The Effect of Corporate Governance on the Relationship between Accounting Quality and Trade Credit: Evidence from Japan

Transcription

1

2 The Effect of Corporate Governance on the Relationship between Accounting Quality and Trade Credit: Evidence from Japan Masahiro Enomoto Research Institute for Economics and Business Administration, Kobe University, Japan 2-1 Rokkodai, Nada, Kobe JAPAN; Phone: ; April 2018 Acknowledgments I acknowledge and appreciate the helpful comments from Nobuhiro Asano, Keishi Fujiyama, Shu Inoue, Fumihiko Kimura, Yuya Koga, Kenji Kometani, Yoshitaka Ohashi, Tomoyasu Yamaguchi, Yuto Yoshinaga, and workshop participants at Tohoku University Accounting School (Tohoku University Accounting School Workshop).

3 The Effect of Corporate Governance on the Relationship between Accounting Quality and Trade Credit: Evidence from Japan Abstract The purpose of this paper is to investigate the effect of shareholdings on the relationship between accounting quality and trade credit in Japan. First, this study focuses on well-known features of Japanese corporate governance, which are cross- and stable shareholdings, as a system substituting public information for private information. The relationship between cross- and stable shareholdings, accounting quality, and trade credit is tested. The results show that trade credit of customers without either cross- or stable shareholdings increases with accounting quality and that cross- and stable shareholdings weaken the relationship between accounting quality and trade credit. The findings suggest that a close tie to cross- and stable shareholder substitutes public information for private information and thereby results in reducing the importance of accounting information. Next, this study considers the main bank, which is a key characteristic of Japanese corporate governance. When cross- and stable shareholdings are replaced with bank related variables (cross-shareholdings with the main bank, main bank shareholdings, bank shareholdings, and main bank borrowings), all bank variables except main bank borrowings have a similar effect on the relationship. The findings imply that suppliers delegate banks to monitor customers and accounting information is substituted by private information. Key words; accounting quality; cross-shareholdings; main bank; stable shareholdings; trade credit

4 1. Introduction This paper investigates the effect of corporate governance on the relationship between accounting quality and trade credit in Japan. Trade credit is a major source of short-term financing for firms in Japan as well as other countries. In 2014, trade credit of listed firms in Japan constituted 12.9% of total assets on average and was 1.46 times that of short-term debts. 1 Previous literature shows that high-quality accounting information mitigates information asymmetry and reduces agency costs. In particular, the association of accounting quality with debt and equity financing has been reported in many studies (e.g. Lambert et al. 2007; Francis et al. 2004; 2005; Bharath et al. 2008; Ge and Kim 2010; Hasan et al. 2012). They show that firms with high-quality accounting information gain favorable contract terms from debt and equity providers. In contrast, trade credit financing has not been sufficiently investigated in accounting literature despite its economic importance. Applying the previous studies discussion to trade credit financing, firms with better accounting quality could obtain offerings of trade credit from suppliers. Meanwhile, firms with lower accounting quality could demand more trade credit to access short-term financing due to limited availability of traditional financing. Thus, the prediction of the relationship between accounting quality and trade credit is two-directional. Garcia-Teruel et al. (2014) show the positive relationship between accounting quality and trade credit among the Spanish non-listed firms, whereas Chen et al. (2017) report the negative relationship among U.S. listed firms. Evidence for the relationship is mixed. Japan has the third largest stock market in the world and is characterized as stakeholders corporate governance under code-law. This study considers cross- and stable shareholdings and the main bank, which previous literature has paid attention to, as well-known features of Japanese corporate governance (e.g., Aoki and Patrick 1994; Hoshi and Kashyap 2001; Aoki et al. 2007). Firms have some kinds of transitional relationship with these corporate shareholders such as banks, insurance, suppliers, customers. They reciprocally hold their shares with each other (cross-shareholdings). In addition, many firms have a transitional relationship with the shares of other firms and they form stable shareholders with cross-shareholders (Sheard 1994). 2 Cross- and stable shareholders play a crucial role for Japanese corporate groupings known as keiretsu. Stable shareholders can be viewed as friendly or sympathetic, allowing managers to protect themselves from the external takeover market. Moreover, they comprise principally long-term transaction partners (including financial institutions) (Sheard 1989: 409). The main bank is a bank that has the closest link to the client firms and provides daily and primary financial services. It is usually a long-term lender and/or shareholder of the client firm. Cross- and stable shareholders and main banks play an important role in business economies, complementing capital markets. This 1 Chen et al. (2017) report that trade credit to average total assets is on average 11% among their sample firms in the U.S. 2 This paper posits that stable shareholders consist of cross-shareholders and other stable shareholders. 1

