BANKING PRACTICES AND BORROWING FIRMS FINANCIAL REPORTING: EVIDENCE FROM BANK CROSS-SELLING

Transcription

1 BANKING PRACTICES AND BORROWING FIRMS FINANCIAL REPORTING: EVIDENCE FROM BANK CROSS-SELLING by Minhui Su A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Joseph L. Rotman School of Management University of Toronto Copyright by Minhui Su (2017)

2 Banking Practices and Borrowing Firms Financial Reporting: Evidence from Bank Cross-Selling Minhui Su Doctor of Philosophy Joseph L. Rotman School of Management University of Toronto 2017 Abstract This dissertation studies whether banking practices affect borrowing firms financial reporting. Specifically, I examine the effect of bank cross-selling activities (i.e., a bank s joint provisions of lending and underwriting services to the same firm) on borrowers financial reporting quality for debt contracting. Cross-selling enhances lenders monitoring incentives because banks face additional reputational risks associated with underwriting services. In addition, cross-selling improves lenders monitoring abilities because information sharing between the underwriting and the lending divisions enables them to better discipline borrowers financial reporting for debt contracting purposes. Consistent with these arguments, I show that cross-selling is associated with an improvement in the debt contracting value (DCV) of accounting information at borrowing firms. I also find that banks use more accounting-based monitoring mechanisms for loans cross-sold with underwriting services than for other comparable loans. These findings are robust to alternative identification strategies. ii

3 Acknowledgments I would like to express my deepest thanks to my dissertation co-chairs, Scott Liao and Gord Richardson, for their close guidance and support. I thank Gord for being the most caring and encouraging advisor I could ever imagine. I am deeply indebted to Scott for his mentorship since my first year in the program and his investment in my academic career. I also owe my sincere gratitude to my committee members Dushyant Vyas and Baohua Xin for their invaluable time and intellectual input. I owe special thanks to Jeff Callen, Alex Edwards, Ole- Kristian Hope, Daehyun Kim, Ping Zhang, and other faculty members at Rotman for their encouragement, advice, and support toward completion of my doctoral degree. I also thank my fellow students and the staff at Rotman, who have made my past five years enjoyable. I am thankful to my parents, who have been always there for me. Finally, I am forever grateful to my friends in Toronto for their love and encouragement at all times. This accomplishment would have been impossible without them. iii

4 Table of Contents Acknowledgments... iii Table of Contents... iv List of Tables...v List of Figures... vi List of Appendices... vi 1. Introduction Institutional Background and Literature Review Evolution of Regulations on the Scope of Bank Activities Research on Combining Lending and Underwriting Research on Bank-Firm Relationships Research on the Role of Accounting Information in Debt Contracting Hypothesis Development Research Design Data and Sample Selection Model Specifications Empirical Results Descriptive Statistics Main Findings Additional Analyses Additional Identification Strategies Evidence from Public Debt Contracting Details of Financial Covenants in Cross-Sold Loans Cross-selling When Information Firewall Existed Robustness Checks Conclusion...33 References...35 iv

5 List of Tables Table 1 The Distribution of Cross-Sold Loans by Year...51 Table 2 Descriptive Statistics...52 Table 3 The Impact of Cross-Selling on Debt Contracting...53 Table 4 The Impact of Banks Cross-Selling on Firm Financial Reporting...55 Table 5 Cross-Sectional Variations in the Impact of Cross-Selling on Debt Contracting and Firm Financial Reporting...57 Panel A: Cross-Sectional Variations in the Impact of Cross-Selling on Debt Contracting...57 Panel B: Cross-Sectional Variations in the Impact of Cross-Selling on Firm Financial Reporting...59 Table 6 Additional Identification Using Refined Samples...61 Panel A: The Impact of Cross-Selling on Debt Contracting...61 Panel B: The Impact of Cross-Selling on Firm Financial Reporting...63 Table 7 Additional Identification Using a Difference-in-Differences Design...65 Panel A: The Impact of Cross-Selling on Debt Contracting...65 Panel B: The Impact of Cross-Selling on Firm Financial Reporting...67 Table 8 Corroborating Evidence from Public Debt Contracting...69 Table 9 Details of Financial Covenants in Cross-Sold Loans...71 Table 10 Robustness Check Using the Sample Prior to the Recent Financial Crisis...73 Panel A: The Impact of Cross-Selling on Debt Contracting the Sample Prior to the Recent Financial Crisis...73 Panel B: The Impact of Cross-Selling on Firm Financial Reporting Using the Sample Prior to the Recent Financial Crisis...75 Table 11 Robustness Check Controlling for General Covenants...77 Table 12 An Alternative Measure of the DCV of Accounting Information...79 v

6 List of Figures Figure 1 Illustration of the Treatment and Control Samples in the Main Analyses...48 Figure 2 Illustration of the Difference-in-Differences Research Design...49 Figure 3 Illustration of the Definition of Cross-Sold Bonds...50 vi

7 List of Appendices Appendix A Evolution of Major Regulations or Proposed Regulations on the Scope of Bank Activities...39 Appendix B Variable Definitions...42 Appendix C Interest Rates Charged on Cross-Sold Loans...46 Appendix D The First Stage of PSM...47 vii

