Insider ownership and default risk: What does the data reveal about Japanese firms?

Abstract

Purpose

The ownership structure in Japanese firms has experienced a significant change recently, fueled primarily by regulatory changes. This has important repercussions on corporate performance and risk. This paper examines the impact of insider ownership on the default risk of Japanese firms.

Design/methodology/approach

We collected data from the Nikkei Corporate Governance Evaluation System (CGES) database for the period 2004–2019. Our final dataset yields 36,116 firm-year observations. We apply a firm fixed effect model for baseline regression. Endogeneity was checked by applying propensity score matching (PSM) and two-stage least squares (2SLS) techniques. Furthermore, the robustness of baseline regression results was checked using alternative estimation techniques.

Findings

Results show a significant positive influence of insider ownership on default risk. Furthermore, ROA volatility and stock price volatility appear to be the major channels through which insider ownership affects a firm’s default risk. We further document that external monitoring mechanisms, including traditional main bank ties, institutional ownership and analyst coverage, are the key risk-mitigating factors.

Research limitations/implications

Our research deals with Japanese firms only. Future research may attempt to analyze the cases of emerging economies. Furthermore, future research might examine the ownership-default risk relationship for financial institutions to see if this relationship differs between financial and nonfinancial firms.

Practical implications

Insider ownership enhances the probability of default. Hence, policymakers may consider instituting a ceiling for insider ownership in Japanese firms. Moreover, we highlight various risk-mediating channels that would help policymakers adopt guidelines for mitigating corporate risk.

Originality/value

Our study is the first to investigate the effect of insider ownership on default risk in Japanese settings. Prior studies identified various determinants that affect firms’ default risk. Our study contributes to this stream of literature by examining the impact of insider ownership on default risk and extending the limited literature related to insider ownership.

Citation

Haque, H., Kabir, M.N., Abedin, S.H., Miah, M.D. and Sharma, P. (2024), "Insider ownership and default risk: What does the data reveal about Japanese firms?", China Accounting and Finance Review, Vol. 26 No. 3, pp. 354-384. https://doi.org/10.1108/CAFR-06-2023-0059

1. Introduction

The relationship between insider ownership and firms’ default risk can be viewed from the classical agency theory, which holds that modern firms are owned by dispersed and atomistic shareholders; hence, they appoint professional managers as agents with substantial control rights over operations. This separation of ownership and control provides managers with considerable latitude in personal empire building at the expense of shareholders (Fama & Jensen, 1983). Owners, thus, have tended to adopt various mechanisms to mitigate the ensuing agency problems. One such mechanism has been an attempt to encourage managers to become part owners themselves, with the expectation that providing insiders – CEOs, employees and directors – the opportunity to own shares will align the interests of manager–owners with those of outside shareholders (Jensen & Meckling, 1978), motivating them thus to enhance firm value since their financial interests would be tied to shareholder wealth.

Insider ownership provides an intriguing ownership setting to test implications for various firm performance outcomes. Indeed, extant literature has examined the relationship between insider ownership and accounting performance (Bhagat & Bolton, 2019; Tsafack & Guo, 2021; Fan & Wang, 2022), stock return (Kaserer & Moldenhauer, 2007), investment decisions (Panousi & Papanikolaou, 2012), firms’ innovative capacity (Choi, Lee, & Williams, 2011), the cost and uses of corporate debt (Lugo, 2019) and corporate value (Bhabra, 2007). However, studies investigating the association between insider ownership and default risk remain scarce, thus providing the primary impetus for the present study. Our study focuses on Japan to examine the insider ownership-default risk nexus, motivated by the uniqueness of Japanese corporate governance and ownership structures. During the country’s 1955–1990 super-growth period, Japanese corporate shares were co-owned by main banks, non-bank financial institutions, parent companies, subsidiary firms and others. Referred to as “cross and stable shareholding” (Johnston & McAlevey, 1998), this structure of shareholding was required in instances where financial markets did not exert adequate discipline on corporate borrowers (Nguyen, 2011); in lieu, cross-shareholders were expected to monitor the managers. A downside of cross-shareholding is that managers may go unpunished, including termination for poor performance (Ikeda, Inoue, & Watanabe, 2018). Consequently, managers may take on less risky investments, which may not lead to optimal performance for stakeholders. Japan’s super-growth period was followed by a period of economic stagnation, commencing in the mid-1990s, which, among others, compelled the country to initiate distinct reforms in corporate governance and ownership structures.

The noticeable reforms in ownership structures are perceived to have important ramifications for corporate default risk. To date, the literature has examined the effect of institutional ownership on default risk for Japanese firms (Kabir, Miah, Ali, & Sharma, 2020; Sakawa, Watanabel, Duppati, & Faff, 2021). There has yet to be an empirical study that investigates the effect of insider ownership on default risk in Japan. Shuto and Kitagawa (2011) investigated the effect of managerial ownership (MO), estimated as a percentage of shares owned by the directors, on the cost of debt in Japanese firms. Analyzing 643 firm-year observations for the period 1997–2004, they document a positive impact of MO on interest rate spread, implying that MO increases firms’ risk reflected in their risk premium. Similarly, Tanaka (2016) investigates the effect of MO, measured as a percentage of equity ownership held by managers and the maturity structure of corporate public debt. Analyzing 1,454 firm-year observations for the period 2005–2012, this study concludes that firms with higher MO issue shorter maturity debt and receive lower credit ratings.

We build upon and contribute to the above stream of literature by analyzing 36,116 firm-year observations for the period 2004–2019 and include firms of all types, ages and financial status. Our study differs from Shuto and Kitagawa (2011) as well as Tanaka (2016) in several important ways. First, the studies of Shuto and Kitagawa (2011) as well as Tanaka (2016) are unlikely to account for the dynamics of the recent changes that took place in corporate governance and ownership settings in Japan (section 2 describes these changes in detail). Moreover, unlike the studies of Shuto and Kitagawa (2011) and Tanaka (2016), our dataset comprises a substantially large sample and a longer period. A large sample size reduces estimation bias and facilitates the generalization of findings to a broader context. In addition, a large sample size boosts the rigor and reliability of research because it provides more information, greater variability, reduced collinearity among variables and higher degrees of freedom (Baltagi, 2008). Hence a large observation drawn from a longer period facilitates reliably tackling endogeneity problems using sophisticated techniques.

Second, the default risk proxies used in Shuto & Kitagawa (interest rate spread) and in Tanaka (yield spread and credit rating) are similar, to a greater extent, which requires additional measures of default risk. Literature argues that it is instructive to use more than one predictor of firms’ default risk (Hilscher & Wilson, 2017). Hence, our measure of default risk, Merton’s distance to default, complements the above studies in an incremental way. Moreover, Shuto and Kitagawa estimate managerial ownership as the ratio of shares owned by the directors, whereas Tanaka takes the percentage of equity ownership held by managers. We provide an explanation of three distinct proxies: directors, CEOs and employee ownership. This exercise is crucial because there may be differences in interests among various groups of corporate insiders. As a result, the overall ownership by insiders is expected to have minimal variation within a company because of the confounding effects of changes in the number and composition of insider groups over a period of time.

Applying the firm fixed effect (FE) model, we find a negative effect of insider ownership on firms’ distance to default, implying that the larger the insider ownership, the greater the default risk. Our results thus support the “managerial entrenchment” hypothesis in the context of control rights. Our results remain robust to a battery of robustness tests, including alternative proxies of default risk, estimation methods and endogeneity concerns. We further show that the effect of insider ownership on firms’ default risk is partially mediated by return on asset volatility and stock price volatility. Finally, we illustrate that the positive effect of insider ownership on default risk is weaker for firms with larger institutional shareholders, greater analyst coverage and stronger ties with main banks. This risk-mitigating effect of external governance mechanisms can be attributed to their monitoring role, which partially reduces the excessive risk-taking tendency of managers.

Our study makes three primary contributions to the literature. First, to the best of our knowledge, this study is the first to investigate the effect of insider ownership on default risk in Japanese settings. We believe that Japan offers itself as an important case for our study because of its unique tradition of ownership structure. Traditionally, Japanese corporations were tightly protected from external takeovers, facilitated by unique institutions including cross- and stable shareholdings. However, these protective institutions are gradually winding up with the aim of reinvigorating the ailing Japanese economy by reducing corporate bankruptcies as well as tapping a diversified base of shareholders. Potential investors may be interested in knowing the effect of a particular type of ownership on managerial risk-taking behavior. In this regard, our research provides new information that is helpful for investors. Moreover, very little is known about Japanese corporate characteristics compared to their Western counterparts, including those in the USA and UK Given this academic backdrop, the current study is expected to enhance our understanding of the dynamics of corporate ownership and the default-risk nexus in the Japanese context.

Second, the expanding literature investigates the impact of institutional (Suto & Toshino, 2005; Kabir et al., 2020), foreign (Ahmadjian & Robbins, 2005; Nagaoka, 2006; Sueyoshi, Goto, & Omi, 2010) and concentrated ownership (Nguyen, 2011) on firm risk-taking. In addition, the extant literature provides evidence on the impact of insider ownership on firm performance (Iturralde, Maseda, & Arosa, 2011; Jain, Gunasekar, & Balasubramanian, 2020). However, the impact of insider ownership on default risk remains largely unexplored. Hence, our study is novel in providing evidence as to the effect of insider ownership on the default risk of Japanese firms.

Third, prior studies show that stock market volatility (Naifar, 2011), bank–firm relations (Fukuda, Kasuya, & Akashi, 2009), corporate social responsibility (Suganda & Kim, 2023) and financial mismatch (Zhitao & Xiang, 2023) affect firms’ default risk. Our study contributes to this stream of literature by identifying another important, but previously unexplored, determinant that affects firms’ default risk. Moreover, our research identifies the channels through which the effect of insider ownership on corporate default risk is transmitted. We also show the governance mechanisms that mitigate the negative impact of insider ownership on the distance to default. These findings bear important policy relevance as to how firms and regulatory authorities formulate policies to effectively manage corporate risks.