5 type of corporate governance tends to have and mitigate information asymmetry using private information instead of public information (Ball et al. 2000). The Japanese corporate governance, therefore, offers an interesting avenue for this study, which focuses on trade credit financing, the relationship between accounting quality and trade credit, and the effect of a close tie to shareholders on the relationship. Biddle and Hilary (2006) show that as firms with higher accounting quality can access better financing their investment is more efficient. However, in their international study, they do not find such a relationship in Japan. Their results imply the substitution between private and public information. 3 They infer that bank financing and keiretsu as important sources of financing could serve as a private channel to mitigate information asymmetry. Their inference is based on the fact that Japan is a bank oriented country and its primary financing is bank borrowing. Shuto and Kitagawa (2011) and Shuto et al. (2017) document that Japanese ownership structure affects the relationship between the accounting quality and debt contract in previous studies. They predict and report that Japanese ownership structure has a monitoring role and substitute public information for private information. However, little is known about the effect of such structure on trade credit. Given that the private information substitutes public information under the relationship between firms and cross- and stable shareholders, it is predicted that a strong tie to them decreases the effect of accounting quality on trade credit. Garcia-Teruel et al. (2014) and Chen et al. (2017) have not investigated the substitution that is generated from the close tie to shareholders. A regression model is designed following Chen et al. (2017) and cross- and stable shareholdings are incorporated. Three abnormal accruals measures are summarized by principal component analysis as a measure of accounting quality. Cross- and stable shareholders are derived from the NRL (NLI Research Institute) database. Based on 23,460 firm-years from Japanese firms between 2001 and 2014, this study finds that trade credit of firms without stable shareholdings increases with accounting quality. However, stable shareholdings reduce the association of accounting quality with trade credit. When splitting stable shareholdings into cross-shareholdings and other stable shareholdings, cross-shareholdings has a similar effect on the association. In additional tests, cross- and stable shareholdings are replaced with bank related variables (cross-shareholdings with the main bank, main bank shareholdings, bank shareholdings, and main bank borrowings). Bank variables except main bank borrowings have a similar effect on the association. This study contributes to the existing literature in several ways. First, this is the first study indicating that a close tie to shareholders weakens the relationship between accounting quality and 3 In addition to Biddle and Hilary (2006), Ball and Shivakumar (2005), and Beatty et al. (2010) highlight the substitution of accounting information with private information. 2

6 trade credit. This study extends the literature on the substitution between private and accounting information by focusing on the trade credit. Considering the importance of trade credit in financing, this study contributes to a growing body of literature on accounting quality and financing. The additional findings of this study imply that suppliers as a financing provider delegate the monitoring of customers to banks that are closely tied to them. Previous studies on trade credit have not focused on this phenomenon. Moreover, the relationship between accounting quality and trade credit financing among firms without cross- and stable shareholdings complements the previous studies that show the same relationship between accounting quality and debt and equity financing (e.g. Francis et al. 2004; 2005; Bharath et al. 2008). The substitution of private information for public information, which is highlighted by Ball et al. (2000), and Biddle and Hilary (2006), through close ties to stakeholders is not necessarily dominant in Japanese firms. The findings of this study document that the phenomenon is observed in proportion to cross- or stable shareholdings. Next, primary results are consistent with Garcia-Teruel et al. (2014), who used a sample from Spanish private firms, but inconsistent with Chen et al. (2017) using U.S. listed firms. Japan and Spain are characterized as having code-law and bank-oriented financial systems in common, different from the U.S. Although Garcia-Teruel et al. (2014) do not have the interaction term of accounting quality with bank variables, under the bank-oriented financial system, the results of this current study could show similar phenomena. The results suggest that the feature of corporate governance at a country level could affect the role of accounting quality. The remainder of this paper is organized as follows. Section 2 describes the institutional setting in Japan, literature review, and development of hypotheses. Section 3 presents the research design, data, and sample selection procedures used in this study. Section 4 provides the empirical results. Section 5 reports additional and robustness tests and Section 6 concludes this paper. 2. Japanese setting and hypothesis development 2.1. Japanese Setting Japan is a code-law country and characterized by stakeholder corporate governance. A major group such as banks, business association, and labor unions form the agents involved with a firm. In code-law countries, insider communication between managers and stakeholders tend to be a primary resolution system for information asymmetry (Ball et al. 2000). Well-known characteristics of the corporate governance of Japanese firms are cross- and stable shareholdings between firms and banks (e.g. Aoki and Patrick 1994; Hoshi and Kashyap 2001; Aoki et al. 2007). This is a system of interlocking shareholdings between firms. The main bank also plays an important role in Japanese corporate governance as well as reciprocally held cross-shareholdings. Japanese corporate groupings, keiretsu typically consists of the extensive 3