8 1. Introduction While banking regulations and practices have a significant impact on non-financial firms operations, most research studying the consequences of banking regulations focuses on the direct effects of these regulations on banks; however, less is known about their spillover effects on non-financial firms. This paper attempts to provide insights into this important issue by exploring whether banking practices affect borrowing firms financial reporting. I examine the effect of bank cross-selling activities, defined as a bank s joint provisions of lending and underwriting services to the same firm. Specifically, I examine the effect of bank cross-selling on borrowing firms financial reporting quality for debt contracting and the use of financial information in debt contracts. I am interested in the impact of banking practices on accounting choices at borrowing firms in particular because, while a large body of accounting literature suggests that banks as lenders affect firms accounting policies, only limited empirical evidence exists on how banking sector characteristics affect firm financial reporting (Watts, 2003). I examine the effect of bank cross-selling in part because the question of whether commercial banks should be allowed to engage in investment banking activities has been under debate for decades. After the most recent financial crisis, some critics argue that the removal of barriers between commercial banking and investment banking contributed to the crisis, and they call for a revival of the Glass-Steagall Act. 1 Prior research on this debate 1 Their rationale is that breaking up universal banks can help solve the too big to fail problem and thus protect taxpayers from having to bail out banks due to their excessive risk-taking in non-traditional bank activities. For example, the 21st-Century Glass-Steagall Act was introduced to Congress in both 2013 and See for the details of the bill. In addition, the Volcker Rule (known as Section 619 of the Dodd-Frank Act) and the Lincoln 1

9 primarily focuses on the impact of cross-selling on banks underwriting practices, and there is only limited evidence on its implications on banks lending practices. Studying the impact of bank cross-selling on loan contracting and banks demand for financial reporting for debt contracting helps broaden our understanding of the impact of the regulatory changes in banking scopes. Cross-selling is likely to affect debt contracting and firm financial reporting through banks increased monitoring incentives and monitoring abilities. Compared to issuing a stand-alone loan, when a bank cross-sells lending and underwriting services to the same firm, it faces additional reputational risks associated with underwriting services: potential losses of business or legal liabilities as a result of failures to perform due diligence (e.g., Beatty and Ritter, 1986; Chemmanur and Fulghieri, 1994). 2 The heightened reputational risks incentivize the cross-selling bank to increase monitoring of the borrower and demand more relevant accounting information to facilitate monitoring. Cross-selling also enhances a bank s monitoring abilities. In the underwriting process, a bank gathers information to conduct due diligence and prepare registration statements. In addition to information collected during the underwriting process, the increased monitoring incentives motivate the bank to acquire more information during the lending process, thereby allowing the bank Amendment (known as Swaps Push-Out Rule or Section 716 of the Dodd-Frank Act) are also both manifestations of this sentiment. 2 Anecdotally, in one of the largest class-action settlements ever, Citigroup Inc. agreed to pay $2.65 billion to settle a suit brought by WorldCom investors. The lawsuit alleges that Citigroup and other investment banks did not conduct adequate due diligence before bringing WorldCom bonds to the market in May 2000 and May See for details. In addition, Blackwell et al. (1990) suggest that the reason we observe so few due diligence cases reaching court is the sensitivity of the value of the underwriters reputation to adverse publicity. 2

10 to better understand the underlying economics of the firm and better discipline the borrower s financial reporting for debt contracting purposes. Based on these arguments, I predict (1) a greater use of accounting-based monitoring mechanisms in cross-sold loans than in non-cross-sold loans and (2) a higher debt contracting value (DCV) of accounting information for cross-sold loan receivers than for non-cross-sold loan receivers following the reception of loans. I further explore whether the impact of cross-selling on the increase in banks monitoring varies with their past lending relationships with borrowers. I predict that past lending relationships mitigate the extent to which cross-selling affects debt contract design and firm financial reporting. Uncertainty about a firm s financial condition is heightened especially when the bank has no prior lending relationships with the firm. In addition, cross-selling allows the bank to enjoy more significant benefits from increased informational economies of scope in the absence of prior knowledge about the borrower s credit quality. Therefore, I expect that the increase in reliance on accounting-based monitoring mechanisms and the improvement in the DCV of accounting information due to cross-selling is more prominent when the bank does not have a prior lending relationship with the firm. To test these hypotheses, I define a cross-sold loan as a loan received by a firm that has issued bonds or shares within two years prior to the loan origination date, for which the same lead lender acts as an underwriter. To provide a valid comparison, I choose a control group of non-cross-sold loans as loans received within two years following an issue of a debt or equity security that is underwritten by a different bank. Therefore, all loans in my sample have debt or equity issues within two years prior to loan initiations, and the 3