We have organized the paper as follows: Section 2 briefly discusses the corporate ownership structures and recent changes in Japan. Section 3 studies the relevant literature and formulates the hypothesis. The research methodology to test the hypothesis is presented in Section 4. Section 5 presents and interprets the empirical results, followed by the channel analysis in Section 6. Section 7 discusses the external corporate governance mechanisms that may moderate the association between insider ownership and default risk, which is followed by the conclusion and policy recommendations in Section 8.

2. Corporate ownership and regulatory changes in Japan

On the one hand, Japan is the third largest economy in the world in 2022, with a GDP of USD 4.26 trillion. However, it has been stuck in a state of financial and economic stagnation for many years. Furthermore, business performance was significantly and negatively affected by two successive financial crises: the Asian financial crisis in 1997 and the global subprime disaster in 2007, resulting in a sweeping and adverse impact on corporate performance. For instance, the number of corporate bankruptcies remained high (13,414 cases per year) for the last three decades (1990–2020). The estimated amount of liabilities of all bankrupted firms in the year 2000 alone amounted to the historic high of about 24 trillion Japanese yen. With the normalization of the Asian crisis, the number of corporate bankruptcies was declining, reaching its lowest point (6,030 firms) in 2021. However, the number remains at such a high level that an economy like Japan, at its revitalizing stage and struggling with low GDP growth, cannot afford to absorb.

Consequently, Japan initiated various reforms in the realms of corporate governance and finance. For instance, stock options were legalized after strong pressure from corporations, leading to the rise of managerial shareholdings. At the same time, smaller boards, in terms of size, replaced the conventional larger boards (Ahmadjian, 2000). Moreover, an amendment to the Commercial Code was made in 2002 requiring companies to emphasize selecting outside directors for the board. In continuation, the Commercial Code was revised in April 2003, allowing Japanese firms to choose either the prevalent auditing system or a US-style committee system, thus legally separating monitoring functions from the executives’ functions (Yoshikawa & McGuire, 2008).

Amidst this, the Japanese corporate environment witnessed an unprecedented event in 2005: Livedoor’s takeover battle for venerated Fuji TV. This episode was considered one of the greatest uproars in Japanese corporate governance settings, which paved the way for reshaping corporate governance and takeover markets (Miah & Uddin, 2017; Whittaker & Hayakawa, 2007). Livedoor’s successful takeover bids, capitalizing on regulatory loopholes, accelerated the ongoing reforms in the basic pillars of the country’s corporate governance that affected ownership as well. The Company Law was reformed in May 2006, introducing different classes of stocks to streamline corporate ownership, particularly for start-ups (Whittaker & Hayakawa, 2007). In addition, the Companies Act, which was revised in 2014, mandated listed companies to appoint a minimum of one independent external director (Armour, Enriques, Hansmann, & Kraakman, 2017). Subsequently, the introduction of the Corporate Governance Code in 2015 changed the ownership structure, unwinding relational shareholdings and increasing foreign, institutional and individual shareholdings (Mielcarz, Osiichuk, & Puławska, 2023). For example, total shareholdings by corporate firms and financial institutions dropped from 73.1% in 1990 to 50% in 2021. Shareholdings by city and regional banks declined to just 2.5% in 2021. At the same time, foreign shareholdings increased significantly, from 5% in 1990 to 30.4% in 2021. There has also been an increase in individual and dispersed shareholding. As of 2021, individuals together held up to 3% of shares in Japanese firms.

In proportion to this trend, insider shareholding rose significantly over the years. Kato, Skinner, and Kunimura (2009) estimated that the parentage of ownership held by managers and board members accounted for 8.08% for the period 1997–2007. In our sample, the respective estimate accounts for 13.17% (average for the period 2004–2019). Including employees’ ownership (2.9%) in this figure yields an insider ownership percentage equal to 16. Although it is difficult to compare this data with other developed countries due to differences in sample size and definitions of insider ownership, the literature, however, offers indicative statistics. Lewellen and Lewellen (2022), for example, estimate that insiders and affiliates own 9.0% shares of US mid-sized firms (in terms of total capital) and only 4.9% of large firms. In the context of the UK, Kyere and Ausloos (2021) show that the proportion of shares owned by insiders accounts for 3.87% (as of 2014). Shan, Troshani, and Tarca (2019) provide data from 44 developed and developing countries and show that insiders hold on average only 3.43% of shares (2003–2014). These estimates suggest that insiders hold a comparatively higher percentage of shares in Japanese corporate firms. These ownership characteristics may have an important ramification for firms’ risk and performance. Hence, it is imperative to examine the effect of insider ownership on default risk in a Japanese setting.

3. Literature review and hypothesis development

Following the seminal work of Jensen and Meckling (1978), studies have attempted to assess if, and under what conditions, insider ownership might mitigate or worsen agency problems. Literature provides evidence supporting both “managerial entrenchment” and “interest alignment” hypotheses. The “managerial entrenchment” hypothesis, conjecturing a “positive” relationship between insider ownership and default risk (Saunders, Strock, & Travlos, 1990; Gorton & Rosen, 1995; Wright, Ferris, Sarin, & Awasthi, 1996; Hobdari, 2008), argues that manager–owners become more empowered to expropriate firm resources, leading to a higher probability of default. We rely on two primary arguments as to how insider ownership increases firm default risk.

First, insider ownership gives managers substantial power to expropriate firms' resources (Saunders et al., 1990). Insiders with large equity ownership are likely to take on higher levels of risk to enhance returns (Switzer, Tu, & Wang, 2018; Switzer & Wang, 2013) because entrenched managers prioritize short-term personal benefits, eroding firms’ long-term value. External capital markets are likely to take this risk into account when pricing loans. Tanaka (2016) and Shuto and Kitagawa (2011) exhibit that higher managerial ownership is associated with lower credit ratings and higher yield spreads, implying that bondholders perceive a higher managerial equity stake as a risk-enhancing factor and consequently ask for a higher spread margin. Moreover, insider dominant firms face barriers to access to external financing (Hobdari, 2008), which forces them to resort to more risky and expensive sources of funds. A higher risk premium paid by firms with greater insider ownership adversely affects firms’ profitability, resulting in higher profit volatility.

Second, the literature shows that insider ownership can drive managerial behavior in a direction that affects stock price volatility, resulting in higher default risk. In an environment of information asymmetry, the stock price reacts to the news released by a firm. For instance, stock prices react positively to the timely release of favorable news, whereas they react negatively to bad news. Higher insider ownership provides managers with the controlling power to dictate firms’ disclosure decisions. While executives like to disclose good news immediately, they tend to hoard bad news (Haw et al., 2004). Hong, Lim, and Stein (2000) contend that in the absence of strong investor protection, executives are likely to withhold bad information to materialize a private gain. However, when the amount and time of information hoarding exceed the market tolerance, an unexpected release of information occurs, leading to a stock price crash (Zhong, 2022; Hutton, Marcus, & Tehranian, 2009). Hu et al. (2022) analyzed data from 40 countries and documented an inverse U-shaped association between insider ownership and stock price crash risk.

Empirical studies provide evidence supporting managerial entrenchment behavior when insider shareholding is high. For example, Berger, Imbierowicz, and Rauch (2016) found that insider ownership (non-CEO higher-level and lower-level managers) increases the probability of failure in the case of US banks. Barry, Lepetit, and Tarazi (2011) show a mixed relationship between insider ownership and default probability in the case of privately owned European banks. Using an international sample of financial firms, Switzer et al. (2018) document a positive effect of insider ownership on default risk. For a large sample of Taiwanese nonfinancial firms, Chiang, Chung, and Huang (2015) found a negative (positive) effect of managerial (directors) ownership on default risk.

Contrary to the management entrenchment hypothesis, the “alignment of interest” proposition, conjecturing a “negative” effect of insider ownership on default risk (Huang, Lin, & Huang, 2011; Pham, Suchard, & Zein, 2012; Core, Hail, & Verdi, 2014; Andreou, Antoniou, Horton, & Louca, 2016), argues that when the level of insider ownership is low or nil, managers are insufficiently motivated to exert optimum efforts and take on value-enhancing risks. Accordingly, increasing insider ownership helps align the interests of managers with those of outside shareholders; thus, the larger the insider ownership, the greater the motivation to engage in relevant and appropriate levels of risk-taking activities. Managers may financially benefit from such behavior via enhanced returns; at the same time, manager–owners are more likely to be interested in better managing the probability of default, leading to a negative insider ownership–default risk relationship. Moreover, as the insider stake increases, manager–owner’s interests coincide with those of outside shareholders, which, in turn, require less monitoring and supervision of managers, resulting in a decline in agency costs (Jensen & Meckling, 1978). Studies further show that managers with higher equity stakes tend to use lower corporate debt (Jensen, Solberg, & Zorn, 1992), resulting in limited default risk.

Literature further highlights managerial reputation as a risk-reducing factor. Managers might show risk-averse behavior if they weigh reputation more than pecuniary benefits. As a result, they are more likely to avoid high-risk projects that result in failure or financial difficulties (Bhattacharyya & Cohn, 2010). Instead, managers tend to preserve stability within the company by upholding their standing as competent agents of the shareholders’ wealth by avoiding undue risk-taking. This results in a decline in firms’ default risk.

In light of the foregoing, we draw the following non-directional hypothesis:

H1.

Insider ownership affects the default risk of Japanese firms.

4. Methodology

4.1 Data and sample

Our sample consists of 36,116 firm-year observations from 2004 to 2019. Consistent with prior studies, we exclude financial firms because the regulatory requirements are different for financial firms. The sample selection procedure has been discussed in detail in panel A of Table 1. Data relating to the main variables of interest were collected from the NIKKEI Corporate Governance Evaluation System (CGES). We collected the control variables from the EIKON database.