7 cross-shareholdings among the member firms and the main bank. Cross- and stable shareholdings is a mechanism for protecting firms from hostile takeover and pressure from the capital market (Sheard 1989; 1991; Osano 1996). Member firms of industrial groups actively transact with each other on a regular basis in the long horizon and strongly rely on trade credit financing within the group (Berglof and Perotti 1994). In addition, through cross- and stable shareholding mechanism, managers and the shareholders share firms non-financial information. Specifically, the information sharing comes through interlocking directorates and presidential club (Shacho-kai) meetings that are held regularly to exchange inside information in affiliated firms (Sheard 1991; Goto 1992; Cooke 1997; Douthett and Jung 2001). 4 Through this mechanism, cross-shareholders can gain access to strategic information such as the performance and business plans of associated firms. Thus, for Japanese firms, cross-ownership of stock and access to information is particularly prevalent among firms in the same industrial group (keiretsu) (Jacobson and Aaker 1993, 403). 5 The mechanism also provides the effective monitoring of managerial performance such as operating and investment (Kaplan and Minton 1994; Kang and Shivdasani 1997). Okabe (2002, 50-51) states that cross-shareholdings are a form of implicit contract that mutually enables the enjoyment of a reduction of transaction costs as a result of a long-term transactional relationship. Several studies report that the influence of cross- and stable shareholdings has weakened since around the middle of 1990s (e.g. Miyajima and Kuroki 2007). However, in the study sample, the ratio of cross- and stable shareholdings stabilized after around The mean value of cross-shareholdings (stable shareholdings) is 11.9% (30.0%) in 2001, 9.5% (22.7%) in 2007, and 10.4% (21.3%) in Miyajima et al. (2015) document the changes in the ownership structure of corporate governance in Japan and report that the unwinding of cross-shareholdings is driven by bank shareholders, not by industrial firms. Rather cross-shareholdings between industrial firms have been strengthened for the business alliance after Several studies present evidence that crossand stable shareholders still influence accounting numbers and debt contracts in Japan (Shuto and Kitagawa 2011; Nagata and Nguyen 2017). This information can be seen as justification for the focus of this study. The main bank holds a prominent position for providing financing to the client firm. In Japan, banks are permitted to hold shares in non-financial firms up to a specified ratio, as in European countries. 6 The main bank is at the core of the cross-shareholder group and in many cases 4 Goto (1992) reports the coordination of R&D within member firms. He also states that member firms undertake various types of interfirm coordination, but that outsiders to this group does not have accessing information. 5 Further, Jiang and Kim (2001) report that cross-shareholdings are an influential factor for the mitigation of information asymmetry in Japanese equity markets. 6 In Japan, banks are permitted to hold the shares of non-financial firms. Based on previous studies, Ono et al. (2017, 1) explain two reasons why banks, both the main bank and non-main bank, hold the equity claim of borrowers. They summarize that the motivation of holding borrower s shares is to (1) obtain a competitive advantage from 4

8 has close ties to the borrowers firm. Douthett and Jung (2001, 597) describe how Japanese firms relate to each other under the mechanism that the main bank is the principal lender to the group members, it owns a significant number of shares in group firms, and often participates in their management. The main bank within the group can monitor the borrowers daily transactions through its account and dispatch directors to lender firms with financial difficulties (Sheard 1994). Lender information is accumulated by the main bank in the form of long-term transactions. In addition, other banks form long-tern relationships with cross- and/or stable shareholders by holding borrowers shares with the main bank. Moreover, previous research documents that the main bank in the relationship influences management turnover of Japanese firms. Kang and Shivdasani (1995, 1997) find that firms in financial difficulties increase the possibility of management turnover when under the main bank. Miyajima et al. (2018) document that, even after 2006, the main bank still has an influence on management turnover and disciplines managers through monitoring. Although it is said that the influence of banks has weakened since around 2000, Hirota (2009) reports that the main bank borrowing to total debt and shareholder s equity has not changed substantially since In summary, cross- and stable shareholders based on the main bank system still play an important role in business economies in Japan. The mitigation of information asymmetry by private information through close ties to stakeholder offers an interesting setting for this research by focusing on an effect of such ties on the association of accounting quality with trade credit Hypothesis development There are two main avenues for explaining why firms use trade credit instead of cash payment, which are non-financial and financial theory. 7 Some financial theory based studies highlight that suppliers are able to obtain private information through business relationships. For example, suppliers may often meet with buyers more regularly than the bank to exchange information regarding business plan, industry and demand trends, and R&D, as well as financial information. Moreover, suppliers are able to know the size and timing of customers orders through daily operating activities (Petersen and Rajan 1997, Miwa and Ramseyer 2008; Uesugi et al. 2009). Thus, suppliers have an advantage in private information compared to financial institutions and are complementary effect between shareholdings and lending activity and (2) mitigate the conflict of interest between shareholder and borrowers. 7 From non-financial theory, trade credit allows saving of transaction cost by reducing the frequency of settlement (Ferris 1981). Taking a longer time for payment procedures (e.g. verification of goods) increases trade credit. In addition, price discrimination (Brennan et al. 1988) and quality guarantees (Emery and Nayar 1998) are discussed in previous literature. From financial theory, monitoring of suppliers has an advantage over financial institutions because suppliers have daily operating transactions with customers. Suppliers also have a comparative advantage in collateral liquidation (Frank and Maksimovic 2005; Longhofer and Santos 2003). Buyer s moral hazard is reduced through strong supplier-customer relationships using trade credit (Cuñat 2007). Furthermore, trade credit is beneficial for both suppliers and customers in the long-term relationship (Cheng and Pike 2003). 5