11 only difference between cross-sold loans and non-cross-sold loans is whether or not the lead lender acts as the security underwriter. The sample in my debt contracting analysis includes 8,046 loans (i.e., 4,115 cross-sold loans and 3,931 non-cross-sold loans) issued from 1997 to Consistent with the first hypothesis, I find that cross-sold loans contain more financial covenants and accounting-based performance pricing provisions than noncross-sold loans. 3 For the analysis of whether cross-selling has an impact on the DCV of accounting information at borrowing firms, the treatment (control) sample consists of the period within five years after the initiation of the cross-sold loan (the non-cross-sold loan). I measure the DCV of accounting information using the relation between accounting profitability and future credit ratings (Bushman and Wittenberg-Moerman, 2012). I find that, compared to non-cross-sold loan receivers, cross-sold loan receivers have a greater DCV of accounting information, suggesting that bank cross-selling practices likely increase firms financial reporting quality for debt contracting. Cross-sectionally, I find that the impact of crossselling on debt contract design and firm financial reporting is more pronounced in the absence of prior lending relationships. In the additional analyses, to provide a further understanding of the use of accounting-based covenants, I also examine the tightness of financial covenants and the types of covenants in cross-sold loans. I find that cross-sold loans are associated with 3 In this dissertation, I use the term accounting-based covenants to refer to financial covenants and accounting-based performance pricing provisions. Financial covenants serve as tripwires that detect early signals of financial distress and facilitate transfers of control rights. Performance pricing provisions allow banks to charge higher interest rates when borrowers performance deteriorates and can also be used by banks as another mechanism for the state-contingent allocation of bargaining power (Asquith et al., 2005; Christensen et al., 2016). 4

12 tighter financial covenants, supporting the argument that banks have greater incentives to monitor cross-sold loans than non-cross-sold loans. I also find a greater use of performance covenants relative to capital covenants in cross-sold loans versus in non-cross-sold loans based on Christensen and Nikolaev s (2012) classification. This result is consistent with Christensen and Nikolaev s findings that the relative use of performance covenants to capital covenants is positively associated with the extent to which accounting information portrays credit risk. To address the selection bias (between the treatment and the control groups) and the related omitted variable concern, I employ three additional identification strategies. First, I match cross-sold loans with non-cross-sold loans on observable firm characteristics and loan characteristics using the propensity score matching (PSM) method. The results continue to hold using the PSM sample, suggesting that the impact of cross-selling on debt contracting and firm financial reporting is not driven by different firm or loan characteristics associated with cross-sold loans versus non-cross-sold loans. In addition, I refine the sample to further rule out the possibility that the findings are driven by heterogeneity across the borrowers or lenders assigned in the treatment and control samples. I first modify the sample by holding the borrowers constant. Specifically, I restrict the sample to firms that have received both cross-sold and non-cross-sold loans at different points in time and compare the contractual terms and firm financial reporting for crosssold loans versus for other loans within this refined sample. Next, I utilize a different control group by holding the lenders constant. Similarly, I restrict the sample to banks that have provided cross-sold loans and compare the contractual terms and firm financial 5

13 reporting for cross-sold loans versus for the other loans in this subsample. The inferences remain unchanged using these refined samples. In addition to the level analysis used in my primary tests, I also use a difference-indifferences design to study the impact of cross-selling on debt contract design and on firm financial reporting around the time that a firm first receives a cross-sold loan. Specifically, I construct a treatment sample using two consecutive loans received by a firm including the first cross-sold loan and the immediate preceding non-cross-sold loan. I then construct a control sample of two consecutive non-cross-sold loans. I find that the treatment sample exhibits a greater increase in the use of accounting-based covenants than the control sample, and that the improvement in the DCV of accounting information is more significant around the issuance of the first cross-sold loans than around the issuance of the first non-cross-sold loans. My results are robust to restricting the sample to the period before the financial crisis, suggesting that my findings are not affected by banks changes in lending behavior due to the financial crisis or related monetary policies such as quantitative easing. I also show that my findings for financial covenants are robust to controlling for general (non-financial) covenants. Lastly, I measure the DCV of accounting information using the timely loss recognition models in Basu (1997), and the results continue to hold. The paper makes the following contributions. First, my study adds to the emerging literature that examines the spillover effects of banking sector regulations or practices on non-financial firms. While a growing number of papers study the real effects of banking deregulations on geographical expansions (e.g., Amore et al., 2013; Chava et al., 2013; Cornaggia et al., 2015), very little research examines how banking regulations or practices 6

14 affect firm financial reporting. One exception is Gormley et al. (2012) who study how banking competition from international banks affects firms timely loss recognition in India. My paper complements Gormley et al. by focusing on a different aspect of banking pracitices (i.e., deregulations on bank product lines) and by introducing the effect of banks underwriting on firm financial reporting. Second, this research provides insights into the debate on the appropriate scope of bank activities by focusing on a bank s joint provisions of lending and underwriting services to the same firm. I show that banks use more accounting-based monitoring mechanisms in cross-sold loan contracts and demand more relevant accounting information from cross-sold loan receivers. This study points to a potential benefit arising from permitting cross-selling activities: namely, increased informational economies of scope that allow for more efficient debt contracting. Although Drucker and Puri s (2005) findings that banks charge lower interest rates on cross-sold loans are potentially consistent with the informational economies of scope mechanism, their results cannot rule out the possibility that allowing concurrent lending and underwriting leads to banks pursuit of underwriting business at the expense of loan quality. My study complements Drucker and Puri by providing evidence on cross-selling banks increase in monitoring and demand for financial reporting with high debt contracting value. 4 4 It is beyond the scope of this paper to assess whether returning to the Glass-Steagall separation would reduce social welfare overall. 7