4.2 Dependent variable: Merton’s distance-to-default

We apply Merton’s distance-to-default as the primary dependent variable and a proxy of default risk. DD is widely applied in the literature as a market-based estimate of default risk (Chava & Purnanandam, 2010; Jokipii & Monnin, 2013; Anginer, Demirguc-Kunt, & Zhu, 2014). The DD model takes a firm’s market-related information into account in calculating the risk based on the theories of Merton (1974) and Black and Scholes (1973). The model can be described as follows:

(1)DDt=ln (VA,tXt)+(μ−12σA2)(T−t)σAT−t

Here, DD indicates the distance of a firm from its default threshold. A smaller value of DD represents a higher probability of default and vice versa. V_A = reflects the market value of assets with the assumption that market value follows a geometric Brownian motion; σA indicates the volatility of underlying assets; X_t indicates the default point value of a firm, which is calculated by adding all the short-term liability to the half of long-term liability; μ indicates the average return of the asset and T−t is set to 1 year.

The standard measure of DD, however, suffers from some important complications. For instance, it is extremely difficult to measure μ with an acceptable level of accuracy using high-frequency data over a long period of time. Moreover, the conventional estimate considers all short-term liabilities and half of the long-term liabilities while ignoring other liabilities. In addition, it is almost infeasible to accurately calculate the market value of assets. Hence, Duan and Wang (2012) suggest the following model to overcome the problems associated with the traditional measure of DD:

(2)DD=ln (VA,tXt) σAT−t

In equation (2), the default threshold is indicated by point X_t, which is estimated as follows:

current liability+0.5*Long−term Liability+δ *other liability

where δ is determined and calculated for sectors within each calibration group. specified and estimated for sectors in each calibration group. Empirical estimation provides evidence that equation (2) is a better measure of DD than the conventional one (Duan & Wang, 2012). Accordingly, we collect DD data from the “Credit Research Initiative (CRI),” a database maintained by the National University of Singapore. The CRI is based on the estimation technique proposed by Duan and Wang (2012), who tackle the problems associated with the conventional DD estimation technique. Moreover, the CRI data are widely used in the literature (Shin & Kim, 2015; Leroy & Lucotte, 2017; Nadarajah, Duong, Ali, Liu, & Huang, 2021).

4.3 Key independent variables: ownership structure

In our model, insider ownership is the primary independent variable. We consider three different proxies of insider ownership, namely, director (DIR), CEO (CEOIR) and employee (EMPIR) ownership. Each is defined as the percentage of shares they hold. The data source for insider ownership is the NIKKEI CGES database.

4.4 Control variables

Following the prior literature (Ali, Liu, & Su, 2018; Baghdadi, Nguyen, & Podolski, 2020), we adopted an array of control variables. First, we use the logarithm of total sales (LNS) to control for firm size. It is widely perceived that large firms are more stable than their smaller peers because they have greater access to resources and make sound investment decisions. Thus, firm size is perceived to positively influence DD. Next, we control for the capital-sales ratio (K/S) to assess the impact of a firm’s hard capital (tangible or fixed asset) on DD. Firms with increased investment in tangible assets are supposed to be more stable and have a lower probability of default. From this perspective, a positive relationship between K/S and DD is expected. Third, we control for firms’ market power or “free cash flow,” measured by the ratio of EBITDA to sales (Y/S). Since we are not able to measure free cash flows directly, we assume that free cash flow is strongly correlated with operating income (EBITDA). Higher free cash flow at a firm’s disposal indicates greater strength and capability to meet firms’ credit obligations and may enhance firms DD.

Firms may have numerous other discretionary capital expenditures that may not be known or adequately monitored. To control for such expenditures, we apply the ratio of R&D expenditures to property, plant and equipment (PPENT), denoted by R&D/K. We assign a value of zero for missing observations of R&D/K to retain the sample size. It is expected that firms with a higher R&D/K ratio will make more informed and cautious investment decisions, leading to better use of discretionary investments resulting in higher returns. Hence, firms with a higher R&D/K ratio are assumed to have a lower default probability. Further, a dummy variable (RDUM) is introduced by assigning a value of 1 for firms that report R&D data and zero otherwise. This will allow the intercept term to capture the mean of R&D/K for missing values, indicating the availability of R&D data.

We also control for firms’ growth opportunity, which is proxied by the ratio of firms’ capital expenditure to PPENT (I/K). Firms with higher I/K are likely to invest in riskier projects, which may result in higher default risk. We then apply the firm’s profitability, net income divided by total assets. Return on assets (ROA), a standard accounting performance indicator, reflects a firm’s ability to generate sufficient income to run its operations smoothly. As a result, we can expect ROA to have a negative effect on firms' default probability. Leverage, the ratio of total debt to total equity, indicates a firm’s strength in meeting its debt obligations. The lower the leverage ratio, the greater the ability of a firm to meet its debt obligations and vice versa. Hence, a higher leverage ratio increases a firm’s default probability. Finally, we control for firm age, the natural log of the difference between the date of incorporation and the data collection period. Firms with a longer operating history are deemed to be more stable than younger firms; thus, we can hypothesize a positive effect of firm age on DD. In addition, we winsorize the control variables at the 1% level in both tails to reduce the possible impact of outliers. The detailed description of variables is provided in the Appendix.

4.5 Estimation model

We test our hypothesis (H1) applying the following FEs model:

(3)DDi,t=α0+β1INSOWNi,t+β2LNSi,t+β3K/Si,t+β4Y/Si,t+β5R&D/Ki,t+β6RDUMi,t+β7I/Ki,t+β8ROAi,t+β9LEVERAGEi,t+β10AGEi,ti,t+Yeart+εi,t

Here, DD indicates the measure of default risk calculated based on equation (3) (Merton’s distance-to-default, DD); INSOWN = insider ownership structure proxy measure (DIR = director ownership, CEOIR= CEO ownership and EMPIR = employee ownership); LNS = firm size; K/S = capital to sales ratio; Y/S = EBITDA to sales ratio; R&D/K = ratio of research and development expenditures to PPENT; RDUM is used as a dummy variable that takes 1 if R&D data exists or 0 for missing data; LEVERAGE = total debt to firms’ total assets’ I/K = ratio of firms’ capital expenditure to PPENT and ROA = return on assets. We also include firms’ Age calculated as the natural logarithm of AGE; YEAR = indicates year dummies and ε_i = the error term.

5. Empirical analysis

5.1 Summary statistics and correlation matrix

The descriptive statistics are presented in Panel B of Table 1. We report a mean of 4.46 and a standard deviation of 2.43 for DD. The mean (median) for the three proxies of insider ownership are as follows: director ownership (DIR) 7.54% (1.65%), CEO ownership (CEOIR) 5.64% (0.57%) and employee ownership (EMPIR) 2.89% (1.80%). Our first insider ownership proxy (DIR) ranges from 0 to 99.40% with the corresponding standard deviation of 12.08. The second measure of insider ownership, CEOIR, varies from 0 to 88.52%, with a standard deviation of 10.53. The employee ownership ranges from 0%to 39.90%, with a standard deviation of 3.45.

As for the control variables, we find the mean (median) for LNS to be 10.67 (10.54), which ranges between 2.08 and 16.96. The mean (median) of K/S is 0.33 (0.25). We then report the mean (median) of Y/S to be 0.084 (0.075), with a range between −0.25 and 0.35. On average, firms’ average I/K was 0.15% during the sample period, with a standard deviation of 0.36%. The firms included in our sample have an average ROA of 5.56% with a standard deviation of 6.66%. We also note that sample firms had a 19% leverage ratio during the sample period. Finally, the log of the average age of the firm is 2.99, with a standard deviation of 0.94.

5.2 Correlation matrix

The correlation matrix is displayed in Panel C of Table 1. The univariate test revealed a negative and statistically significant association at the 5% level between DD and all three insider ownership proxies. This result provides preliminary support for the theory that an uptick in insider ownership leads to a drop in DD. In the case of the control variables, LNS, ROA, K/S, Firm Age, Y/S have a positive correlation with DD and are significant at the 5% level. Although not significant, R&D/K and RDUM also showed a positive correlation with DD. On the other hand, LEVERAGE and I/K show a negative correlation with DD, although the latter is not statistically significant. These results are coherent with common perceptions and prior literature. Noticeably, the highest positive and significant correlation among the independent variables was 0.95 (CEO ownership and director ownership). However, this relationship does not cause multicollinearity problems because we treat them separately. To check whether potential multicollinearity issues do not bias our subsequent tests, the variance inflation factor (VIF) is calculated for independent variables used in equation (3). We find the value of each independent variable less than 10, indicating that the bias concerns have been mitigated.

5.3 Baseline regression results

We present the results of the baseline regression in Table 2. To dissect the unincorporated impact of insider ownership on DD, we initially run the regression with proxies of insider ownership – director, CEO and employee ownership – without including the control variables. The results are exhibited in columns 1–3 of Table 2. All the coefficients of insider ownership show a negative association with distance-to-default, which suggests that insider ownership decreases the distance to default, thus increasing the probability of bankruptcy. Our first variable of interest, director ownership (DIR), has a negative coefficient (−0.005) at the 1% significance level. As mentioned earlier, increased director ownership provides them with greater control rights over firms’ resources. Moreover, directors with higher equity ownership may feel more secure in terms of job losses attributed to outside pressure. Hence, directors tend to disregard internal governance and risk control mechanisms, resulting in higher firm-specific risks. Similarly, we find a negative association of CEO (−0.003) and employee (−0.033) ownership (both significant at the 5% level) with DD. We also show the impact of total insider ownership (sum of all three forms of insiders) on firms’ default risk in column (7) of Table 2, and our results indicate that insider ownership is negatively associated (−0.005) with DD at the 1% significance level. Our findings concur with earlier studies by Tanaka (2016) and Yen, Lin, Chen, and Huang (2014), which document a positive influence of insider ownership on credit risk for Japanese and Taiwanese firms, respectively.

Subsequently, we introduce the control variables and rerun the regressions. Results are reported in columns 4–6. Consistent with the earlier results, we find that all the coefficients of insider ownership are negative with DD. For director and CEO ownership, the values of the coefficients increase and are significant at the 1% level; although the coefficient of employee ownership decreases, it remains statistically significant at the 5% level.