9 likely to offer the amount of trade credit to customers based on private information as well as public information. 8 The exchange of private information could make the relationship between accounting quality and trade credit ambiguous. 9 Previous studies on the effect of accounting quality on debt and equity financing indicate that high accounting quality lowers information risk and thereby reduces the cost of debt and capital (e.g. Francis et al. 2004; 2005; Bharath et al. 2008). Based on this relationship, suppliers could be willing to offer trade credit to firms with better accounting quality. Pike and Cheng (2001) report that according to their questionnaire survey, customers financial statements are one of the most popular information sources of risk screening for trade credit supply. Garcia-Teruel et al. (2014, 1196) state that customers with a higher credit rating should be offered more trade credit from suppliers, given that these face a lower moral hazard problem. Raman and Shahrur (2008) show that restraint in earnings management, which is an important factor of accounting quality, encourages supplier-customer relationships. Meanwhile, when suppliers estimate the risk of customers with low accounting quality to be higher, suppliers may offer lower trade credit to the customers. Accounting quality, thus, is crucial information for offering trade credit. In line with this argument, using a sample of 8,396 firm-years from Spanish private firms between 1995 and 2005, Garcia-Teruel et al. (2014) report that firms with higher accruals quality tend to obtain more trade credit from suppliers. Li et al. (2018) assume that mandatory IFRS adoption improves financial reporting transparency and find that the IFRS adoption increases trade credit. 10 The results are consistent with evidence from previous research that financing providers are willing to offer more financing to firms with higher accounting quality. 11 Based on the above discussion, trade credit increases with accounting quality. Typically, trade credit has higher interest rates than short-term debt. Higher interest rates lead to another prediction regarding the relationship between accounting quality and trade credit. Firms with higher accounting quality can finance short-term debt from financial institutions more easily and less costly. In line with previous research, customers with low accounting quality appear to have difficulty accessing debt and equity financing. Such customers may need to increase trade credit due to their financial constraints and if so, trade credit is negatively related to accounting quality. Based on U.S. sample of 115,703 firm-years from 1985 to 2011, Chen et al. (2017) present evidence that accounting quality is negatively associated with trade credit. 8 In addition, Ng et al. (1999) show that credit analysis for offering trade credit to customers is under economies of scale that increase with the number of customers, decreasing cost. Miwa and Ramseyer (2008) claim that suppliers are familiar with short term lending through daily activities. 9 This exchange of private information is likely to occur under the relationship between suppliers and customers withor without cross- or stable shareholdings. 10 Financial reporting transparency is a concept that closely relates to accounting quality used in this study. Li et al. (2018) do not directly show the relationship between the improvement of financial reporting transparency and the increase in trade credit. 11 In addition to accrual quality, they show similar results when using lower earnings variability, higher income smoothing and earnings predictability. 6

10 Accounting quality has a positive and negative effect on the amount of trade credit. In addition, private information collected from the supplier-customer transactional relationship could weaken the effect. However, financial information is used to facilitate transactions between suppliers and customers. In particular, it is a major source of credit rating that suppliers usually rely on for offering trade credit. Pike and Chen (2001) report that credit rating is the most popular source of information for credit risk. Thus, this study predicts that the positive effect is likely to dominate the negative effect. 12 Hypothesis 1: Trade credit increases with accounting quality. Next, the effect of cross- and stable shareholders on the relationship between accounting quality and trade credit is discussed to develop Hypothesis 2. Several studies present the evidence that cross- and stable shareholders have a significant effect on accounting figures and debt contracts in Japan. Shuto and Iwasaki (2014) report that firms with higher stable shareholdings increase the informational component of income smoothing. Shuto and Kitagawa (2011) investigate the monitoring role of cross- and stable shareholders. They document that stable shareholdings by financial institutions lower the interest rate of the bond. Nagata and Nguyen (2017) focus on the relationship between Japanese ownership structure and disclosure quality (management forecast revision) in Japan. In particular, they find that bank shareholdings lead to worse disclosure quality due to the accessibility of private information. Ball and Shivakmar (2005) point out that private information substitutes accounting quality by comparing the conservatism of listed firms with that of non-listed firms. Beatty et al. (2010) also suggest that when outside fund suppliers can access private information, accounting quality plays a less important role in investment efficiency. This result implies the substitution between private information and accounting information with respect to debt. With regard to corporate governance, Ge et al. (2012) report that higher quality corporate governance leads to less strict contracting terms. Robin et al. (2016) provide evidence consistent with the substitution between high-quality audit and debt covenant. This study assumes that the close tie to cross- and stable shareholders is inclined to encourage managers to use private instead of public information. Cross- and stable shareholders form a business group, including customers, suppliers and the main and non-main banks. Managers within the business group monitor each other through trade credit and cross-shareholdings (Berglof 12 Based on a questionnaire survey of Japanese private firms, Uesugi et al. (2009) find that only 6.3% of trade credit contracts include early payment discounts. Early payment discounts tend to be offered to risky customers (Klapper et al. 2009). The percentage (6.3%) in Uesugi et al. (2009) is much less than that of U.S. firms in Ng et al. (1999) (24.5%), and in Giannetti et al. (2008) (21.3%). This could lead to suppliers in Japan focusing on accounting quality if offering trade credit without discount. 7

