Middle East and North Africa regions (Alnabulsi et al., 2022) and Bangladesh (Akhter, 2023). Conversely, studies from New Zealand and Italy indicate that the relation between NPLs and bank profitability is statistically insignificant or positive (Kumar et al., 2020; Pancotto et al., 2024).

The second factor is bank size. Larger banks are generally expected to achieve higher profitability based on economies of scale. However, smaller banks may also be profitable by lending to higher-risk customers. Therefore, previous studies have reported mixed results regarding the relationship between bank size and profitability (Athanasoglou et al., 2008; Kumar et al., 2020; Liu & Wilson, 2010; Mehzabin et al., 2023). Recently, bank size has been regarded as an indicator of a bank’s ability to adapt to changes in the regulatory environment. The implementation of Basel III has reportedly had a positive impact on the profitability of large banks, while exerting a negative impact on the profitability of smaller banks (Gržeta et al., 2023).

The third factor is the cost-to-income ratio, which reflects a bank’s operational efficiency. Prior studies have consistently found that a higher cost-to-income ratio significantly reduces banks’ profitability (Dietrich & Wanzenried, 2011; Kumar et al., 2020, 2022; Liu & Wilson, 2010).

Equity-to-assets ratio has been identified as the fourth factor influencing bank profitability, although prior studies reported mixed results regarding its effect. A high equity-to-assets ratio enables banks to pursue business opportunities more effectively and absorb unexpected losses, potentially increasing profitability (Athanasoglou et al., 2008; O’Connell, 2023). In contrast, Dietrich and Wanzenried (2011) report a significantly negative relationship between the equity-to-assets ratio and bank profitability. They argue that safer banks attracted additional savings deposits during the financial crisis but were unable to convert them into significantly higher earnings due to reduced lending demand during this period. Liu and Wilson (2010) report mixed findings regarding the relationship between equity ratio and profitability.

The fifth factor is non-interest income. In the United States, deregulation has enabled banks to offer non-traditional services and increase non-interest income. While non-interest services can improve bank profitability (Mehzabin et al., 2023), they also tend to increase risk significantly, resulting in less favorable risk-return trade-offs on average (DeYoung & Rice, 2004; Stiroh & Rumble, 2006). In Europe, non-interest income is associated with a decline in profitability and increased risk among small banks, a trend attributed to their limited experience in generating such income (Mercieca et al., 2007). However, evidence from Japanese regional banks shows mixed effects of revenue diversification on profitability (Liu & Wilson, 2010). When the ratio of non-interest income is low, it tends to have a positive effect on profitability (Gambacorta et al., 2014). Higher levels of non-interest income have been linked to both higher profitability and reduced risk, especially during the COVID-19 pandemic (Li et al., 2021). Furthermore, the impact on profitability varies between fee and commission income and other forms of non-interest income (Nisar et al., 2018).

3.3. Research Questions

For over a decade, Japanese financial authorities have encouraged lending practices that do not rely on personal guarantees. However, the impact of loans without personal guarantees on bank performance remains underexplored. This study examines the impact of such loans on bank profitability, considering the recent advances in disclosure among Japanese regional banks.

If banks can appropriately evaluate borrowers’ creditworthiness without relying on personal guarantees, they may enhance profitability by charging higher interest rates. Since companies that do not provide personal guarantees tend to achieve higher profitability after loan agreements, compared to those that do (Uesugi et al., 2018), they are likely to afford the higher interest rates. Meanwhile, loans without personal guarantees are also expected to involve higher screening and monitoring costs to mitigate adverse selection and moral hazard, which could negatively affect bank profitability. Moreover, if the impact of such loans is negligible, no significant impact will be observed. Accordingly, this study poses its first research question (RQ1) without hypothesizing the direction of the effect:

RQ1. 

Do loans without personal guarantees have a significant impact on bank profitability?

The impact of such loans on bank profitability is influenced by the quality of bank management. Banks with high-quality management are more likely to benefit from such loans through effective screening and monitoring. Several studies have shown that bank profitability is significantly affected by non-performing loan ratio (Kumar et al., 2022; Nisar et al., 2018), indicating higher credit risks and lower management quality (Louzis et al., 2012; Podpiera & Weill, 2008). Accordingly, this study poses the second research question (RQ2):

RQ2. 

Is the positive impact of loans without personal guarantees on bank profitability significant in banks with high-quality management?

4. Materials and Methods

This section presents the research design to address the research questions, describing the regression models employed, the data sources used, and the definitions of all variables.