15 2. Institutional Background and Literature Review 2.1 Evolution of Regulations on the Scope of Bank Activities Prior to 1933, commercial banks in the U.S. were permitted to underwrite public securities. In 1933, as a legislative response to bank failures during the Great Depression, the U.S. government enacted the Glass-Steagall Act, imposing a rigid separation between commercial banking and investment banking to address the concerns over the conflicts of interest within financial institutions and the potential for increased riskiness of the financial system. One of the goals of the Glass-Steagall Act was to ensure the safety and soundness of the U.S. banking industry. Starting from April 1987, commercial banks were allowed to establish separate Section 20 securities affiliates as investment banks and generate up to 5% of total revenue. The erosion of the Glass-Steagall Act continued into the 1990s until the final passage of the Gramm-Leach-Bliley (GLB) Act in 1999, which completely removed the restrictions on banks, securities firms and insurance companies from engaging in more than one of these three lines of business in the same organization. The recent financial crisis has re-energized the debate over the regulations on the scope of bank activities. The 21st Century Glass-Steagall Act was introduced to the Congress in 2013 and The purpose of the bill is stated as follows: (1) to reduce risks to the financial system by limiting banks ability to engage in activities other than socially valuable core banking activities; (2) to protect taxpayers and reduce moral hazard by removing explicit and implicit government guarantees for highrisk activities outside of the core business of banking; and (3) to eliminate conflicts of interest that arise from banks engaging in activities from which their profits are earned at the expense of their customers or clients. 8

16 In addition, the Volcker Rule (known as Section 619 of the Dodd-Frank Act), first publicly endorsed by President Obama on January 21, 2010, prohibited banks from engaging in short-term proprietary trading of securities, derivatives, commodity futures, and options for their own account. The final rules became effective April 1, In addition, the Lincoln Amendment (known as Swaps Push-Out Rule or Section 716 of the Dodd-Frank Act), first enacted in 2010, prohibited the provision of Federal assistance to a Federal depository institution that is a swap entity. This provision was further amended (or repealed ) in December 2014 due to considerable public criticisms, and the new amendment significantly narrows the scope of swaps subject to the push-out requirement. See Appendix A for the timeline of the evolution of major regulations or proposed regulations on the scope of bank activities in the U.S. history. Several studies examine the implications of abandoning Glass-Steagall restrictions from the perspective of financial institutions. For example, Yildirim et al. (2006) examine the wealth effects of the events surrounding the passage of the GLB Act of 1999 and find a positive market reaction to the GLB Act for investment banks and insurance firms. My study complements this literature by focusing on the implications of this deregulation for non-financial firms rather than financial institutions. 2.2 Research on Combining Lending and Underwriting Given the controversy concerning the appropriate scope of banking activities, a substantial literature attempts to speak to this issue by studying the consequences of allowing commercial banks to underwrite public securities. This line of literature can be broadly classified into two categories. First, much of the literature focuses on how a bank s 9

17 concurrent lending and underwriting affects its underwriting business (Drucker and Puri, 2008). Prior research suggests that there are benefits and costs associated with concurrent lending and underwriting. On the benefit side, previous research suggests that the increased scope of bank activities enhances the bank s ability to credibly certify the quality of its borrowers to outside investors due to its information advantage (e.g., Puri, 1996; Gande et al., 1997; Duarte-Silva, 2010). On the cost side, some research also raises concerns associated with conflicts of interests. Specifically, a bank as a lender has more private information about a firm than public investors do due to previous loan monitoring activities, and it may have incentives to underwrite public securities for a bad firm so that the firm can use the proceeds to repay bank loans. A few theory studies (e.g., Kanatas and Qi, 1998; Puri, 1999) model the trade-off between informational economies of scope and conflicts of interest. The empirical literature in this area finds little support for banks exploiting of conflicts of interest. Instead, most studies find that cross-selling improves the certification ability of commercial banks. Second, a few papers investigate the impact of a bank s concurrent lending and underwriting on its lending practices. Drucker and Puri (2005) show that concurrent lending and underwriting allows banks to offer discounted yield spreads on cross-sold loans due to economies of scope. Their results are potentially consistent with the informational economies of scope mechanism. However, lower interest rates could also be a result of banks tying practices, where banks alter interest rates based on a firm s decision to use their underwriting services. A related concern that banks may pursue underwriting business at the expense of loan quality. My study alleviates this concern by documenting banks increase in monitoring. Neuhann and Saidi (2016) model the effect of bank scope 10