Besides evaluating the statistical significance of our variable of interest, we further assess, following Kabir et al. (2020), the economic significance of this variable. As an illustration, in column 4 of Table 2, a one standard deviation rise in director ownership (12.079) is related to a 9.66% drop in DD (calculated as 12.079*0.008). A similar pattern can be observed in columns 5 and 6 of Table 2, where an increase of one standard deviation in CEO and employee ownership is associated with a reduction in DD by 12.62 and 2.41%, respectively.

Most of our control variables show statistically significant effects, with expected signs on DD, and the effects are consistent across the models. We find firms’ size (LNS) to be positively associated with the DD, which implies that large firms are subject to a lower default probability. This finding is similar to that of Nguyen (2011), who found that larger firms have a low default probability in Japan. Results also show a positive effect of firms’ fixed assets (K/S) on DD, which implies that firms investing substantially in tangible assets can generate steady earnings by utilizing long-term assets more efficiently. Therefore, capital-intensive firms are subject to a low probability of default. Similarly, consistent with our prediction, Y/S and ROA exhibited a positive and statistically significant (at the 1% level) association with DD. The results further indicate that R&D expenditure increases firms’ DD or decreases the default probability. Higher expenditures in R&D provide firms with long-term stability and enhanced earnings capacity, which in turn reduces firms’ default probability. We also find I/K and leverage are negatively related to DD. This suggests that the higher the usage of leverage, the greater the risk of default for a firm.

5.4 Addressing endogeneity

In the previous section, we showed that insider ownership increases default risk. There is a chance that such a causal effect could be due to an endogeneity issue. Unobserved heterogeneity, sample selection bias and simultaneity are the three primary factors that contribute to endogeneity issues. Our baseline regression results deal with unobserved heterogeneity as we used a firm-fixed effect regression model. We consider sample selection bias and simultaneity problems in this section.

Results derived from the baseline regression could be biased owing to endogeneity concerns arising from simultaneity. When a firm faces high default risk, insiders (such as executives, key stakeholders or significant shareholders who are part of the firm’s management or board) may become more risk averse. Insiders, fearing potential financial loss and damage to their reputation, may reduce their ownership stake to mitigate personal financial risk. This behavior is particularly likely if insiders perceive that the firm’s default risk is outside their control or too high to manage effectively.

5.4.1 PSM regression results

Our results could be biased due to a possible endogeneity issue arising from sample selection bias. In the context of studying the association between insider ownership and default risk, sample selection bias could occur if the firms included in this study are not representative of the broader population of firms we are interested in or if the way they were selected into the sample is related to the variables of interest (insider ownership and default risk). To mitigate the potential endogeneity concerns, we adopt the approach of Ikeda et al. (2018) and Lehmann (2019), who employ the propensity score matching technique. The process consists of two sequential steps. First, we introduce a dummy variable. The variable INSDUM is assigned a value of 1 for firms with insider ownership (including directors, CEOs and employees) exceeding 5% and a value of 0 for all other firms. For the second stage, we perform a logit regression on the dummy variable and other explanatory factors, excluding the proxies for insider ownership. The propensity scores for each firm-year observation are derived from the estimated values of each model. We perform matching without replacement and mandate that the propensity scores for each matched pair must differ by no more than +1.0%. The results are displayed in Panel A of Table 3. The pseudo R² for the regression exhibits a substantial value (0.255); however, it experiences a large decrease for the post-match sample (0.001). The test involves recalculating the logit model for the post-match sample. Panel B of Table 3 exhibits results where the t-statistic is not significant among all variables for the postmatch difference in firms’ characteristics. Finally, we present the multivariate regression results in Panel C of Table 3, where a significant negative impact of CEO and director ownership on DD is observed. Most of the control variables also showed signs and effects on DD, similar to the results of baseline regression. Overall, the PSM technique suggests that the baseline regression results are not driven by the endogeneity issue.

5.4.2 Endogeneity test: 2SLS regression

Another possible source of endogeneity in our study is reverse causality, which refers to the possibility that, instead of insider ownership affecting default risk, it is actually the default risk of a firm that influences the level of insider ownership. The two-stage least squares (2SLS) regression technique is a useful tool to address a possible endogeneity concern arising from reverse causality. However, this technique requires identifying an instrument variable. Studies frequently apply the industry average as an exogenous instrument for firm-level endogenous variables (Liu, Miletkov, Wei, & Yang, 2015). We can expect a firm’s ownership structure to be positively associated with the industry average. However, it cannot be said, with a reasonable level of accuracy that a firm’s default risk is mediated by the average ownership structure of the industry. Rather, an individual firm’s distinctive ownership structure is likely to affect the firm’s probability of default.

Panel A of Table 4 displays the findings of the initial regression analysis, where we regress the proxies of insider ownership on the chosen instrumental variable as well as the same set of control variables used in the baseline regression model. In accordance with the theory of instruments, there is a positive correlation between the industry average of insider ownership proxies and the insider ownership variables (DIR, CEOIR and EMPS) at the 1% significance level. The calculated F-statistics for all three regressions are significantly high, suggesting that the industry average, which we used as an instrument, is appropriate.

Furthermore, we conduct the Cragg–Donald’s Wald F weak-instrument test and find that the p-value of the statistics is 0.000 for both regressions, thereby rejecting the null hypothesis that the model’s instrument is weak. The F-statistic values we obtained exceed the critical values reported by Stock and Yogo (2005), confirming that our instrument is not weak.

In the second stage, we utilize the residuals and fitted values of insider ownership proxies obtained from the first-stage regression model as a test variable. The findings from the second-stage regression analysis are displayed in Panel B of Table 4, validating the results obtained from the initial regression analysis. In line with our prior findings, we observed a negative correlation between the coefficient of the residual values of all three insider ownership proxies and the DD, and all coefficients were significant at the 1% level. Hence, our results confirm our hypothesis that insider ownership increases firms’ default risk.

5.5 Robustness test

5.5.1 Robustness test: alternative proxy for default risk

To reaffirm our earlier results, we apply the Altman Z-score, a well-established model for predicting bankruptcy. It is used in multiple discriminant analysis (MDA) to assess a firm’s probability of bankruptcy within a two-year period. This model combines five different financial ratios, and each of these ratios is given a specific weight, and the sum of these weighted ratios forms the Z-score. A lower Z-score indicates a higher risk of bankruptcy (lower solvency), while a higher score suggests a firm is financially stable. The Altman Z-score is calculated as follows:

Altman Z−Score=1.2A+1.4B+3.3C+0.6D+1.0E

where A = working capital/total assets; B = retained earnings/total assets; C = earnings before interest and tax/total assets; D = market value of equity/total liabilities and E = sales/total assets. Panel A of Table 5 exhibits the results. It is observed that all three variables representing insider ownership – director, CEO and employee – show a decreasing effect on firms’ stability, as measured by the Altman Z-score. This finding reaffirms the managerial entrenchment hypothesis, consistent with the baseline regression.

5.5.2 Robustness check: Fama Macbeth test

We further checked the robustness of our baseline regression results by applying the Fama and MacBeth (1973) estimation technique. Ownership structure does not vary significantly across time; hence, cross-sectional analysis is believed to suit our data well. To test this proposition, we perform a new regression analysis in two stages. First, we performed cross-sectional regression for each period. Second, the average is estimated for the intercept and slope coefficients obtained from the cross-sectional regression, while the standard errors are adjusted for cross-sectional correlations. The results are shown in Panel B of Table 5. The results confirm our earlier finding that insider ownership increases firms’ default probability.

6. Channel analysis

Our analysis so far shows that insider ownership increases firms’ default risk. We identify the channels that mediate the effect of insider ownership on default risk. To achieve this objective, we measure firms’ risk using accounting and market-based factors, including ROA volatility and stock price volatility. Following Baron and Kenny (1986) and Francis, Hasan, Liu, Wu, and Zhao (2021), we use two-stage procedures to investigate if insider ownership proxies affect DD through these risk channels. We first regress insider ownership variables on risk proxies (ROA volatility and stock price volatility). Then we include risk proxies as independent variables alongside the main variable of interest and regress on DD.

6.1 ROA volatility

We estimate the ROA volatility, an accounting-based measure of risk, by calculating the standard deviation of the yearly ROA for the current year (t) and the two preceding years (t-2). The results, presented in columns 1–3 of Table 6, demonstrate a strong positive and statistically significant (at a 1% level) influence of all the indicators of insider ownership on the volatility of return on assets (ROA), implying that a firm’s risk, reflected in ROA volatility, increases as insider ownership increases. When we include ROA volatility in the regression along with the proxies of insider ownership, the degree of negative effect of insider ownership on DD decreases, which implies that ROA volatility functions as a channel that mediates the relationship between insider ownership and DD. Following Hasan (2020) and Francis et al. (2021), we estimate the mediation effect of ROA volatility. The mediation effect refers to the reduction in the coefficient of insider ownership proxies that occurs when the firm’s ROA volatility is included as an additional explanatory variable in the regression model. The mediation effect on the volatility of return on assets is of significant economic magnitude. For instance, the total impact of director ownership on the probability of default is −0.008 (shown in column 4 of Table 2), whereas the direct impact of director ownership on the probability of default is −0.004 (shown in column 4 of Table 6). The indirect mediation effect, defined as the difference between the total effect and the direct effect, has a value of −0.004. Therefore, the mediation effect accounts for 50% (i.e. 0.004/0.008) of the total effect. Similarly, the mediation effects of return on assets (ROA) volatility with CEO ownership and employee ownership are 75 and 57%, respectively. The mediation effect of ROA volatility is only partial, as seen by the fact that the coefficient on insider ownership proxies remains substantial in columns 4–6 of Table 6. With increased ownership, insiders establish greater control over firms’ resources, which tempts them to invest in high-risk return projects; as a result, returns may fluctuate enormously. Such unsteady investment returns make it difficult for firms to meet their debt obligations. Hence, our results confirm that a firm’s risk-taking tendency, measured by ROA volatility, functions as a channel between insider ownership and default risk.