4.1. Models

The statistical analyses in this study are conducted using Stata version 19.5. To explore the research questions, this study employs two-way fixed effects models that incorporate both bank-specific and year-specific fixed effects, with bank-clustered robust standard errors.10 Equations (1) and (2) are used to evaluate RQ1 and RQ2, respectively.

$$\begin{array}{cc}\hfill PRO{F}_{i,t}& =\alpha +{\beta }_{1}NP{G}_{i,t}+{\beta }_{2}NP{L}_{i,t}+{\beta }_{3}SIZ{E}_{i,t}+{\beta }_{4}C{R}_{i,t}\hfill \\ & +{\beta }_{5}E{R}_{i,t}+{\beta }_{6}REN{T}_{i,t}+\epsilon \hfill \end{array}$$

(1)

$$\begin{array}{cc}\hfill PRO{F}_{i,t}& =\alpha +{\beta }_{1}NP{G}_{i,t}+{\beta }_{2}NP{L}_{i,t}+{\beta }_{3}NP{G}_{i,t}\times NP{L}_{i,t}\hfill \\ & +{\beta }_{4}SIZ{E}_{i,t}+{\beta }_{5}C{R}_{i,t}+{\beta }_{6}E{R}_{i,t}+{\beta }_{7}REN{T}_{i,t}+\epsilon \hfill \end{array}$$

(2)

Details of the variables are provided in Appendix A. $PRO{F}_{i,t}$ represents bank profitability. This study employs three distinct measures of profitability: $RO{L}_{i,t}$, $RO{A}_{i,t}$ and $RO{E}_{i,t}$. $RO{L}_{i,t}$ is the return on loans, directly reflecting the profitability of loans. $RO{A}_{i,t}$ and $RO{E}_{i,t}$ represent ROA and ROE, respectively. These two indicators reflect the bank’s overall profitability.

$NP{G}_{i,t}$ represents the proportion of new loans not contingent upon personal guarantees. RQ1 focuses on the coefficient of $NP{G}_{i,t}$ in Equation (1), which captures the relationship between loans without personal guarantees and bank profitability. $NP{L}_{i,t}$ represents the proportion of non-performing loans. A high non-performing loan ratio is considered to reflect poor management quality in the literature (Louzis et al., 2012; Podpiera & Weill, 2008). Banks with high-quality management are more likely to conduct effective screening and monitoring, enabling them to lend to creditworthy borrowers even without requiring personal guarantees. Therefore, this study focuses on the coefficient of the interaction term $NP{G}_{i,t}\times NP{L}_{i,t}$ in Equation (2) to explore RQ2.

Both models contain control variables based on the prior studies on bank profitability stated in Section 3.2. $SIZ{E}_{i,t}$ is the bank size and $C{R}_{i,t}$ denotes the cost-to-income ratio. $E{R}_{i,t}$ represents the equity-to-assets ratio. $REN{T}_{i,t}$ is lending services ratio of revenue.

4.2. Data and Variables

This study uses data from 100 regional banks listed in the 2023 annual data on the utilization of the guidelines on personal guarantees, published by the JFSA, to calculate $NP{G}_{i,t}$ (Japan Financial Services Agency, 2024). All of these banks have fiscal year-ends in March. $NP{G}_{i,t}$ is the proportion of new loans not contingent upon personal guarantees. Given the improvements in disclosure practices among regional banks since FY 2019, this study manually collects data concerning $NP{G}_{i,t}$ from the JFSA data and publications by regional banks from FY 2019 to 2023. The manually collected dataset is archived in a public data repository for reproducibility (Yoshinaga & Onishi, 2025).

Other variables are calculated using annual non-consolidated financial data obtained from NEEDS Financial Quest 2.0 for each bank. $RO{L}_{i,t}$ is the ratio of interest on loans and discounts (FINFSTA’D11023) to loans and bills discounted (FINFSTA’B11053). $RO{A}_{i,t}$ is the ratio of ordinary income (FINFSTA’D11115) to total assets (FINFSTA’B11098). $RO{E}_{i,t}$ is the ratio of net income (FINFSTA’D11145) to equity capital (FINFSTA’C11113). The deflators of the profitability variables are the averages of previous and current periods. $NP{L}_{i,t}$ is the non-performing loans ratio (FINFSTA’K11179). $SIZ{E}_{i,t}$ is calculated as the natural logarithm of total assets (FINFSTA’B11098). $C{R}_{i,t}$ is the ratio of ordinary expenses (FINFSTA’D11060) to ordinary revenue (FINFSTA’D11021). $E{R}_{i,t}$ is calculated as the ratio of equity capital (FINFSTA’C11113) to total assets (FINFSTA’B11098). $REN{T}_{i,t}$ is the ratio of interest on loans and discounts (FINFSTA’D11023) to ordinary revenue (FINFSTA’D11021).