18 due to cross-selling on banks monitoring incentives and show that banks of broad scope have incentives to monitor a firm even when they retain small loan shares. They further use the deregulation on bank scope activities to empirically test the model by showing that increases in bank scope result in reduced lead shares retained by the bank. Neuhann and Saidi (2017) find that the deregulation on bank scope activities improves the access to finance for risky ventures of public firms. My study compliments Neuhann and Saidi (2016, 2017) by examining the implications of this deregulation for banks monitoring mechanisms and borrowers financial reporting policies. 2.3 Research on Bank-Firm Relationships A large theory and empirical literature examines the costs and benefits of strong bank-firm relationships by focusing on repeated lending interactions. On the one hand, some research suggests that one benefit associated with repeated lending is that it reduces information asymmetry between the borrower and the lender, thereby increasing the availability of credit and lower loan spreads (e.g., Petersen and Rajan, 1994; Bharath et al., 2009). The other benefit is that a strong bank-firm relationship helps to accommodate special contractual features that can improve welfare (Boot 2000). For example, it leaves room for flexibility in loan contracts that allow for better monitoring and possible valueenhancing inter-temporal transfers in loan pricing. On the other hand, studies also show the costs of strong bank-firm relationships. Some research suggests that information advantage gained through repeated lending may allow the bank to extract rents from its information monopoly and potentially to hold up a firm because of information 11

19 asymmetries between other lenders and the firm (e.g., Sharpe, 1990; Rajan, 1992; Santos and Winton, 2008; Schenone, 2009). However, bank-firm relationships extend beyond repeated lending interactions, particularly for universal banks, as universal banks offer other financial services, underwrite and trade securities, and manage investment funds. Ferreira and Matos (2012) use the bank s representation on boards of directors or the bank s holding of shares through bank asset management divisions as measures of bank-firm relationship, and document that stronger bank-firm relationships are associated with higher loan spreads during the credit boom period but lower spreads during the financial crisis. Dass and Massa (2012) measure the strength of the bank-firm relationship by bank-firm proximity, size of the loan, and the bank s equity ownership in the borrowing firm, and find a strong bank-firm relationship improves the borrowing firm s corporate governance due to better monitoring but lowers the firm s stock liquidity due to increased information asymmetry for the investors in the borrower s stock. My study complements this literature by identifying a different form of bank-firm relationship (i.e., a bank s joint provision of lending and underwriting services to the same corporate client) and examining issues related to this broadened definition of bank-firm relationship. 2.4 Research on the Role of Accounting Information in Debt Contracting An extensive theoretical and empirical literature suggests that financial statement information is particularly important for contracting purposes (Watts and Zimmerman, 1986; Ball, 2001; Holthausen and Watts, 2001; Watts, 1993). Provisions such as financial covenants or some performance pricing provisions are directly written on accounting 12

20 variables. Large empirical evidence suggests that the properties of accounting numbers are associated with debt contract design. For example, Demerjian (2011) shows evidence that a reduction in the contracting usefulness of the balance sheet is associated with a lower use of balance sheet covenants. Costello and Wittenberg-Moerman (2011) find that when a firm experiences a material internal control weakness, lenders decrease their use of financial covenants and accounting-based performance pricing provisions. Dou (2016) documents that borrowers with a lower DCV of accounting information are more likely to renegotiate accounting-based terms. One traditional view suggests that contracting is more efficient when accounting rules are conservative, and that the demand from debtholders is the main reason why firms adopt conservative accounting policies (Watts and Zimmerman, 1986; Watts, 2003). Accounting conservatism is desirable to debtholders because they have asymmetric payoffs with respect to a firm s net assets and therefore are more concerned with the lower bound of the earnings and asset distributions. An extensive literature documents the association between accounting conservatism and debt contracting practices. For example, Zhang (2008) finds that interest rates are lower for firms with more conservative financial reporting. Beatty et al. (2008) find that lenders use conservative contract modifications (i.e., income escalators) to meet their demand for conservatism. A few other studies provide direct evidence that debtholders demand accounting information for debt contracting purposes. For example, Gormley et al. (2012) study the entry of foreign banks into India during the 1990s and find that foreign bank entry is associated with an increase in timely loss recognition by borrowing firms. Erkens et al. (2014) find that lender monitoring through board representation reduces demand for accounting conservatism in borrowers 13

21 financial reporting. My study relates to this literature by providing evidence that a bank s provisions of multiple products shape the borrower s financial reporting for debt contracting purposes. 3. Hypothesis Development When a bank cross-sells lending and underwriting services to the same firm, it bears additional reputational risks arising from underwriting services compared to issuing a stand-alone loan (Lu, 2007; Neuhann and Saidi, 2016). Accordingly, the bank has greater incentives to prevent borrower bankruptcy when it jointly provides lending and underwriting services to the same firm. Based on the incomplete contract theory, this can be achieved by including more financial covenants in debt contracts that allow them to acquire the control rights and take actions when firm performance deteriorates, but well before bankruptcy (Aghion and Bolton, 1992; Christensen et al., 2016). A bank s increased incentives to monitor due to cross-selling also motivates it to acquire more information in the loan syndication process compared to a regular non-crosssold loan (Neuhann and Saidi, 2016). In addition, the bank as an underwriter gathers information about the firm s creditworthiness by conducting due diligence research and coordinating the preparation of detailed information about the firm and its financial status (the registration statement) to be filed with the SEC (Liaw, 2012). The incremental information gathered in the lending process and the preceding underwriting process allows the bank to have a closer understanding of the borrower s underlying economics, thereby better disciplining the borrower s ability to withhold bad news. If the bank detects that the initially contracted-upon accounting numbers fail to 14