6.2 Stock price volatility

We use stock price volatility as our second measure of risk, which is a market-based risk factor. Prior studies, including Brogaard, Li, and Xia (2017) and Baghdadi et al. (2020), have used stock price volatility as an indicator of firms’ default risk. In line with this tradition, we examine whether stock price volatility mediates the relationship between insider ownership and a firm’s default risk. We estimate stock price volatility as the standard deviation of the daily stock return, which demonstrates the range within which stock prices may fluctuate. Rapid fluctuations in the stock price usually signal business failure. The regression results are presented in Table 7. Similar to ROA volatility, the results of stock price volatility show a significant positive association across all proxies of insider ownership, indicating a higher risk appetite of insider owners reflected in firm stock price volatility. When stock price volatility is added to the regression alongside insider ownership proxies, the negative impact of insider ownership on DD is reduced. This suggests that stock price volatility acts as a mediating factor in the relationship between insider ownership and DD.

We further calculate the mediation effect of stock price volatility. For example, the total impact of director ownership on the probability of default is −0.008 (shown in column 4 of Table 2), and the direct impact of director ownership on default risk is −0.007. The indirect mediation effect is calculated as the difference between the overall effect and the direct effect, which is −0.001. Therefore, the mediation impact accounts for approximately 12.5% of the overall influence on director ownership. Similarly, we find the mediation effect of stock volatility on CEO ownership and employee ownership to be 58 and 60%, respectively. Similar to ROA volatility, the mediation effect of stock volatility is partial because the coefficients of insider ownership proxies are significant in columns 4–6 of Table 7. Companies with high stock price volatility may find it difficult to attract potential investors, which makes the company financially less solvent, leading to an increased probability of default.

7. The effect of external governance mechanisms

Additionally, we examine whether the external governance mechanisms mitigate the negative effect of insider ownership on DD. Following previous literature, we consider institutional ownership, analysts' coverage and main bank relationships to be significant external governance measures.

7.1 Institutional ownership

Prior studies provide evidence that institutional shareholders can significantly affect firms’ default risk (Li, Kannan, Rau, & Yang, 2022; Kabir et al., 2020). As per the “management disciplining” hypothesis, institutional shareholders actively monitor managers to discourage their self-serving behaviors (Lima & Hossain, 2018). Hence, the presence of large institutional shareholders is likely to mitigate the negative effect of insider ownership on DD. To test this hypothesis, we split firms with higher and lower intuitional ownership based on the median value. Regression results are provided in Table 8. The results show that the coefficient of director ownership is lower for firms with higher institutional shareholdings. On the other hand, the value of the CEO ownership coefficient increases as institutional shareholding increases. Results also show that employees’ equity holding increases (or decreases) DD for firms with high (or low) institutional ownership, although the coefficients are not statistically significant, suggesting that institutional ownership mitigates the negative impact of insiders’ equity stakes on DD. We also performed the Chow test to investigate whether the difference in coefficients of insider ownership proxies between high and low institutional ownership is statistically significant. Our results in Table 8 show that the value of Chi-square is significantly higher than the critical value, indicating that the differences in coefficients of insider proxies between high and low institutional firms are significantly different at the 1% significance level.

7.2 Analyst coverage

The literature also provides evidence that analyst coverage plays a monitoring role in mitigating information asymmetry and controlling agency costs (Luo, Wang, & Wu, 2023; Sun, Liu, & Lan, 2010; Kim, Lu, & Yu, 2018). Analyst coverage can be an important substitute for external governance to discourage managerial opportunism. Analysts extract information from firms’ financial statements, attend corporate conferences to produce reports about the prospects of firms and provide consultancy services by circulating those reports to existing and potential investors. Analysts often forecast firms’ future performance, which serves as an external pressure on managers to achieve the target. In this sense, analyst coverage translates into a good quality governance mechanism (Mouselli, Abdulraouf, & Jaafar, 2014), encouraging managers to be more cautious in risk-taking activities, which, in turn, alleviates default risk. Therefore, we can hypothesize that higher analyst coverage mitigates the negative impact of insider ownership on DD.

To test this hypothesis, we retrieve data from the I/B/E/S database. Analyst coverage simply indicates the total number of unique analysts issuing earnings forecasts for a firm in an accounting year. Following the conventional method, all missing values are replaced by zero. All firms are divided into two groups: the first group includes all firms that have higher than the median number of analyst coverage and take the value 1. The second group includes all the firms with a lower than median number of analyst coverage and takes the value 0.

The results are presented in Table 9, which illustrate that the negative association between insider ownership and DD is greater for firms with low analysts’ coverage (columns 4–6) than for firms with higher analysts’ coverage (columns 1–3). The coefficient for director ownership in column 4 is −0.009, which decreases to −0.004 in column 1, both statistically significant at the 1% level. Likewise, the coefficient of −0.012 for CEO ownership in column 5 decreases to −0.007 in column 2. A reduction in negative impact is observed for employee ownership as well. This implies that the extent of the negative association between insider ownership and default risk is lower for firms with high analyst coverage than for their peers with low analyst coverage. In addition, the Chow test shows that the differences in coefficients of insider ownership proxies between high and low analyst coverage are statistically significant at the 1% significance level.

7.3 Main bank ties

The Japanese corporate governance system has traditionally featured an important characteristic known as the main bank system, which plays a critical corporate disciplinary role. The monitoring activities of main banks not only help reduce information asymmetry and agency costs but also administer the risk-taking behaviors of managers (Prowse, 1990). Miyajima, Ogawa, and Saito (2018) argue that the main banking system still prevails in Japan, which coordinates among different banks to meet the external financing needs of firms. Being the largest external financer, a main bank or a lead bank monitors the client firms. This has an important implication for the default risk of Japanese firms. Hence, it is important to investigate whether the main banks moderate the relationship between insider ownership and default risk.

Following Miyajima et al. (2018), we identify the main bank relationship based on three distinct criteria: a firm has a long-term (five years or more) relationship with a lead bank, the lead bank holds an equity stake in the client firm and client firms’ borrowing from the lead bank amounts to no less than the industry median. We assign 1 to firms that meet the above three criteria and zero to the remaining firms. We run the regression for these two groups of firms separately. In columns 1–3 of Table 10, we present the results for firms linked with a particular lead bank known as the main bank. In columns 4–6 of Table 10, we report the results for firms without main bank linkage. It is observed that the negative relationship between insider ownership and DD is weaker for firms linked with main banks than for firms without such ties, implying that firms with the main bank tie are less exposed to default risk than firms without the main bank relationship. Finally, the Chow test shows that the differences in coefficients of insider ownership proxies between firms with a main bank and firms without a main bank are statistically significant.

8. Conclusion

The implication of insider ownership on firms’ default risk is certainly an intriguing area of research. On the contrary, research in this area remains scarce. Moreover, research in this area in the context of Japanese firms appears yet to be conducted. The forgoing precisely forms the motivation of the present study. Using 36,116 firm-year observations over the 2004–2019 period, we show a negative effect of insider ownership on the distance to default, our proxy for default risk, implying that insider ownership enhances a firm’s probability of default. Our finding supports the well-known “managerial entrenchment” hypothesis of firms. We also demonstrate that firms' risk-taking behavior, as measured by ROA volatility and stock price volatility, transmits this effect. We further show that external corporate governance mechanisms such as institutional ownership, analyst coverage and main bank ties tend to mitigate the risk-enhancing effect of insider ownership. Our results are robust to endogeneity issues and alternative default risk proxies.

Our findings have important policy implications for managers, investors and policymakers, especially in the Japanese setting. First, insider ownership is likely to enhance a firm’s probability of default. Hence, in the first instance, policymakers may consider regulating a ceiling for insider ownership of firms. Second, in light of the various risk-enhancing channels, policymakers may further consider adopting guidelines on the risk behavior of firms, with particular attention to ROA volatility and stock price volatility. Third, given the fact that various factors may help reduce the probability of risk in the presence of insider ownership, policymakers may also consider requiring or encouraging firms to embrace institutional owners in their ownership structures, foster ties with main banks and invite analysts to cover firms’ operations and earning prospects. Our findings may also make existing and future investors more vigilant and prudent; shareholders may keep away from companies characterized by high insider ownership and low or no institutional ownership, bank ties and analysts’ coverage. ROA volatility and stock price volatility may also keep investors away. The foregoing findings and policy implications are expected to influence ownership structures and firms’ inclination for risk-enhancing and reducing factors, in the case of Japanese firms.

There are some issues that the current research cannot cover due to its scope limitations. For instance, future research might analyze the ownership–default risk relationship for financial institutions to compare whether this relationship varies between financial and nonfinancial firms. In addition, future research might delve into a broader analysis of how differences in insider ownership between pre- and post-pandemic correlate with the default risk. Although insider ownership and default risk relationships in the context of Japan provide an important insight for corporate governance and firms’ decision-making, a regional analysis including other advanced countries in the region such as South Korea and Taiwan will shed light on the spillover effects of Japanese corporate ownership and the default risk nexus in the region. Such an analysis will enrich our understanding as to whether a regional common approach to mitigating corporate risk is more effective and conducive to reducing agency costs. In addition, comparing the Japanese case with its Western counterparts and applying common parameters would be interesting.