After constructing these variables for each observation, those with missing values were excluded. Observations with values corresponding to the upper and lower 1% of each variable were excluded as outliers. Descriptive statistics for the final sample are presented in

Table 3. Descriptive statistics.

	Mean	SD	Min	25%	Median	75%	Max	N
$RO{L}_{i,t}$	0.011	0.003	0.007	0.009	0.010	0.012	0.028	352
$RO{A}_{i,t}$	0.002	0.001	−0.003	0.001	0.002	0.003	0.006	352
$RO{E}_{i,t}$	0.032	0.018	−0.074	0.023	0.033	0.041	0.075	352
$NP{G}_{i,t}$	0.388	0.136	0.148	0.284	0.362	0.468	0.784	352
$NP{L}_{i,t}$	0.020	0.009	0.008	0.014	0.018	0.023	0.099	352
$SIZ{E}_{i,t}$	14.987	0.985	12.523	14.239	15.074	15.750	16.865	352
$C{R}_{i,t}$	0.837	0.088	0.632	0.776	0.838	0.883	1.205	352
$E{R}_{i,t}$	0.047	0.011	0.025	0.041	0.046	0.053	0.083	352
$REN{T}_{i,t}$	0.541	0.109	0.326	0.458	0.532	0.621	0.790	352
$\Delta NP{G}_{i,t}$	0.064	0.080	−0.134	0.014	0.045	0.090	0.345	253

Note: The table shows descriptive statistics. SD denotes standard deviation; Min and Max indicate the minimum and maximum values, respectively; N represents the number of observations. $RO{L}_{i,t}$ is the return on loans. $RO{A}_{i,t}$ represent return on assets (ROA). $RO{E}_{i,t}$ denotes return on equity (ROE). $NP{G}_{i,t}$ represents the proportion of new loans not contingent upon personal guarantees. $NP{L}_{i,t}$ represents the proportion of non-performing loans (NPL). $SIZ{E}_{i,t}$ is the bank size. $C{R}_{i,t}$ denotes the cost-to-income ratio. $E{R}_{i,t}$ represents the equity-to-assets ratio. $REN{T}_{i,t}$ is lending services ratio of revenue. $\Delta NP{G}_{i,t}$ is the first difference of $NP{G}_{i,t}$.

, encompassing 352 bank/year observations from FYs 2019 to 2023. The average of $RO{L}_{i,t}$ suggests that the loan interest rate of Japanese regional banks is approximately 1%. On average, $RO{E}_{i,t}$ is more than 15 times greater than $RO{A}_{i,t}$ driven by the high financial leverage of commercial banks. The average value of $NP{G}_{i,t}$ is 38.8%, whereas the minimum and maximum values are 14.8% and 78.4%, respectively. This variation suggests that the implementation of loans without personal guarantee varies substantially across banks. Descriptive statistics for $\Delta NP{G}_{i,t}\left(=NP{G}_{i,t}-NP{G}_{i,t-1}\right)$ are also presented. The mean value of $\Delta NP{G}_{i,t}$ indicates that the proportion of such loans increased by an average of 6.4% annually. Figure 2 illustrates the time-series trend of $NP{G}_{i,t}$, showing a consistent upward trajectory over time.

Table 4. Correlation matrix.

	$RO{L}_{i,t}$	$RO{A}_{i,t}$	$RO{E}_{i,t}$	$NP{G}_{i,t}$	$NP{L}_{i,t}$	$SIZ{E}_{i,t}$	$C{R}_{i,t}$	$E{R}_{i,t}$	$REN{T}_{i,t}$
$RO{L}_{i,t}$		0.038	−0.048	−0.036	0.308	−0.668	0.242	0.214	0.646
$RO{A}_{i,t}$	0.089		0.788	0.108	−0.194	0.337	−0.912	0.455	−0.158
$RO{E}_{i,t}$	−0.004	0.831		0.191	−0.124	0.338	−0.785	−0.014	−0.077
$NP{G}_{i,t}$	0.028	0.143	0.215		−0.103	0.094	−0.122	−0.091	−0.029
$NP{L}_{i,t}$	0.586	−0.079	−0.044	−0.048		−0.443	0.287	−0.057	0.153
$SIZ{E}_{i,t}$	−0.645	0.309	0.297	0.097	−0.433		−0.533	−0.019	−0.591
$C{R}_{i,t}$	0.165	−0.924	−0.840	−0.142	0.206	−0.474		−0.290	0.215
$E{R}_{i,t}$	0.242	0.461	0.035	−0.075	0.054	0.004	−0.313		−0.079
$REN{T}_{i,t}$	0.674	−0.119	−0.075	−0.001	0.300	−0.614	0.186	−0.070