11 and Perotti 1994). The increases in cross- and stable shareholdings means stronger tie to suppliers as shareholders and could lead to more active exchange of private information within a business group. If customers were able to share their private information as well as public information through shareholder networks, including the main bank, they would have less incentive to improve earnings quality in order to access trade credit financing. Suppliers without cross-shareholdings may delegate the monitoring of customers to cross-shareholders including banks. In Japan, as previously mentioned, banks usually hold borrower s shares for long periods of time. The main bank of the cross-shareholders typically represents the relationship. In addition, other banks as well as the main bank hold the borrowers shares in order to have a close tie to them. When customers have a close and stable tie to banks, the banks tend to supply funds even when customers are in financial difficulties (Sheard 1991). Therefore, suppliers have less incentive to grant trade credit according to accounting quality of customers with a close tie to banks. Suppliers can utilize not only financial statement information but also the monitoring of the main bank with private information. In other words, suppliers can delegate monitoring to banks. In summary, given the existence of cross- and stable shareholdings under the main bank system, the firms private information exchanges between managers and them. A dependency on accounting information, therefore, could reduce with the strength of the relationship. According to the aforementioned discussions, the firms that have a close relationship with cross- and stable shareholdings are likely to reduce the relationship between accounting quality and trade credit. Hypothesis 2: The increase of cross- and stable shareholdings reduces the relationship between accounting quality and trade credit. As described in the development of Hypothesis 1, supplier-customer relationships inherently make the effect of accounting quality on trade credit ambiguous even without cross- and stable shareholdings. Thus, in the next section,this paper provides the research designs to focus on the influence of cross- and stable shareholdings. 3. Research design, sample selection procedure, and data 3.1. Research design Hypothesis 1 is tested by running the regression model of equation (1). TradeCreditit = β 0 + β 1AQ it + β 2LiquidCost it + β 3InfoAsym it + β 4 Log(Asset) it + β 5Log(Age + 1) it + β 6MktShare it + β 7POS_ChgSale it + β 8NEG_ChgSale it + β 9ROA it + β 10MTΒ it + β 11AltmanZ it + β 12Leverage it + β 13CA it + β 14CL_XTrade it + β 15CashHold it + β INDUSTRY + β YEAR + ε (1) 8

12 Chen et al. s (2017) model is applied as closely as possible to ensure comparability. The independent variable, TradeCredit, is the ratio of trade credit to total assets. AQ, test variable, is accounting quality measure. AQ is calculated using principal component analysis in line with Bharath et al. (2008) and Chen et al. (2017). Three types of abnormal accruals measures (AA1, AA2, and AA3) are used for principal component analysis. For AA1, the absolute value of the residual of Dechow and Dichev s (2002) accrual model is standardized after multiplying by -1. AA2 is calculated based on Teoh et al. s (1998) model and AA3 is computed based on Dechow et al. s (1995) model in a similar vein. All three accrual models are estimated by industry-year, where industries are identified by the Tokyo Stock Exchange classification codes. 13 AQ is computed as the first principal component score of AA1, AA2, and AA3. Accounting accruals are computed by the balance sheet and income statement because the calculation of accruals is modified to mitigate the influences of change in accounts payable according to Chen et al. (2017). 14 Change in accounts payable is removed from the calculation of accruals. Change in accounts payable is associated with the change in inventories. Hence, the changes in inventory are regressed on the changes in accounts payable and the residuals of the regression are used as the modified change in inventories. 15 Control variables in equation (1) are similar to Chen et al. (2017) and the definitions of them are in the Appendix. 16 Positive β 1 indicates that accounting quality promotes the offering of trade credit from suppliers and this supports Hypothesis 1. To test Hypothesis 2, equation (2) includes corporate governance variables, StableShareholdings, CrossShareholdings, and OtherStableShareholdings. First, TradeCredit is regressed on AQ and its interaction with StableShareholdings in equation (2). As stable shareholdings can be split into cross-shareholdings and other stable shareholdings (StableShareholdings = CrossShareholdings + OtherStableShareholdings), StableShareholdings is replaced with CrossShareholdings and OtherStableShareholdings. Other stable shareholdings consist 13 Sample firms are divided into 33 industries according to Tokyo Stock Exchange classification codes. At least 20 observations in each industry-year group are required to calculated abnormal accruals. As a result, the primary test uses 25 industries for the estimation of equation (1). 14 Firm-years in which total assets and/or sales growth exceeds 100% are excluded to mitigate the effect of outliers from major changes to business fundamentals such as large M&As (Hribar and Collins 2001; Almeida and Campello 2007). 15 Accruals is defined as (Δcurrent asset - Δcash - Δtrading securities - Δshort-term loans receivable) - (Δcurrent liability - Δshort-term loan payable - Δnote payable for PPE - Δaccrued amount payable for PPE) - Δlong-term allowance - depreciation); PPE = the amount of property, plant, and equipment: all of the items are divided by total assets at the end of year t-1. To modify change in accounts payable, the change in accounts payable and the change in inventories are excluded and the residuals of regression of the change in inventories on the changes in accounts payable are added. For Dechow and Dichev s (2002) and Teoh et al. s (1998) models, working capital accruals are used after eliminating non-working capital items. 16 One exception from Chen et al s (2017) model is Altman Z. They use as S&P credit rating (PredRating) instead of Altman Z. The database (Nikkei NEEDS-FinancialQuest) does not include credit rating in it. 9