22 capture new credit risk information, it can renegotiate the accounting-based contractual terms with the firm (Dou, 2016). The borrower has incentives to agree to renegotiate because refusal to negotiate may lead to more unfavorable outcomes once the covenants are violated in the future. For example, the lender may threaten not to grant a waiver upon debt covenant violation or to require more stringent contractual terms (e.g., increase collateral requirement or charge higher interests) in renegotiations. Therefore, the bank s improved understanding of the borrower s underlying economics helps discipline the borrower s financial reporting and enhances the contracting usefulness of reported accounting numbers (e.g., Watts, 2003; Nikolaev, 2010). As a result, banks are better able to utilize more accounting-based contractual terms in cross-sold loans than in non-crosssold loans. 5 Based on the above arguments, the first two hypotheses are stated as follow: H1: Cross-sold loans contain more financial covenants and accounting-based performance pricing provisions than non-cross-sold loans. H2: Following the reception of loans, cross-sold loan receivers have a higher debt contracting value (DCV) of accounting information than non-cross-sold loan receivers. Finally, I explore the cross-sectional variation in the impact of cross-selling on the improvement in banks monitoring. I predict that past lending relationships mitigate the extent to which cross-selling affects debt contract design and firm financial reporting. Uncertainty about a firm s financial condition for the bank as both an underwriter and a 5 It is possible that better-informed banks can worry less about potential conflicts of interest and impose fewer financial covenants. Nevertheless, in general, better information about the borrower should allow for flexibility and discretion in contracting (Boot, 2000). A loan contract is a package of n-contractual terms (Melnik and Plaut, 1986) and agency theory suggests that there is a trade-off between the number of covenants and the interest rate (Jensen and Meckling, 1976; Myers, 1977; Smith and Warner, 1979). Previous empirical research documents that banks charge lower interest rates on cross-sold loans (Drucker and Puri, 2005). Therefore, in this context, I expect that better information about the borrower allows the bank to charge lower interest rates and increase the use of covenants. 15

23 lender is heightened when the bank has no prior lending relationships with the firm. In addition, cross-selling allows the bank to enjoy greater benefits from increased informational economies of scope in the absence of prior knowledge about the borrower s credit quality. Based on these arguments, my third set of hypotheses is as below: H3a: The impact of bank cross-selling on the use of financial covenants and accounting-based performance pricing provisions is more pronounced when the bank does not have a prior lending relationship with the firm. H3b: The impact of bank cross-selling on the DCV of accounting information is more pronounced when the bank does not have a prior lending relationship with the firm. 4. Research Design 4.1 Data and Sample Selection I obtain the lending data from LPC Dealscan, the underwriting data from SDC and Mergent FISD, the accounting data from Compustat Annual, and the stock returns data from CRSP. Also, when Dealscan loan-observations report no financial covenants, I use loan agreements provided by Amir Sufi for the period from 1996 to 2005 and hand-collect loan agreements for 1994, 1995, and 2006 to 2013 from SEC filings (from 10-K, 10-Q, and 8-K filings) to avoid data errors. 6 I manually match the underwriters in SDC and in Mergent FISD with the lenders in Dealscan to identify cross-sold loans. 7 My primary analyses of loan spreads and covenants include loans issued between October 31, 1997 and December 31, This research design choice is made because 6 Professor Amir Sufi at the University of Chicago shares data covenant data for the period from 1996 to 2005 on the website. 7 I also use the Freedom of Information Act to obtain the historical Fed approval dates of the Section 20 subsidiaries to cross-check my matching of the underwriter and the lender. 16

24 banks cross-selling was significantly boosted by a revision to Regulation Y (Docket No. R-0958). This deregulation became effective on October 31, 1997 and removed the information firewalls between the commercial-bank and the securities divisions within bank-holding companies. 8 I exclude borrowers in the financial services industry from the sample due to their different nature of business and industry-specific regulations. 9 My final sample in the loan analysis includes 8,046 loans (4,115 cross-sold loans and 3,931 non-cross-sold loans) with financial covenants data available. 10 I define a crosssold loan as a loan received by a firm that has issued bonds or shares within two years prior to the loan origination date, for which the same lead lender acts as an underwriter. This choice of two years is made following Neuhann and Saidi (2016, 2017). To make the comparison valid, I define a non-cross-sold loan as a loan that occurs within two years following an issue of a debt or equity security that is underwritten by a financial institution other than the lenders. Figure 1 illustrates the details of the treatment sample and the control sample. Therefore, all loans in my sample have security issues within two years prior to their origination dates, and the only difference between cross-sold loans and non-crosssold loans is whether or not the lead lender acts as the security underwriter. Further, the 8 The details of this deregulation can be found on the website: gov/boarddocs/press/ boardacts/1997/ /r-0958.pdf. 9 Based on previous studies (e.g., Black et al., 2004; Vasvari, 2010), one reason for eliminating financial firms in this context is that debt covenants for these firms are redundant. Regulatory monitoring is more efficient than covenants, and therefore financial institutions that borrow have fewer covenants for this external reason. 10 Dealscan includes both loan originations and loan amendments. Loan amendments refer to cases where loans are refinanced or amended to increase loan amount, change loan maturity, or modify other loan characteristics (e.g., change from a revolving loan to a term loan). I do not distinguish loan amendments from my loan originations because the same arguments also apply to loan amendments. 17