Table 2

Baseline regression results

	(1)	(2)	(3)	(4)	(5)	(6)	7
Dependent variable	DD	DD	DD	DD	DD	DD	DD
DIR	−0.005***			−0.008***
	(−4.362)			(−9.013)
CEOIR		−0.003**			−0.012***
		(−2.302)			(−10.559)
EMPIR			−0.033***			−0.007**
			(−5.338)			(−2.407)
INSIDER							−0.005***
							(0.001)
LNS				0.146***	0.124***	0.158***	0.122***
				(19.634)	(15.701)	(21.754)	(0.008)
K/S				0.767***	0.756***	0.779***	0.754***
				(19.734)	(18.275)	(20.044)	(0.041)
Y/S				5.230***	5.177***	5.171***	5.163***
				(31.092)	(28.914)	(30.692)	(0.179)
R&D/K				0.012***	0.009**	0.012***	0.009**
				(3.170)	(2.277)	(3.152)	(0.004)
RDUM				0.335	0.116	0.403	0.106
				(0.560)	(0.190)	(0.673)	(0.614)
I/K				−0.094***	−0.091***	−0.101***	−0.096***
				(−3.737)	(−3.539)	(−4.027)	(0.026)
ROA				0.031***	0.031***	0.029***	0.031***
				(16.587)	(15.841)	(15.720)	(0.002)
LEVERAGE				−5.800***	−5.789***	−5.857***	−5.795***
				(−95.521)	(−88.926)	(−96.919)	(0.065)
Age				0.026*	0.017	0.066***	0.014
				(1.834)	(1.120)	(4.936)	(0.016)
Constant	4.538***	4.510***	4.589***	2.807***	3.258***	2.463***	3.317***
	(404.505)	(427.876)	(247.483)	(4.626)	(5.224)	(4.062)	(0.624)
Observations	40,655	35,337	40,927	36,116	31,618	36,152	31,618
R-squared	0.783	0.801	0.782	0.541	0.546	0.540	0.545
Firm fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Note(s): Table 2 presents estimation results of Equation (3) using firm fixed effect model. The dependent variable is DD. Columns 1–3 shows results without the control variables. Columns 4–6 include the control variables. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denote statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 3

Addressing endogeneity using propensity score matching

Panel A: Pre-match propensity score regression and post-match diagnostic regression
	Pre-match	Post-match
	(1)	(2)
Dependent variable	INSDUM
LNS	−0.689***	0.034
	(−61.248)	(0.496)
K/S	−0.680***	0.042
	(−12.673)	(0.678)
Y/S	−1.713***	−0.620
	(−5.794)	(−0.888)
R&D/K	−0.021***	0.004
	(−3.424)	(0.549)
I/K	−0.492***	−0.028
	(−8.062)	(−0.594)
ROA	0.050***	0.009
	(13.042)	(0.167)
LEVERAGE	0.677***	0.179
	(6.799)	(1.481)
Age	−0.710***	−0.055
	(−33.957)	(−0.214)
Constant	11.145***	−0.234
	(83.717)	(−1.472)
Observation	36,157	17,471
Pseduo R2	0.255	0.001

Panel B: Post-match difference in firm characteristics
Variable	Treated	Control	Difference	t-test
LNS	11.328	11.263	0.06	0.43
K/S	0.35	0.355	−0.01	−0.67
Y/S	0.085	0.085	0.00	−0.46
R&D/K	0.162	0.146	0.02	0.3
I/K	0.152	0.155	0.00	−0.42
ROA	5.1	4.907	0.19	1.08
LEVERAGE	0.187	0.184	0.00	1.06
Age	3.371	3.398	−0.03	−1.01

Panel C: Post-match regression analysis
Dependent variable	Distance to default	Probability of default	CDS spread
	(1)	(2)	(3)
DIR	−0.012***
	(−3.758)
CEOIR		−0.010**
		(−2.681)
EMPIR			0.001
			(0.229)
LNS	0.184***	0.155***	0.168***
	(17.609)	(13.608)	(15.320)
K/S	0.864***	0.956***	0.968***
	(17.782)	(17.578)	(18.450)
Y/S	7.046***	6.798***	6.941***
	(25.612)	(22.592)	(23.726)
R&D/K	0.027***	0.029***	0.031***
	(5.437)	(5.884)	(6.152)
RDUM	0.900	0.730	0.847
	(1.509)	(1.240)	(1.437)
I/K	0.021	0.040	0.047
	(0.686)	(1.273)	(1.505)
ROA	0.062***	0.065***	0.068***
	(16.850)	(16.342)	(17.450)
LEVERAGE	−6.022***	−5.909***	−5.877***
	(−60.706)	(−52.212)	(−54.447)
Age	−0.028	−0.021	−0.009
	(−1.218)	(−0.822)	(−0.379)
Constant	1.501**	1.863***	1.589***
	(2.425)	(3.040)	(2.595)
Observations	17,471	13,391	14,886
R-squared	0.340	0.342	0.342
Fixed year effects	Yes	Yes	Yes
Firm fixed effects	Yes	Yes	Yes

Note(s): Table 3 reports the results of the propensity score matching procedure to investigate the effects of insider ownership on default risk. Panel A reports the parameter estimates from the logit model used to estimate propensity scores. The dependent variable INSDUM in column (1) and (2) of Panel A is an indicator variable set to one if the firm has insider ownership score more than median in a given year, zero otherwise. Panel A reports the pre-match propensity score regression and post-match diagnostic regression. Panel B reports the univariate comparisons of firm characteristics between treated and control firms the corresponding t statistics. Panel C reports multivariate results relating to default risk and insider ownership. The dependent variables are distance-to-default. t-statistics are reported in parentheses. Variables are defined in Appendix and are winsorized at the 1–99% levels. ***, ** and * indicate statistical significance of coefficient estimates at the 1, 5 and 10% levels, respectively

Source(s): Table created by the authors

Table 4

Addressing endogeneity using 2SLS regression

	(1)	(2)	(3)	(4)	(5)	(6)
	Panel A: First stage			Panel B: Two stage regression results
Dependent variable	DIR	CEOIR	EMPS	DD	DD	DD
Industry average	6.885***	0.305***	0.855***
	(9.38)	(9.93)	(4.88)
DIR				−0.438***
				(−7.00)
CEOIR					−0.213***
					(−11.47)
EMPS						−0.210***
						(−13.53)
LNS	−1.500***	−1.168***	−0.206***	0.823***	0.407***	0.128***
	(−3.01)	(−3.73)	(−7.22)	(8.58)	(16.71)	(16.23)
K/S	−1.973***	−1.592***	−0.451***	1.884***	1.372***	0.781***
	(−10.08)	(−8.98)	(−7.26)	(10.96)	(18.44)	(16.87)
Y/S	0.689	1.336	−2.975***	4.849***	5.280***	4.561***
	(0.72)	(1.55)	(−9.80)	(7.83)	(15.92)	(21.04)
R&D/K	−0.087***	−0.050***	−0.042***	0.064***	0.045***	0.022***
	(−3.97)	(−2.55)	(−5.90)	(4.10)	(4.51)	(3.60)
RDUM	−5.156*	−4.350	1.015	3.192***	1.819***	1.075***
	(−1.66)	(−1.56)	(0.97)	(7.23)	(6.63)	(4.25)
I/K	0.359**	0.562***	−0.857***	−0.246	−0.216**	−0.288***
	(2.40)	(4.15)	(7.41)	(−1.57)	(−2.02)	(−2.71)
ROA	0.212***	0.166***	−0.012***	−0.047***	0.008	0.043***
	(19.19)	(16.51)	(−3.71)	(−3.01)	(1.52)	(18.03)
LEVERAGE	8.523***	7.762***	−0.501***	−9.426***	−7.405***	−5.864***
	(24.47)	(24.64)	(−4.58)	(−16.33)	(−37.49)	(−70.56)
Age	−5.311***	−3.802***	−0.074***	2.408***	0.941***	0.007
	(−3.49)	(−4.78)	(−3.50)	(6.95)	(11.32)	(0.50)
Constant	41.867***	29.958***	2.527**	−17.030***	−5.283***	2.882***
	(13.26)	(10.49)	(2.39)	(−6.27)	(−7.67)	(10.06)
Observations	31,621	31,621	36,156	31,621	31,621	36,156
R-squared				0.421	0.422	0.422
Firm fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes
F-statistics	58.23**	45.21**	46.21**
Cragg–Donald F Statistics	42.34	38.19	38.90
Stock–Yogo Weak ID test value	16.38	16.38	16.38

Note(s): Table 4 represents the regression results for the relationship between insider ownership and firm default risk using 2SLS. Panel A presents the result of the first stage regression and panel B presents the second stage regression. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denote statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 5

Robustness test

	(1)	(2)	(3)	(1)	(2)	(3)
	Panel A: Alternative proxy of default risk			Panel B: Alternative estimation technique
Dependent variable	Z_score	Z_score	Z_score	DD	DD	DD
DIR	−0.006***			−0.002**
	(−4.231)			(−2.794)
CEOIR		−0.010***			−0.004***
		(−5.447)			(−3.533)
EMPIR			−0.036***			−0.024***
			(−5.227)			(−3.363)
LNS	−0.115***	−0.023	−0.116***	0.200***	0.176***	0.198***
	(−3.824)	(−0.683)	(−3.842)	(7.146)	(6.649)	(7.388)
K/S	−0.224***	−0.153*	−0.218***	1.047***	1.030***	1.030***
	(−3.104)	(−1.958)	(−3.014)	(17.819)	(15.866)	(17.234)
Y/S	1.147***	1.028***	1.149***	5.199***	5.073***	5.094***
	(5.966)	(5.109)	(5.974)	(15.927)	(15.416)	(16.058)
R&D/K	0.006	0.003	0.006	0.076***	0.052***	0.073***
	(1.419)	(0.646)	(1.419)	(3.812)	(3.930)	(3.884)
RDUM	0.244	0.125	0.248	0.949***	1.036***	0.993**
	(0.350)	(0.141)	(0.355)	(3.228)	(3.163)	(2.961)
I/K	0.072***	0.043*	0.072***	−0.315***	−0.309***	−0.364***
	(3.243)	(1.951)	(3.241)	(−4.024)	(−3.509)	(−4.484)
ROA	0.085***	0.079***	0.085***	0.036***	0.036***	0.035***
	(47.711)	(42.392)	(47.550)	(5.700)	(5.008)	(5.693)
LEVERAGE	−4.245***	−4.284***	−4.233***	−5.753***	−5.764***	−5.789***
	(−40.664)	(−37.427)	(−40.466)	(−18.600)	(−16.311)	(−18.502)
Age	−0.386***	−0.184***	−0.353***	−0.137***	−0.154***	−0.120***
	(−12.523)	(−5.016)	(−12.099)	(−4.371)	(−4.565)	(−3.809)
Constant	5.811***	4.392***	5.768***	2.043***	2.276***	2.047***
	(7.568)	(4.566)	(7.514)	(6.032)	(6.592)	(7.020)
Observations	32,416	28,583	32,438	36,120	31,621	36,156
R-squared	0.806	0.816	0.806	0.341	0.331	0.332
Firm fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes

Note(s): Panel A of Table 5 presents results for analyzing the effect of insider ownership on firm default risk using DD spread as a proxy of default risk. The dependent variable is Altman Z-score. Panel B of Table 5 provides the results of regression analyzing the effect of insider ownership on firm default risk using Fama and Macbath (1973) as an alternative estimation technique. The dependent variable is DD. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denote statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 6

Channel analysis: ROA volatility

	(1)	(2)	(3)	(4)	(5)	(6)
Dependent variable	ROA_VOL			DD
DIR	0.009***			−0.004***
	(3.386)			(−2.974)
CEOIR		0.014***			−0.003*
		(4.756)			(−1.949)
EMPIR			0.070***			−0.003**
			(5.515)			(−2.314)
ROA_VOL				−0.049***	−0.048***	−0.049***
				(−15.840)	(−15.078)	(−15.838)
LNS	0.113**	0.149***	0.106**	−0.200***	−0.211***	−0.200***
	(2.128)	(2.629)	(1.998)	(−6.993)	(−7.044)	(−6.993)
K/S	−0.274**	−0.127	−0.291**	0.219***	0.178**	0.221***
	(−2.182)	(−0.961)	(−2.315)	(3.212)	(2.537)	(3.243)
Y/S	1.104***	1.239***	1.147***	0.712***	0.707***	0.714***
	(3.406)	(3.670)	(3.535)	(4.052)	(3.944)	(4.065)
R&D/K	0.001	0.004	0.001	−0.002	−0.002	−0.002
	(0.215)	(0.611)	(0.176)	(−0.421)	(−0.642)	(−0.425)
RDUM	0.004	0.007	0.011	0.806	0.453	0.809
	(0.003)	(0.004)	(0.008)	(1.092)	(0.534)	(1.096)
I/K	0.051	0.045	0.047	−0.007	−0.004	−0.007
	(1.407)	(1.262)	(1.316)	(−0.381)	(−0.187)	(−0.365)
ROA	−0.034***	−0.039***	−0.036***	0.015***	0.012***	0.014***
	(−11.276)	(−12.420)	(−11.840)	(8.840)	(7.261)	(8.719)
LEVERAGE	1.048***	0.927***	1.113***	−3.351***	−3.167***	−3.348***
	(6.040)	(5.051)	(6.402)	(−35.635)	(−32.512)	(−35.565)
Age	−0.760***	−0.915***	−0.870***	0.333***	0.280***	0.356***
	(−12.816)	(−13.921)	(−15.451)	(10.333)	(7.988)	(11.647)
Constant	3.346**	3.465**	4.013***	5.383***	5.930***	5.302***
	(2.266)	(2.025)	(2.719)	(6.731)	(6.533)	(6.638)
Observations	33,664	31,388	33,673	33,664	31,388	33,673
R-squared	0.518	0.527	0.519	0.810	0.819	0.810
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes

Note(s): Table 6 presents regression results of whether the effect of insider ownership on default risk is affected by the degree of risk taking. We use ROA volatility as a variable to capture firm’s risk-taking behavior. ROA volatility is measured as the standard deviation of ROA in previous three years. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denotes statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 7

Channel analysis: Stock price volatility

	(1)	(2)	(3)	(4)	(5)	(6)
Dependent variable	Stock volatility			DD
DIR	0.004***			−0.007***
	(5.284)			(−5.178)
CEOIR		0.002***			−0.005***
		(2.596)			(−3.047)
EMPIR			0.035***			−0.040***
			(11.304)			(−6.671)
Stock volatility				−0.952***	−0.950***	−0.955***
				(−88.491)	(−79.766)	(−88.590)
LNS	−0.274***	−0.294***	−0.273***	−0.463***	−0.511***	−0.463***
	(−20.514)	(−20.457)	(−20.514)	(−17.994)	(−17.853)	(−17.981)
K/S	−0.314***	−0.279***	−0.317***	−0.020	−0.056	−0.025
	(−9.958)	(−8.446)	(−10.078)	(−0.329)	(−0.857)	(−0.414)
Y/S	1.139***	1.153***	1.139***	1.839***	1.767***	1.840***
	(13.579)	(13.518)	(13.602)	(11.408)	(10.443)	(11.412)
R&D/K	−0.002	−0.001	−0.002	−0.007	−0.008*	−0.007
	(−0.731)	(−0.582)	(−0.800)	(−1.388)	(−1.673)	(−1.420)
RDUM	−0.399	−0.285	−0.396	0.296	0.164	0.299
	(−1.280)	(−0.736)	(−1.272)	(0.496)	(0.214)	(0.501)
I/K	0.023**	0.021**	0.022**	0.005	0.005	0.005
	(2.406)	(2.338)	(2.388)	(0.267)	(0.308)	(0.275)
ROA	0.005***	0.006***	0.004***	0.021***	0.019***	0.020***
	(5.954)	(7.278)	(5.395)	(13.679)	(11.753)	(13.334)
LEVERAGE	0.800***	0.620***	0.824***	−2.878***	−2.734***	−2.849***
	(18.248)	(13.352)	(18.802)	(−34.067)	(−29.664)	(−33.670)
Age	−0.114***	−0.014	−0.103***	0.334***	0.355***	0.368***
	(−6.386)	(−0.671)	(−6.041)	(9.785)	(8.458)	(11.279)
Constant	6.126***	5.890***	6.162***	10.933***	11.406***	10.896***
	(17.792)	(14.078)	(17.945)	(16.480)	(13.740)	(16.446)
Observations	34,827	30,636	34,831	34,827	30,636	34,831
R-squared	0.438	0.440	0.460	0.674	0.675	0.679
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes

Note(s): Table 7 presents regression results of whether the effect of insider ownership on default risk is affected by the degree of risk taking. We use stock volatility as a variable to capture firm’s risk-taking behavior. Stock volatility is measured as the standard deviation of daily stock returns in year t. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denotes statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 8

Insider ownership and default risk: the role of institutional ownership

	(1)	(2)	(3)	(4)	(5)	(6)
	High institutional ownership			Low institutional ownership
	DD	DD	DD	DD	DD	DD
DIR	−0.007***			−0.008***
	(−4.708)			(−6.838)
CEOIR		−0.012***			−0.010***
		(−6.200)			(−7.628)
EMPIR			0.004			−0.001
			(0.838)			(−0.239)
LNS	0.082***	0.053***	0.093***	0.130***	0.117***	0.146***
	(7.013)	(4.200)	(8.084)	(10.537)	(8.986)	(12.035)
K/S	0.728***	0.723***	0.742***	0.730***	0.720***	0.744***
	(11.682)	(10.820)	(11.912)	(14.382)	(13.356)	(14.669)
Y/S	6.492***	6.245***	6.486***	3.327***	3.442***	3.275***
	(25.804)	(23.219)	(25.728)	(14.709)	(14.348)	(14.462)
R&D/K	0.024***	0.021***	0.024***	0.004	0.003	0.005
	(3.204)	(2.712)	(3.156)	(0.971)	(0.630)	(1.069)
RDUM	0.344	−0.318	0.387	1.489	1.473	1.591
	(0.471)	(−0.369)	(0.530)	(0.876)	(0.863)	(0.935)
I/K	−0.357***	−0.321***	−0.375***	−0.080***	−0.078***	−0.080***
	(−4.978)	(−4.339)	(−5.215)	(−2.963)	(−2.828)	(−2.951)
ROA	0.026***	0.027***	0.024***	0.033***	0.032***	0.031***
	(9.280)	(9.269)	(8.698)	(13.013)	(12.058)	(12.483)
LEVERAGE	−6.680***	−6.664***	−6.718***	−5.199***	−5.228***	−5.248***
	(−66.608)	(−61.067)	(−67.344)	(−66.300)	(−62.644)	(−67.101)
Age	0.129***	0.108***	0.164***	−0.047**	−0.040*	−0.004
	(6.088)	(4.585)	(8.203)	(−2.457)	(−1.931)	(−0.226)
Constant	3.430***	4.444***	3.119***	1.940	2.062	1.498
	(4.590)	(5.058)	(4.178)	(1.136)	(1.203)	(0.876)
Observations	15,990	13,789	16,024	20,119	17,822	20,121
R-squared	0.605	0.610	0.604	0.508	0.517	0.507
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes
χ2	27.04	22.18	28.26
Prob > χ2	0.000	0.000	0.000

Note(s): Table 8 reports estimates from regression results for analyzing the moderating impact Institutional ownership on the relationship between insider ownership and firm default risk. High institutional ownership and low institutional ownership indicate the below and above median value of institutional ownership level, adjacent to strong and weak external governance. Columns 1–3 show results for firms with high institutional ownership and columns 4–6 show results for firms with less than median institutional ownership. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denotes statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 9

Insider ownership and default risk: role of analyst coverage

	(1)	(2)	(3)	(4)	(5)	(6)
	High analyst coverage			Low analyst coverage
	DD	DD	DD	DD	DD	DD
DIR	−0.004***			−0.009*
	(−8.743)			(−1.663)
CEOIR		−0.007***			−0.012**
		(−10.168)			(−2.476)
EMPIR			−0.006**			−0.019**
			(−1.965)			(−2.290)
LNS	0.156***	0.135***	0.169***	−0.048*	−0.061**	−0.041
	(19.747)	(15.982)	(21.802)	(−1.776)	(−2.163)	(−1.539)
K/S	0.724***	0.724***	0.736***	0.904***	0.819***	0.905***
	(17.369)	(16.239)	(17.651)	(7.894)	(6.882)	(7.930)
Y/S	5.006***	4.968***	4.946***	5.051***	4.994***	4.959***
	(27.357)	(25.386)	(26.990)	(11.591)	(11.003)	(11.357)
R&D/K	0.010**	0.008*	0.010**	−0.003	−0.013	−0.005
	(2.438)	(1.892)	(2.439)	(−0.333)	(−1.228)	(−0.439)
RDUM	0.544	0.313	0.615
	(0.885)	(0.492)	(1.000)
I/K	−0.068**	−0.067**	−0.075***	−0.367***	−0.332***	−0.385***
	(−2.576)	(−2.485)	(−2.831)	(−4.287)	(−3.836)	(−4.498)
ROA	0.035***	0.035***	0.033***	0.007	0.009*	0.006
	(17.040)	(16.075)	(16.212)	(1.590)	(1.934)	(1.368)
LEVERAGE	−5.571***	−5.584***	−5.631***	−7.172***	−7.131***	−7.213***
	(−84.995)	(−79.226)	(−86.282)	(−41.881)	(−39.051)	(−42.501)
Age	0.009	0.002	0.051***	0.179***	0.155***	0.197***
	(0.573)	(0.134)	(3.517)	(4.811)	(3.799)	(5.630)
Constant	2.485***	2.944***	2.121***	5.298***	5.452***	5.208***
	(3.977)	(4.553)	(3.397)	(17.834)	(17.952)	(18.855)
Observations	30,415	26,631	30,445	5,684	4,962	5,690
R-squared	0.541	0.544	0.539	0.614	0.621	0.614
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes
χ2	20.88	23.87	19.88
Prob > χ2	0.000	0.000	0.000