13 of financial institutions (property account) and parent company. 17 This variable, therefore, is regarded as stable shareholdings by financial institutions for firms without a parent company, because many banks, including non-main banks, hold the shares of listed firms for the purposes of lending and the strength of the relationship. TradeCredit it = β 0 + β 1AQ it + β 2AQ StableShareholdings it + β 3 StableShareholdings it + β 4LiquidCost it + β 5InfoAsym it + β 6Log(Age + 1) it + β 7MktShare it + β 8POS_ChgSale it + β 9NEG_ChgSale it + β 10ROA it + β 11MTΒ it + β 12AltmanZ it + β 13Leverage it + β 14CA it + β 15CL_XTrade it + β 16CashHold it + β INDUSTRY + β YEAR + ε (2) TradeCredit it = β 0 +β 1 AQ it + β 2 AQ CrossShareholdings it + β 3AQ OtherStableShareholdings it + β 4 CrossShareholdings it + β 5 OtherStableShareholdings it + β 6LiquidCost it + β 7InfoAsym it + β 8Log(Age + 1) it + β 9MktShare it + β 10POS_ChgSale it + β 11NEG_ChgSale it + β 12ROA it + β 13MTΒ it + β 14AltmanZ it + β 15Leverage it + β 16CA it + β 17CL_XTrade it + β 18CashHold it + β INDUSTRY + β YEAR + ε (3) StableShareholdings = the ratio of shares held by cross-shareholders and other stable shareholders; CrossShareholdings = the ratio of shares held by cross-shareholders; and OtherStableShareholdings = the ratio of shares held by other stable shareholders. Equation (2) allows the coefficients to differ between firms with and without stable shareholders, the differences in the effect of accounting quality between the two is tested. As previously described, stable shareholdings are predicted to weaken the relationship between accounting quality and trade credit. Then, a positive coefficient of AQ (β 1) suggests that firms with high AQ and without stable shareholdings increase the amount of trade credit and a negative coefficient of AQ it CrossShareholdings (β 2) means that close ties to stakeholders cancels out the effect of AQ. 18 Similarly, this study focuses on the coefficients of the two interaction terms, β 3 and β 4 in equation (3). This study also predicts that the sign of β 1 in equation (3) is positive and that β 3 and β 3 is negative in opposite to β This definition is used by the NRL database. Shuto and Kitagawa (2011) and Shuto and Iwasaki (2014) employ the same definition. They also deal with this variable as a proxy for stable shareholdings by financial institutions. 18 β1 in the equation (2) shows the mixed results of (a) the positive effects of accounting quality on the amount of trade credit and (b) the negative effect of exchange of private information between suppliers and customers (without stable shareholdings) on (a). Positive β1 shows that the former effect dominates the latter one. β3 represents the direct effect of stable shareholdings on trade credit. Because the close tie to stable shareholdings could increase the credibility of a firm, the expected sign of β3 is positive. 19 The expected signs of β4 and β5 are positive similar to β3 in equation (2). 10

14 3.2. Sample selection procedure and data The initial sample consists of non-financial firms with consolidated financial statement data from 2001 to Firm-years in which total assets and/or sales growth exceeds 100% are excluded to avoid the effect of major changes to business fundamentals such as large M&As. Firm-years that do not have the data to calculate accounting accrual are also excluded. Lastly, firm-years without data on cross- and stable shareholdings or without the data to calculate control variables are excluded. The final sample yields 23,630 firm-year observations. The sample selection procedures are summarized in Table 1. Data are obtained from three databases. All of the data on financial statements is obtained from Nikkei NEEDS-FinancialQuest (Nikkei Media Marketing). The data on cross- and stable shareholdings are derived from the Data Package of Cross-Shareholding and Stable Shareholding (NLI Research Institute). Stock price data is obtained from NPM (Financial Data Solutions). 4. Results Table 2 reports descriptive statistics on dependent and all independent variables used in the analysis of the relationship between accounting quality and trade credit. The mean value of TradeCredit is (the median is 0.126). Cross-shareholders hold approximately 10% of shares on average and stable shareholders hold over 20%, implying that cross- and stable shareholders still have a substantially influence on corporate governance in Japan. Table 3 shows the correlation matrix. Since no high correlation coefficient is observed within independent variables, the results of the regressions in this section will not be influenced by multicollinearity. 20 Table 4 presents the regression results of equation (3). 21 In column (1) the coefficient of AQ is not significant, not consistent with Hypothesis 1. Columns (2) and (3) test Hypothesis 2 with the interaction effect of AQ with StableShareholdings, CrossShareholdings, and OtherStableShareholdings. Column (2) reports that the coefficient of AQ is significantly positive, indicating the positive effect of accounting quality on trade credit when stable shareholdings is zero. 22 The results can be interpreted to mean that the higher accounting quality of firms without stable shareholders increases the amount of trade credit. Hypothesis 1 is conditionally supported if 20 The VIFs (variance inflation factor) for the primary tests are below five, the multicollinearity does not appear to affect the results. 21 All continuous variables at the top and bottom 1% are winsorized to mitigate the influence of outliers. 22 In Japan and Spain, the majority of trade credit contracts do not have cash discounts contracts in common (Uesugi et al. 2009; García-Teruel et al. 2014). In the U.S., Ng et al. (1999, 1110) report that "2/10 net 30" contracts are frequently observed in their sample. This contract means the combination of 2 percent discount for payment within 10 days and a net period of 30 days. The implicit interest rate is 43.9 percent. The results of this study imply that Japanese and Spanish suppliers pay attention to accounting quality when offering trade credit without discount. Meanwhile U.S. firms with little access to traditional financing due to low accounting quality would seek to obtain trade credit in spite of the high-interest rate. 11