Note: The table shows correlation matrix. Pearson (Spearman) correlations are below (above) the diagonal. $RO{L}_{i,t}$ is the return on loans. $RO{A}_{i,t}$ represent return on assets (ROA). $RO{E}_{i,t}$ denotes return on equity (ROE). $NP{G}_{i,t}$ represents the proportion of new loans not contingent upon personal guarantees. $NP{L}_{i,t}$ represents the proportion of non-performing loans (NPL). $SIZ{E}_{i,t}$ is the bank size. $C{R}_{i,t}$ denotes the cost-to-income ratio. $E{R}_{i,t}$ represents the equity-to-assets ratio. $REN{T}_{i,t}$ is lending services ratio of revenue.

presents the correlation matrix for the variables. $NP{G}_{i,t}$ and the profitability variables ($RO{L}_{i,t}$, $RO{A}_{i,t}$ and $RO{E}_{i,t}$) are not strongly correlated. $RO{L}_{i,t}$ does not show a strong correlation with $RO{A}_{i,t}$ and $RO{E}_{i,t}$, suggesting that higher loan interest rates do not necessarily translate into greater overall bank profitability. Furthermore, $RO{L}_{i,t}$ shows a positive correlation with $NP{L}_{i,t}$ and a negative correlation with $SIZ{E}_{i,t}$, whereas $NP{L}_{i,t}$ and $SIZ{E}_{i,t}$ are negatively correlated. This suggests that regional banks that are more active in lending to high-risk borrowers tend to charge higher interest rates, whereas larger banks tend to lend to lower-risk borrowers. $RO{A}_{i,t}$ and $RO{E}_{i,t}$ are strongly negatively correlated with $C{R}_{i,t}$, aligning with previous studies.

5. Results

Table 5. Loans without personal guarantee and bank profitability: fixed effect approach.

	(1)	(2)	(3)	(4)	(5)	(6)
	$RO{L}_{i,t}$		$RO{A}_{i,t}$		$RO{E}_{i,t}$
$Intercept$	−0.011	−0.008	0.019 ***	0.020 ***	0.401 *	0.394 *
	(0.012)	(0.009)	(0.003)	(0.004)	(0.231)	(0.232)
$NP{G}_{i,t}$	0.002 *	0.005 ***	0.000	0.001	0.010	0.003
	(0.001)	(0.002)	(0.000)	(0.000)	(0.008)	(0.014)
$NP{L}_{i,t}$	0.002	0.079 ***	−0.017	−0.003	−0.268	−0.427
	(0.018)	(0.028)	(0.013)	(0.015)	(0.210)	(0.389)
$NP{G}_{i,t}$ $\times NP{L}_{i,t}$		−0.186 ***		−0.033		0.386
		(0.055)		(0.027)		(0.680)
$SIZ{E}_{i,t}$	0.001 *	0.001 *	−0.000 **	−0.000 **	−0.011	−0.010
	(0.001)	(0.001)	(0.000)	(0.000)	(0.015)	(0.015)
$C{R}_{i,t}$	0.001	0.001	−0.012 ***	−0.012 ***	−0.212 ***	−0.212 ***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.016)	(0.016)
$E{R}_{i,t}$	0.003	−0.001	0.011	0.011	−0.290	−0.281
	(0.019)	(0.019)	(0.007)	(0.007)	(0.188)	(0.189)
$REN{T}_{i,t}$	0.004 *	0.005 **	−0.002 ***	−0.002 ***	−0.028	−0.029 *
	(0.002)	(0.002)	(0.001)	(0.001)	(0.017)	(0.017)
$Adj.R^2$	0.951	0.954	0.955	0.956	0.852	0.852
$N$	352	352	352	352	352	352
Year fixed	Included	Included	Included	Included	Included	Included
Bank fixed	Included	Included	Included	Included	Included	Included

Note: The table reports the results from two-way fixed effect regressions. In parentheses, bunk-clustered robust standard errors are reported. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

presents empirical results. Columns (1) and (2), (3) and (4), and (5) and (6) present the results with $RO{L}_{i,t}$, $RO{A}_{i,t}$ and $RO{E}_{i,t}$ as the dependent variable, respectively. The consistently high adjusted R-squared values across all columns indicate that the models are well-specified and include sufficient control variables.