25 firm financial reporting analyses use data within five years after the loan origination date. In this analysis, I require a firm to have non-missing data for at least five years after the loan initiation so that I can estimate a reasonably stable measure of the DCV of accounting information. If a firm receives more than one loan facility in the year, only one facility is included in the analysis to avoid the impact of the weight carried by multiple loan facilities a firm receives within the same year. The unit of analysis is the firm-year. The final sample contains 16,011 observations at the firm-year level. Among these 16,011 observations, 7,944 firm-years are associated with cross-sold loans, while 8,067 are associated with noncross-sold loans. 4.2 Model Specifications To test H1, I estimate Model (1) to study the use of accounting-based contractual terms in cross-sold loans. Dep Var = β 0 + β 1 CrossSoldLoan+ β 2 LoanMaturity+ β 3 LoanSize+ β 4 Secured + β 5 LoanPurpose+ β 6 PastLend+ β 7 FirmSize+ β 8 ROA+ β 9 Leverage + β 10 MarketBook+ β 11 Rated+ β 12 Volatility+ β 13 Zscore+ β 14 LoanSpread +IndustryFE+YearFE+ε (1) The dependent variable is the natural log of the number of financial covenants (LogFinCov) and the use of accounting-based price performance provision (PerformPricAcc). 11 I estimate the model using the OLS and Probit regressions, respectively. The variable of interest is CrossSoldLoan, which is equal to one for the 11 Accounting-based performance pricing provisions refer to provisions based on values derived from financial statement numbers such as current ratio, debt service coverage, debt to equity, debt to tangible net worth, debt to borrowing base rating, EBITDA, fixed charge coverage, interest coverage, leverage, liquidity, loan to value, net income, quick ratio, senior debt to cash flow, senior leverage, and total debt to cash flow. 18

26 treatment sample (i.e., cross-sold loans) and zero for the control sample (i.e., non-crosssold loans). A greater β 1 suggests that banks use more accounting-based covenants in cross-sold loans than in non-cross-sold loans. Based on H1, I predict β 1 to be positive. I control for a battery of loan characteristics and borrower characteristics motivated by other studies (e.g., Bharath et al., 2009; Costello and Wittenberg-Moerman, 2011). The loan characteristics I control for include interest rate, loan maturity, loan size, collateral, loan purpose, previous lending relationship. I also control for firm characteristics such as firm size, return on assets, rating status, leverage, market-to-book ratio, and firm riskiness. I further include industry and year fixed effects to control for the unobserved heterogeneity across industries and time. I allow the errors in Model (1) to be correlated across loan facilities within a given firm, so the standard errors are clustered by firm. Appendix B provides variable definitions. To test H2, I estimate Model (2) below to study the impact of cross-selling on the DCV of accounting information at borrowing firms. CreditRatingt+1 or t+2 = β 0 + β 1 CrossSoldLoan+ β 2 ROA + β 3 CrossSoldLoan ROA + β 4 FirmSize + β 5 Leverage+ β 6 MarketBook + β 7 CreditRating +β 8 Rated +β 9 LoanSpread +β 10 LoanSize +β 11 LoanMaturity +IndustryFE +YearFE+ε (2) The model is adapted from Bushman and Wittenberg-Moerman (2012) where the DCV of accounting information is measured as the extent of accounting profitability predicting future credit ratings. 12 The dependent variable is a firm s crediting rating in year 12 Other studies provide different measures of debt contracting values at the industry level. For example, Ball et al. (2008) use the ability of changes in quarterly accounting earnings to predict credit quality downgrades (Somer s D from the Probit regression). Christensen and Nikolaev (2011) measure the ability of levels of accounting variables to predict credit qualities (pseudo R 2 from the ordered Logit regression). These 19

27 t+1 and t+2, measured by the numerical equivalent of the S&P Domestic Long Term Issuer Credit Rating. It is set from one (when the rating is D) to 22 (when the rating is AAA). I assign a value of zero if the firm is not rated and include an indicator variable to denote firms with ratings. CrossSoldLoan takes a value of one (zero) if a year falls in the five years subsequent to the initiation of a cross-sold loan (a non-cross-sold loan). The variable of interest is CrossSoldLoan ROA. I also control for firm characteristics that are known to affect firm financial reporting quality. Similarly, I include industry and year fixed effects to control for the unobserved heterogeneity across industries and time. I allow the errors in Model (2) to be correlated across different years for a given firm, so the standard errors are clustered by firm. Based on H2, I predict the coefficient on β 3 to be more positive (i.e., a greater mapping between firms accounting profitability and future credit ratings) for crosssold loan receivers than for non-cross-sold loan receivers. To test H3a, I re-estimate Model (1), adding the interaction term CrossSoldLoan PastLend. PastLend is equal to one if a firm has borrowed from the same lead lender(s) in the past five years and zero otherwise. H3a predicts a negative coefficient on this interaction term, suggesting that past lending relationships mitigate the impact of cross-selling on debt contract design. Further, to test H3b, I re-estimate Model (2) to study the differential impact of cross-selling on firm financial reporting based on whether the bank has a lending relationship with the firm in the past five years. Based on H3b, I predict the coefficient on β 3 to be more positive when the bank has no past lending relationship with the firm. industry-level measures are less suitable in my context because firms within the same industry could be crosssold loan receivers and non-cross-sold loan receivers. 20