15 there are zero stable shareholdings. The results of this study are in contrast to those of Chen et al. (2017). The coefficient of interaction term, AQ StableShareholdings has a significantly negative sign. The finding supports Hypothesis 2 and show that the increase in stable shareholdings decreases the importance of accounting quality for supply and/or demand of trade credits. In other words, a close tie to stakeholders, including suppliers, could encourage the exchange of private information instead of public disclosure. With respect to control variables, the almost all of the control variables are significant except CashHold. In column (3), StableShareholdings is replaced with CrossShareholdings and OtherStableShareholdings. The coefficients of the interaction term of AQ with CrossShareholdings and OtherStableShareholdings are significantly negative, indicating that a close tie to cross-shareholdings and other financial shareholders reduces the association of accounting quality with trade credit. 23 The coefficients of StableShareholdings, CrossShareholdings, and OtherStableShareholdings are significantly positive in columns (2) and (3), indicating that a close tie to stable shareholders increases trade credit. In other words, when the customers are in the same business group as suppliers, they can receive more trade credit from suppliers. 24 The aforementioned results can be interpreted another way. Given that either accounting quality or stable shareholdings increase the amount of trade credit, the effects of accounting quality and stable shareholdings cancel each other out because of the negative sign of their interaction term (AQ StableShareholdings). The reason for such cancelling out would be the substitution between public and private information. 5. Additional tests and robustness checks 5.1. Additional tests Effect of bank shareholdings As the first additional test, this study analyzes bank shareholdings as one of the features of Japanese-style shareholdings as well as cross- and stable shareholdings. Specifically, bank shareholdings variables are employed to extract the effect of the bank from StableShareholdings and CrossShareholdings. Shareholdings of borrowers reinforce a close tie to each other. Such a tie would encourage the substitution of public information for private information. Shuto et al. (2017) report the substitution between public and private information and the monitoring effect under the main bank. 23 The coefficient of AQ CrossShareholdings is significantly negative when CrossShareholdings- and OtherStableShareholdings-related variables are separately incorporated. 24 The results are consistent with Hoshi et al. (1990), which state that financial links between Japanese firms could encourage suppliers to extend trade credit to customers and customers to buy from suppliers within the industrial group. 12

16 In line with the primary test, this study focuses on bank shareholdings to observe the effect of the main bank (and non-main bank) on trade credit by including four bank related variables as replacement for stable shareholdings in equation (2). The variables are the ratio of cross-shareholdings with the main bank (MainBankCrossShareholdings), the ratio of main bank shareholdings (MainBankShareholdings), the ratio of bank shareholdings (BankShareholdings), and the ratio of borrowing from the main bank to total debt (MainBankBorrowing). Each firm's main bank is defined as the top-ranked bank on the list of Nikkei Primary Trading Banks (Nikkei Shuyo-Torihiki-Ginko), which is compiled via questionnaire survey distributed to all listed firms regarding the trading banks of each firm. 25 The former two variables represent the shareholding relationship with the main bank. Cross-shareholdings with the main bank show the extent of strong capital tie to each other. MainBankShareholdings represents the influence of the main bank as a shareholder. BankShareholdings is a proxy for the strength of all of bank shareholdings, comprising the main bank and non-main banks. Higher those values mean a stronger tie to banks and thus increase the potential use of private instead of public information. In addition to capital ties, main bank dependency is added as a borrowing tie (MainBankBorrowing) to the main bank. The measure also represents another type of main bank tie. The following regression is run to test the bank relationship: TradeCredit it = β 0 + β 1AQ it + β 2AQ Bank it + β 3 Bank it + β 4LiquidCost it + β 5InfoAsym it + β 6 Log(Asset) it + β 5Log(Age + 1) it + β 8MktShare it + β 9POS_ChgSale it + β 10NEG_ChgSale it + β 11ROA it + β 12MTΒ it + β 13AltmanZ it + β 14Leverage it + β 15CA it + β 16CL_XTrade it + β 17CashHold it + β INDUSTRY + β YEAR + ε (4) Bank = MainBankCrossShareholdings, MainBankShareholdings, BankShareholdings, or MainBankBorrowing. MainBankCrossShareholdings = the ratio of shares mutually held with the main banks, BankShareholdings = the ratio of shares held by banks, MainbankShareholdings = the ratio of shares held by the main bank, and MainBankBorrowings = the ratio of borrowings from the main bank to total borrowings. β 2 is predicted to have an opposite sign of β 1, as in the primary tests. Table 5 shows the results from equation (2), which tests an effect of banks on the relationship between accounting quality and trade credit. All coefficients of the interaction terms of 25 In this questionnaire survey, each firm ranks the banks from which it receives primary financial services, irrespective of shareholdings or loan shares. This study posits that the bank at the top of the list serves as the main bank. Similar procedures are used in other research on the main bank. 13