Based on Columns (1), (3) and (5), which address RQ1, the coefficient of $NP{G}_{i,t}$ is statistically significantly positive when $RO{L}_{i,t}$ is used as a dependent variable. However, based on the coefficient values, an average change in the ratio of $NP{G}_{i,t}$ (6.4%) is associated with a 0.013% (=0.002 × 6.4%) increase in the return on loans, a 0.000% (=0.000 × 6.4%) increase in ROA, and a 0.064% (=0.010 × 6.4%) increase in ROE. These values are substantially lower than the standard deviations of $RO{L}_{i,t}$ (0.3%), $RO{A}_{i,t}$ (0.1%), and $RO{E}_{i,t}$ (1.8%). Therefore, the results suggest that the economic impact of the proportion of loans without personal guarantee on bank profitability is minimal.

In Columns (2), (4) and (6), where $RO{L}_{i,t}$ is used as the dependent variable, only the coefficient of the interaction term $NP{G}_{i,t}\times NP{L}_{i,t}$ is significantly negative. This finding may suggest that the positive impact of such loans on loan interest rates is more pronounced in banks with high-quality management. Accordingly, the marginal effects of $NP{G}_{i,t}$ are examined in

Table 6. Marginal effect of loans without personal guarantee and bank profitability: fixed effect approach.

Panel A Marginal effect of $NP{G}_{i,t}$ on $RO{L}_{i,t}$ for various levels of $NP{L}_{i,t}$
$NP{L}_{i,t}$ percentiles	$NP{L}_{i,t}$ level	marginal effect	standard errors
10th	0.012	0.003 ***	(0.001)
25th	0.014	0.002 **	(0.001)
50th	0.018	0.002 *	(0.001)
75th	0.023	0.001	(0.001)
90th	0.028	−0.000	(0.001)
Panel B Marginal effect of $NP{G}_{i,t}$ on $RO{A}_{i,t}$ for various levels of $NP{L}_{i,t}$
$NP{L}_{i,t}$ percentiles	$NP{L}_{i,t}$ level	marginal effect	standard errors
10th	0.012	0.000	(0.000)
25th	0.014	0.000	(0.000)
50th	0.018	0.000	(0.000)
75th	0.023	0.000	(0.000)
90th	0.028	−0.000	(0.000)
Panel C Marginal effect of $NP{G}_{i,t}$ on $RO{E}_{i,t}$ for various levels of $NP{L}_{i,t}$
$NP{L}_{i,t}$ percentiles	$NP{L}_{i,t}$ level	marginal effect	standard errors
10th	0.012	0.008	(0.009)
25th	0.014	0.009	(0.008)
50th	0.018	0.010	(0.008)
75th	0.023	0.012	(0.010)
90th	0.028	0.014	(0.012)

Note: The table reports the marginal effects of $NP{G}_{i,t}$ on bank profitability for the percentiles of $NP{L}_{i,t}$ using Equation (2). In parentheses, bank-clustered robust standard errors are reported. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

.

Table 6 presents and Figure 3 illustrates the marginal effects of $NP{G}_{i,t}$ at the 10th, 25th, 50th, 75th and 90th percentiles of $NP{L}_{i,t}$, based on Equation (2). Similar to Table 5, Table 6 shows that a statistically significant and positive marginal effect of $NP{G}_{i,t}$ is observed when $RO{L}_{i,t}$ is used as the dependent variable. Panel A indicates that the effect on $RO{L}_{i,t}$ becomes stronger as $NP{L}_{i,t}$ decreases. However, between the 10th and 90th percentiles of $NP{L}_{i,t}$, an average change in $NP{G}_{i,t}$ results in only a modest improvement in $RO{L}_{i,t}$: 0.020% (=0.003 × 6.4%) at most. In Panel B, the marginal effects of $NP{G}_{i,t}$ on $RO{A}_{i,t}$ are approximately zero across the same percentile range. Panel C shows that the positive marginal effect of $NP{G}_{i,t}$ on $RO{E}_{i,t}$ is not statistically significant within the 10th to 90th percentile range of $NP{L}_{i,t}.$ Additionally, within this percentile range, an average change in $NP{G}_{i,t}$ leads to only a modest improvement of 0.090% (=0.014 × 6.4%) in $RO{E}_{i,t}$, at most. Overall, all marginal effects are substantially smaller than the standard deviations of the respective profitability measures. When $NP{L}_{i,t}$ falls within the 10th to 90th percentiles, the marginal effects of $NP{G}_{i,t}$ on bank profitability are not economically significant.