Note(s): Table 9 represents regression results for analyzing the moderating impact of analyst coverage on the relationship between Insider ownership and firm default risk. High analyst coverage and Low analyst coverage indicate the below and above median value of analyst coverage level, adjacent to strong and weak external governance. Columns 1–3 show results for firms with higher analyst coverage and columns 4–6 show results for firms with less than median analyst coverage. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. All regression control for industry and year fixed effects. Superscripts ***, ** and * denotes statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table 10

Insider ownership and default risk: the role of the main bank

	(1)	(2)	(3)	(4)	(5)	(6)
	Firms with main bank			Firms without main bank
	DD	DD	DD	DD	DD	DD
DIR	−0.007***			−0.008***
	(−4.276)			(−6.194)
CEOIR		−0.011***			−0.014***
		(−6.129)			(−6.605)
EMPIR			−0.011			−0.008
			(−0.090)			(−1.484)
LNS	0.100***	0.098***	0.106***	0.206***	0.171***	0.219***
	(9.558)	(9.414)	(10.320)	(18.965)	(13.388)	(20.642)
K/S	0.643***	0.646***	0.630***	0.816***	0.811***	0.837***
	(10.811)	(10.854)	(10.570)	(15.263)	(13.252)	(15.674)
Y/S	6.477***	6.472***	6.401***	4.467***	4.189***	4.409***
	(22.981)	(22.977)	(22.672)	(20.788)	(17.455)	(20.505)
R&D/K	0.097**	0.092**	0.099**	0.011***	0.008*	0.011***
	(2.368)	(2.252)	(2.421)	(2.810)	(1.886)	(2.816)
RDUM				1.186	1.355	1.233
				(1.513)	(1.198)	(1.570)
I/K	−0.333***	−0.319**	−0.382***	−0.058**	−0.062**	−0.062**
	(−2.664)	(−2.554)	(−3.057)	(−2.188)	(−2.240)	(−2.305)
ROA	0.036***	0.036***	0.035***	0.027***	0.027***	0.026***
	(10.858)	(10.920)	(10.569)	(11.875)	(10.542)	(11.248)
LEVERAGE	−5.926***	−5.913***	−5.951***	−5.469***	−5.330***	−5.531***
	(−69.070)	(−68.945)	(−69.539)	(−62.525)	(−51.576)	(−63.610)
Age	−0.059***	−0.063***	−0.034*	0.081***	0.141***	0.117***
	(−2.886)	(−3.174)	(−1.761)	(3.997)	(5.378)	(6.028)
Constant	3.784***	3.822***	3.664***	1.273	1.273	0.953
	(30.917)	(32.351)	(31.648)	(1.599)	(1.117)	(1.199)
Observations	19,358	19,358	19,362	16,754	12,249	16,785
R-squared	0.580	0.580	0.580	0.543	0.559	0.541
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year effects	Yes	Yes	Yes	Yes	Yes	Yes
χ2	29.34	19.28	26.85
Prob > χ2	0.000	0.000	0.000

Note(s): Table 10 reports estimates from regression results to analyze if the presence of main bank system moderates the relationship between insider ownership and firm default risk. Firms with main bank portray strong external governance and firms without main bank portray weak external governance. Columns 1–3 show results for firms with main bank ties and columns 4–6 show results for firms with no main bank ties. T-statistics are in parentheses. Definitions of all variables are provided in Appendix. Superscripts ***, ** and * denote statistical significance at 1, 5 and 10% level, respectively

Source(s): Table created by the authors

Table A1

Definitions of variables

Variable	Definition	Source
Dependent variable
Distance-to-default (Miah and Uddin)	Risk neutral distance-to-default based on Merton (1974)	Credit research initiative
Independent variable
Director ownership	Percentage of shareholdings by Directors	NIKKEI NEEDS CGES database
CEO ownership	Percentage of shareholdings by CEO	Same as above
Employee ownership	Percentage of shareholdings by Employees	Same as above
Control variable
LNS	The natural log of sales, used to measure firm size	DataStream
K/S	The ratio of tangible, long-term assets (property, plant and equipment) to sales is used to measure the mitigation of agency problems as it can be stated that these assets can be easily monitored and provide proper collateral	DataStream
Y/S	The ratio of EBITDA (earnings before interest, tax, depreciation and amortization) to sales	DataStream
R&D/K	The ratio of research and development expenditures to property, plant and equipment. We set missing observations of R&D/K equal to zero to maintain sample size	DataStream
RDUM	A dummy variable is equal to one if R&D data are available, and zero otherwise. This variable allows the intercept term to capture the mean of R&D/K for missing values	DataStream
I/K	The ratio of capital expenditures to property, plant and equipment	DataStream
ROA	Net income before extraordinary items and discontinued operations divided by total assets multiplied by 100	DataStream
Leverage	Long-term debt divided by the book value of total assets	DataStream
Age	One plus the listing age of a firm as measured by the number of years from its IPO as reported in CRSP	DataStream
ROA volatility	The rolling standard deviation of ROA for the year t plus the previous two years	Authors’ calculation
Stock volatility	The rolling standard deviation of stock return for the year t plus the previous two years	Authors’ calculation
Altman Z score	Z score = 1.2A + 1.4B + 3.3C + 0.6D+1.0E A = Working capital/Total Assets; B= Retained Assets/Total Assets; C = Earning before tax and interest/Total Assets; D = Market value of equity/Total liabilities and E = Sales/Total Assets	Authors’ calculation
Institutional ownership	% of shareholding by trust account+ % of shareholding by foreign investor+ % of shareholding by special account
Analyst coverage	The number of analysts following the firms in a given year	Refinitiv Datastream

Source(s): Table created by the authors

Appendix

Table 1

Sample selection, summary statistics and correlation matrix

Panel A: Sample selection
	Observation
Initial observations from NIKKEI CGES database	54,106
Less: Financial companies	11,539
Remaining observations	42,567
Less: Merging data for distance-to-default from CRI database	1,912
Remaining observation	40,655
Less: Missing data from control variables	4,539
Remaining observation	36,116

Panel B: Summary statistics
	N	Mean	Std. Dev	Median	Min	Max
DD	36116	4.464	2.429	3.986	0.185	12.690
DIR	36116	7.535	12.079	1.653	0.000	99.400
CEOIR	31618	5.637	10.529	0.574	0.000	88.520
EMPIR	36116	2.893	3.454	1.800	0.000	39.900
LNS	36116	10.674	1.652	10.538	2.079	16.955
K/S	36116	0.328	0.338	0.256	0.004	2.290
Y/S	36116	0.084	0.078	0.075	−0.251	0.354
R&D/K	36116	0.182	2.438	0.018	−0.091	216.857
RDUM	36116	1.000	0.017	1.000	0.000	1.000
I/K	36116	0.156	0.361	0.108	0.000	39.957
ROA	36116	5.561	6.666	4.672	−20.459	35.202
LEVERAGE	36116	0.190	0.174	0.152	0.000	2.309
Age	36116	2.991	0.945	3.045	0.000	4.234
Note(s): Panel A of Table 1 presents the sample election process and Panel B of Table 1 presents summary statistics of DD, insider ownership proxies and other control variables used in this study. All variables are defined in Appendix and winsorized at the 1–99% levels

Panel C: Pairwise correlations
Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
(1) DD	1.00
(2) DIR	−0.06*	1.00
(3) CEOIR	−0.07*	0.95*	1.00
(4) EMPIR	−0.06*	−0.01	−0.04*	1.00
(5) LNS	0.13*	−0.36*	−0.33*	−0.10*	1.00
(6) K/S	0.05*	−0.11*	−0.11*	−0.10*	−0.05*	1.00
(7) Y/S	0.35*	0.00	0.00	−0.15*	0.14*	0.28*	1.00
(8) R&D/K	0.01	0.02*	0.02*	−0.04*	−0.09*	−0.04*	−0.10*	1.00
(9) RDUM	0.01	0.01*	0.01	0.01	−0.04*	0.00	0.00	0.00	1.00
(10) I/K	−0.01	0.08*	0.09*	−0.10*	−0.07*	−0.11*	0.00	0.14*	0.00	1.00
(11) ROA	0.33*	0.15*	0.13*	−0.08*	0.11*	−0.14*	0.56*	−0.07*	0.01	0.06*	1.00
(12) LEVERAGE	−0.43*	0.04*	0.06*	−0.04*	0.04*	0.28*	−0.13*	−0.04*	−0.01	−0.03*	−0.29*	−0.33*	1.00
(13) Age	0.03*	−0.54*	−0.49*	−0.03*	0.41*	0.20*	0.02*	−0.06*	−0.02*	−0.12*	−0.19*	−0.35*	0.05*	1.00
VIF					1.38	1.44	1.84	1.04	1.00	1.04	1.84	1.22	1.63

Note(s): Panel C of Table 1 presents the pairwise correlations matrix for all variables used in this study. * denotes statistical significance of coefficient estimates at the 5% level. All variables are defined in Appendix

Source(s): Table created by the authors