17 accounting quality with bank shareholdings variables are significantly negative. The results are consistent with those of cross- and stable shareholdings. In other words, even when extracting main bank shareholdings from cross- and stable shareholdings, the results remain unchanged. The tie to banks (main bank) also appears to lower the importance of accounting information for trade credit. The coefficient of interaction term of AQ with MainBankBorrowings, however, is not significant. Although the strengthening of the main bank tie to shareholdings increases trade credit, the tie by debt dependency does not have an effect on trade credit. This result can be interpreted as follows. The increase of bank borrowing strengthens a tie to banks. On one hand, the increase could reduce the importance of accounting figures according to strengthening the tie through bank borrowings. On the other hand, the increase could make banks pay more attention to accounting information. Thus, the mixed effect could make the results ambiguous. All of the coefficients of MainBankCrossShareholdings, BankShareholdings, and MainBankShareholdings are not significant, but the coefficient of MainBankBorrowings is significantly positive. The results mean that trade credit does not increase with the shareholding of banks Effect of financial constraint As second additional tests, this study focuses on the influence of financial constraints on the relationship between accounting quality and trade credit. When the internal funds of customers are less available, customers should increase the demand for receiving trade credit from suppliers (Petersen and Rajan 1997). The results shown in Table 4 would depend on the customer s financial constraints. Therefore, it is assumed that customers with high level of financial constraints have the incentive to mitigate the constraint through the increase of trade credit and suppliers consider such constraint before offering trade credit. In other words, firms with low financial constraint should improve accounting quality in order to obtain trade credit and suppliers should monitor accounting quality. The results of Table 4 are predicted to remain unchanged under firms with high level of financial constraints. To test this prediction, financial constraint measures, KZ index (Kaplan and Zingales, 1997) are calculated and the sample is split into the two sub-samples: the financial constraint is high and low. 27 Table 6 presents the estimation results of the coefficients for high financial constraint firms in columns (1), (2), and (3) and low high financial constraint firms in columns (4), (5), and (6). The coefficient of AQ is significantly positive and the interaction term of AQ with StableShareholdings 26 However, when bank variables are replaced with their dummies, which are 1 if the variables are not zero, the dummies of BankShareholdings, MainbankShareholding, and MainBankBorrowings are significantly positive. The tie to banks through shareholdings appears to increase trade credit. 27 When using WW index (Whited and Wu, 2006), the results are similar to those of the KZ index. 14

18 among financially constrained firms is significantly negative. 28 However, among financially unconstrained firms, the coefficients of AQ are also significant positive, but the values are smaller than those of the constrained group. Moreover, none of the coefficients of the interaction terms of accounting quality with cross- and stable shareholdings variables are significant. The observed findings suggest that, with respect to trade credit, accounting quality among firms with low financial constraints is less important than that of firms with high financial constraints. It can be said that firms with high financial constraint and without stable shareholdings could increase trade credit to relax the constraint through the improvement of accounting quality Effect of market share Next, market share is considered as a proxy for buying power. When customers have strong bargaining power based on buying power, they may receive more trade credit from suppliers without using accounting information. The entire sample is then divided into two groups, where the firm-years in which MktShare values are above (equal or less than) the median in the industry-year group are classified into high (low) market share groups. Table 7 provides the estimation results of the coefficients for firms with high market share in columns (1), (2), and (3) and firms with low market share in columns (4), (5), and (6). All of the coefficients of AQ and AQ with StableShareholdings and CrossShareholdings are significant in columns (5) and (6). The results are similar to those of the primary results in Table 4. In contrast, firms with low market share do not exhibit statistically significant coefficients of AQ nor the interaction term of AQ with StableShareholdings in column (2), which differs from the results of Table 4. Only the coefficient of AQ is marginally significant in column (3), but the value is smaller than in column (6). These findings suggest that for firms with the low market share (low buying power), accounting quality and cross- and stable shareholdings still play an important role for trade credit, whereas for firms with high market share, they are less important for trade credit Robustness checks Three tests are carried out for robustness checks. First, two alternative accrual models are employed: Jones s (1991) and Kothari et al. s (2005) models. When using both model, the results remain unchanged. In addition, three accounting quality measures are used separately before summarizing by principal component analysis (non-tabulated). For AA1 and AA2, the coefficients of AQ and the interaction term of AQ with StableShareholders and CrossShareholders are consistent with Hypotheses 1a and 2. When using AA3, only the coefficient of the interaction term of AQ with 28 Strictly, the extent of suppliers financial constraint may be incorporated in these tests, but their data are not included in Nikkei NEEDS-FinancialQuest. 15