In summary, Table 5 and Table 6 indicate that the marginal effects of loans without personal guarantees remain economically insignificant regardless of the management quality of banks. Prior studies report that, in Japan, personal guarantees have an insignificant effect on borrowers’ profitability (Ono et al., 2012) and credit scores (Ono & Uesugi, 2009). Given these weaker effects, their impact on bank profitability is also likely to be insignificant. Alternatively, another plausible explanation is that personal guarantees are discontinued only for clients whose business performance is sufficiently stable, so that risk and interest of lending do not increase even without personal guarantees. These findings challenge the intended policy outcomes promoted by Japanese financial authorities. Japanese financial authorities have encouraged such loans, with one of the primary objectives being the stabilization of bank profitability. However, the results indicate that even if the proportion of such loans increases, the resulting improvement in bank profitability appears to be negligible.

As Table 4 shows, the control variables used in the models exhibit significant correlations, potentially introducing bias owing to multicollinearity. To address this issue, a robustness check was conducted by re-estimating the models using only the main principal components of the control variables. The untabulated results confirm that the coefficients and significance levels of the variables of interest remain largely consistent with those reported in Table 4 when the models incorporate the first through third principal components in place of the original control variables.

6. Robustness to Endogeneity

The results from the fixed-effects model presented in Section 5 may be biased due to endogeneity. To address this potential problem, this study also estimates a dynamic panel data model using the two-step system GMM estimator (Blundell & Bond, 1998), consistent with research on bank profitability (Alnabulsi et al., 2022; Dietrich & Wanzenried, 2011; Gržeta et al., 2023; Kumar et al., 2020, 2022; Liu & Wilson, 2010; Pancotto et al., 2024). The empirical analysis is based on Equations (3) and (4)11:

$$\begin{array}{cc}\hfill PRO{F}_{i,t}& =\alpha +{\beta }_{1}PRO{F}_{i,t-1}+{\beta }_{2}NP{G}_{i,t}+{\beta }_{3}NP{L}_{i,t}+{\beta }_{4}SIZ{E}_{i,t}\hfill \\ & +{\beta }_{5}C{R}_{i,t}+{\beta }_{6}E{R}_{i,t}+{\beta }_{7}REN{T}_{i,t}+\sum \gamma Yea{r}_t+\epsilon \hfill \end{array}$$

(3)

$$\begin{array}{cc}\hfill PRO{F}_{i,t}& =\alpha +{\beta }_{1}PRO{F}_{i,t-1}+{\beta }_{2}NP{G}_{i,t}+{\beta }_{3}NP{L}_{i,t}+{\beta }_{4}NP{G}_{i,t}\times NP{L}_{i,t}\hfill \\ & +{\beta }_{5}SIZ{E}_{i,t}+{\beta }_{6}C{R}_{i,t}+{\beta }_{7}E{R}_{i,t}+{\beta }_{8}REN{T}_{i,t}+\sum \gamma Yea{r}_t+\epsilon \hfill \end{array}$$

(4)

The lagged dependent variable ($PRO{F}_{i,t-1}$) and other potentially endogenous variables—specifically, all explanatory variables except the year dummies in Equations (3) and (4)—are treated as GMM-style instruments. Following Blundell and Bond (1998) and Roodman (2009a, 2009b), only the second lag of these endogenous variables is used as instruments to avoid weak instruments and mitigate instrument proliferation. To further address the issue of instrument proliferation and ensure estimation stability, the instrument set is collapsed (Roodman, 2009b). Year dummies ($Yea{r}_t$) are included as exogenous instruments in levels to control for time-specific effects12. Robust standard errors with Windmeijer’s correction are employed to address heteroskedasticity and finite-sample bias in the two-step estimation. Model validity is assessed using the Hansen test for overidentifying restrictions and the Arellano–Bond test for the second-order autocorrelation (AR(2)) in first differences (Roodman, 2009b).

Table 7. Loans without personal guarantee and bank profitability: GMM approach.