28 5. Empirical Results 5.1 Descriptive Statistics Table 1 shows the distribution of cross-sold loans by year. The stepwise repeal of the Glass-Steagall Act in the late 1990s significantly boosted banks cross-selling of loans and underwriting services to the same corporate customer. The incidences of cross-sold loans grow from 149 (accounting for 2.5% of the total U.S. syndicated loan market) in 1996 (the year before the deregulation) to 820 (accounting for 11.9% of the total U.S. syndicated loan market) in The growth in dollar amount is even more significant: the amount of cross-sold loans in 2013 is almost 13.4 times as much as in Table 2 reports summary statistics for all the variables used in the primary analyses. On average, the loans in my sample contain 2.50 financial covenants, and 37 % of the loan agreements include an accounting-based performance pricing provision. The mean of CrossSoldLoan suggests that cross-sold loans consist of 51% of the sample for the loan analysis. For the DCV analysis, the average of credit rating for the current period is 8.43 (between B and B+). The distribution of other control variables is largely consistent with the prior research. 21

29 5.2 Main Findings Table 3 presents the results of the use of accounting-based contractual terms in cross-sold loans. 13 Columns 1 and 2 report the results using the full sample. In column 1, the coefficient on CrossSoldLoan is positively significant (0.0504, t-statistic = 4.38), indicating that cross-sold loans contain more financial covenants than do non-cross-sold loans due to banks enhanced monitoring incentives and monitoring abilities. Specifically, a cross-sold loan contains 1.05 more financial covenants than a non-cross-sold loan. This incremental effect is economically significant compared to the sample mean of 2.5 covenants. In column 2, the coefficient on CrossSoldLoan is positively significant (0.1503, z-statistic = 3.15), suggesting banks are more likely to use accounting-based performance pricing provisions in cross-sold loans than in non-cross-sold loans. The findings are consistent with H1 that cross-selling leads to a greater reliance on accounting-based contractual terms as monitoring mechanisms. One may argue that cross-selling is an endogenous choice made by the bank and the firm, and the factors leading to this choice may also affect debt contract design. To address this concern, I match cross-sold loans with non-cross-sold loans on observable firm characteristics and loan characteristics that may affect cross-selling decisions using the PSM method. 14 Columns 3 and 4 report the results using the PSM sample. The positive 13 In Appendix C, I present results to confirm the findings in the prior literature that cross-sold loans overall have lower interest rates than non-cross-sold loans. 14 I match cross-sold loans and non-cross-sold loans based on loan size, loan maturity, firm credit rating status, firm size, leverage, volatility, past lending relationship, industry and year. These matching variables are chosen following Drucker and Puri (2005). I use nearest neighbor PSM matching without replacement and perform covariate balance checks to ensure that the matching is successful. This matching process results in a sample size of 2,720 loan facilities (1,360 cross-sold loans and 1,360 non-cross-sold loans). The firststage PSM results in Appendix D suggest that firms with a larger size, a higher leverage, a credit rating, and 22

30 coefficients on CrossSoldLoan are consistent with previous findings that banks use more accounting-based contractual terms in cross-sold loans than in non-cross-sold loans. These results further support the idea that the differential use of accounting-based covenants in cross-sold loans versus in non-cross-sold loans is not driven by observable firm or loan characteristics that may affect the cross-selling choice. Panel A of Table 4 presents the impact of cross-selling on the DCV of accounting information using the full sample. In both columns 1 and 2, the coefficient on CrossSoldLoan ROA using the full sample is positively significant (coefficient = and ), suggesting that accounting profitability can better predict credit ratings for cross-sold loan receivers than for non-cross-sold loan receivers. Columns 3 and 4 report the results utilizing the PSM sample. Similarly, I find that compared to non-cross-sold loan receivers, cross-sold loan receivers exhibit a greater DCV of accounting information after obtaining the loan. Again, this finding suggests that my inferences using the full sample are not affected by different firm or loan characteristics across the treatment and the control samples. The evidence suggests that firms respond to banks increased monitoring by improving financial reporting quality for debt contracting. I examine the cross-sectional variations in the impact of cross-selling on the increase in bank monitoring and present the results in Table 5. Because the magnitude of the interaction effect in nonlinear models is conditional on the independent variables, the conventional coefficients on interaction variables in Probit regressions do not provide a statistical test of whether the economic interaction of interest is statistically significant (Ai previous borrowing relationships with the bank are more likely to receive cross-sold loans, and that crosssold loans are larger and have longer maturities on average. 23