	(1)	(2)	(3)	(4)	(5)	(6)
	$RO{L}_{i,t}$		$RO{A}_{i,t}$		$RO{E}_{i,t}$
$Intercept$	−0.011	−0.011	0.027 ***	0.024 ***	0.515 ***	0.463 ***
	(0.018)	(0.015)	(0.007)	(0.007)	(0.125)	(0.170)
$RO{L}_{i,t-1}$	1.209 ***	1.109 ***
	(0.374)	(0.261)
$RO{A}_{i,t-1}$			−0.042	−0.053
			(0.051)	(0.057)
$RO{E}_{i,t-1}$					−0.057	−0.045
					(0.048)	(0.061)
$NP{G}_{i,t}$	0.000	0.000	0.001	0.002	0.026	0.047
	(0.004)	(0.003)	(0.002)	(0.001)	(0.040)	(0.033)
$NP{L}_{i,t}$	−0.014	−0.001	−0.024	0.015	−0.376	0.337
	(0.040)	(0.086)	(0.023)	(0.035)	(0.534)	(1.008)
$NP{G}_{i,t}$ $\times NP{L}_{i,t}$		−0.061		−0.047		−0.573
		(0.180)		(0.074)		(2.250)
$SIZ{E}_{i,t}$	0.001	0.000	−0.001 **	−0.001 **	−0.012 ***	−0.010 *
	(0.001)	(0.001)	(0.000)	(0.000)	(0.004)	(0.006)
$C{R}_{i,t}$	0.003	0.004	−0.016 ***	−0.015 ***	−0.268 ***	−0.276 ***
	(0.006)	(0.006)	(0.002)	(0.004)	(0.045)	(0.081)
$E{R}_{i,t}$	−0.002	0.011	0.004	0.004	−0.799 ***	−0.807 ***
	(0.039)	(0.043)	(0.018)	(0.019)	(0.230)	(0.240)
$REN{T}_{i,t}$	−0.002	−0.001	−0.004 *	−0.003 *	−0.090 *	−0.065
	(0.005)	(0.005)	(0.003)	(0.002)	(0.048)	(0.040)
Year dummy	Included	Included	Included	Included	Included	Included
N of obs	253	253	253	253	253	253
N of group	79	79	79	79	79	79
N of inst	18	20	18	20	18	20
Hansen p	0.385	0.238	0.128	0.071	0.501	0.326
AR(1) p	0.396	0.387	0.159	0.254	0.018	0.120
AR(2) p	0.425	0.383	0.678	0.489	0.886	0.851

Note: The table reports the results from two-step system GMM regressions. In parentheses, Windmeijer-corrected robust standard errors are reported. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. N of obs, group and inst are the number of observations, group (bank), and instruments, respectively.

reports the results of this robustness check. The number of observations is smaller than in Table 5 due to the inclusion of lagged dependent variables. The p-values from the Arellano–Bond test for AR(2) in the first-differenced residuals are consistently insignificant across all specifications, supporting the validity of the moment conditions. The p-values from the Hansen test, excluding Column (4), are also insignificant, supporting the validity of the instrument specification. In Column (4) only, the p-value from the Hansen test is slightly significant at the 10% level, which may indicate the instrument specification in this column is not valid. Therefore, to further assess the robustness of the results, the two-step difference GMM estimator is also employed as discussed in the final paragraph in this section (untabulated).

In contrast to Table 5, all coefficients of $NP{G}_{i,t}$ and $NP{G}_{i,t}\times NP{L}_{i,t}$ are statistically insignificant. This finding suggests that the statistical significance of the positive effect of $NP{G}_{i,t}$ is not robust even when accounting for bank management quality.

Table 8. Marginal effect of loans without personal guarantee and bank profitability: GMM approach.

Panel A Marginal effect of $NP{G}_{i,t}$ on $RO{L}_{i,t}$ for various levels of $NP{L}_{i,t}$
$NP{L}_{i,t}$ percentiles	$NP{L}_{i,t}$ level	marginal effect	standard errors
10th	0.012	−0.001	(0.003)
25th	0.014	−0.001	(0.003)
50th	0.018	−0.002	(0.004)
75th	0.023	−0.002	(0.004)
90th	0.028	−0.002	(0.005)
Panel B Marginal effect of $NP{G}_{i,t}$ on $RO{A}_{i,t}$ for various levels of $NP{L}_{i,t}$
$NP{L}_{i,t}$ percentiles	$NP{L}_{i,t}$ level	marginal effect	standard errors
10th	0.012	0.001	(0.001)
25th	0.014	0.001	(0.001)
50th	0.018	0.001	(0.001)
75th	0.023	0.001	(0.002)
90th	0.028	0.000	(0.002)
Panel C Marginal effect of $NP{G}_{i,t}$ on $RO{E}_{i,t}$ for various levels of $NP{L}_{i,t}$
$NP{L}_{i,t}$ percentiles	$NP{L}_{i,t}$ level	marginal effect	standard errors
10th	0.012	0.041	(0.041)
25th	0.014	0.039	(0.044)
50th	0.018	0.037	(0.051)
75th	0.023	0.034	(0.060)
90th	0.028	0.031	(0.069)

Note: The table reports the marginal effects of $NP{G}_{i,t}$ on bank profitability for the percentiles of $NP{L}_{i,t}$ using Equation (4). In parentheses, Windmeijer-corrected robust standard errors are reported. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

reports and Figure 4 illustrates the marginal effects of $NP{G}_{i,t}$ at the 10th, 25th, 50th, 75th, and 90th percentiles of $NP{L}_{i,t}$, based on Equation (4). Across these percentiles, all marginal effects of $NP{G}_{i,t}$ are statistically insignificant, consistent with Table 7. Within the 10th to 90th percentiles of $NP{L}_{i,t}$, an average change in $NP{G}_{i,t}$ leads to an increase of −0.006% (=−0.001 × 6.4%) in $RO{L}_{i,t}$, 0.006% (=0.001 × 6.4%) in $RO{A}_{i,t}$, and 0.262% (=0.041 × 6.4%) in $RO{E}_{i,t}$ at most. All of their absolutes are substantially smaller than the standard deviations of the respective profitability measures. These results support the conclusion that the effect of loans without personal guarantees on bank profitability is economically insignificant.

To ensure the robustness of the results obtained from the two-step system GMM estimator, an alternative specification, the difference GMM estimator, is also employed (untabulated)13. The p-values from the Hansen test and the Arellano–Bond test for AR(2) in the first-differenced residuals are consistently insignificant across all specifications, supporting the validity of the moment conditions. Consistent with the findings from the system GMM specifications, the coefficients of $NP{G}_{i,t}$ and $NP{G}_{i,t}\times NP{L}_{i,t}$ remain statistically insignificant, and their economic significance is minimal. These results reinforce the robustness of the overall conclusion.

7. Conclusions

This study empirically investigates the impact of loans without personal guarantees on the profitability of regional banks. The findings confirm that the statistical significance of the impact is not robust and that such loans have no economically significant impact on overall bank profitability, despite the level of non-performing loan ratios of the banks. Therefore, even as the proportion of the concerned loans increases, it does not significantly influence bank profitability.

This study contributes to literature in both academic and practical terms. First, although previous studies have analyzed the effects of loans without personal guarantees at the loan contract level, the broader impact on lenders’ overall profitability has remained unexplored. This study addresses the gap by demonstrating the insignificant impact of the ratio of such loans on bank profitability. Considering that both previous Japanese studies (Ono et al., 2012; Ono & Uesugi, 2009) and non-Japanese studies (Cadot, 2013; Peltoniemi & Vieru, 2013) highlight the limited impact of personal guarantees at the contract level, the concerned loans will not have a significant impact on lenders’ profitability outside Japan.

Second, financial authorities in Japan have been actively promoting lending without relying on personal guarantees for over a decade, while enhancing transparency through the introduction of KPIs. In 2019, the JFSA shifted its focus from flexible benchmarks, allowing banks to select multiple indicators for financial intermediation, to the standard disclosure of two selected KPIs. One of these KPIs is the percentage of new loans that do not rely on business owner guarantees. However, the findings of this study indicate that an increase in the proportion of such loans does not contribute significantly to bank profitability. This suggests that the proportion of such loans may not serve as an effective KPI for improving banks’ financial performance and management stability.

This study has several limitations and suggestions for future research. The first limitation is that the modest sample size and short sample period may have been affected by the COVID-19 pandemic. Due to data constraints regarding the loans examined, the sample comprises 352 bank/year observations and spans five years from 2019 to 2023, which includes the COVID-19 pandemic period. Although the results are robust under both the two-way fixed effects and GMM approaches, they may still be subject to distortion. Future research could address this limitation: reanalyzing the effects once longer-term data become available, conducting similar analyses in other countries, and utilizing international datasets.

The second limitation is that this study does not answer the question of why such loans do not significantly impact bank profitability. One possible explanation is that many new loans without personal guarantees are rollovers of existing loans extended to dependable borrowers, rather than credit extended to unfamiliar clients. Collateral and personal guarantees play a key role in mitigating lenders’ risk. When banks extend loans to new borrowers without collateral or personal guarantees, they face higher screening and monitoring costs, along with increased credit risk. However, new loans to existing reliable borrowers entail lower costs and risks for banks, reducing the need to significantly increase interest rates. Consequently, it is plausible that banks primarily issue such loans to existing reliable borrowers, which may explain why these loans do not have a significant impact on bank profitability. Addressing the mechanism is one of the promising future research projects.

As the third limitation, this study does not investigate the other social benefits of loans without personal guarantees, because it focuses on their effects on bank profitability. Hoshi and Shibuya (2023) report that reducing reliance on personal guarantees facilitates business succession in SMEs. Investigating the other social benefits of reducing personal guarantees is a possible future research.