Sharing the surplus with clients: Evidence from the protection of bank proprietary information

Transcription

1 Sharing the surplus with clients: Evidence from the protection of bank proprietary information Abstract We examine the effect of the increased protection of banks proprietary information on the surplus-sharing between banks and clients. Our identification strategy relies on the passage of the inevitable disclosure doctrine (IDD), which prevents a bank s former employees from leaking proprietary information to rivals. We show that, after the passage of the IDD, banks offer loans with lower interest rates and longer time to maturity, and the loan interest rate is further lowered for bank-borrower pairs that have a prior lending relationship. Moreover, IDD adoptions result in a more long-lasting lending relationship. Further results show that banks whose trade secrets are protected by the IDD can sell loans at higher loan bidding prices and lower bid-ask spreads in the secondary loan market. Our findings suggest that the protection of proprietary information increases the value of establishing long-run lending relationships for banks; in response, banks share part of the surplus with their borrowers in the primary loan market. Keywords: proprietary information; relationship lending; primary loan market; secondary loan market. JEL Classification: D80; D40; G21

2 1. Introduction Banks are secret keepers. Through repeated interactions with a customer, a bank can obtain a significant amount of proprietary information. Such information acquisition allows the bank to achieve a preferred position in competing for future business (Drucker and Puri, 2005; Bharath et al., 2007) and intertemporal smoothing in loan contract terms (Haubrich, 1989; Greenbaum et al., 1989). On the other hand, the risk of information leakage discourages banks from collecting proprietary information or providing relationship-specific services, as they anticipate that the leaked information would aid competitors in poaching clients. Therefore, keeping customers proprietary information from competitors is essential in the banking industry. 1, 2, 3 Even so, preventing information leakage is challenging due to high labor mobility in the banking sector. In practice, proprietary information regarding a particular customer is acquired through communications between the loan officer and the borrower s management (Uchida et al., 2012). When a loan officer who is privy to proprietary information switches to a new employer, it is highly possible that the former employer s proprietary information will leak to the new employer. This will exacerbate lending market competition and jeopardize the existing bank borrower relationship (Peterson and Rajan, 1995). Gregory Fleming, the president of Morgan Stanley s wealth-management arm, said in an interview, While the situation is disappointing, it is always difficult to prevent harm caused by those 1 A relationship bank s information advantage can also benefit the borrower (Bharath et al., 2011). Without an information advantage, banks cannot share future surplus with the borrower, which can result in borrowers not receiving credit at all (Rajan, 1992; Peterson and Rajan, 1994). 2 Padilla and Pagano (1997) discuss an equilibrium in which a certain degree of information sharing can emerge when banks have an incentive to increase the effort of an entrepreneur of the borrowing firm, even though profits via rent extraction would be reduced. 3 A large body of accounting literature examines the proprietary-cost hypothesis in real-sector firms (e.g., Harries, 1998; Botosan and Stanford, 2005; Berger and Hann, 2007; Dedman and Lennox, 2009; Berger, 2011; Bens et al., 2011; Ali et al., 2014). 1

3 willing to steal, 4 demonstrating such a concern about information leakage in the banking business. In this paper, we seek to examine the economic consequences of the protection of banks proprietary information. We exploit the staggered recognition of the inevitable disclosure doctrine (IDD) by U.S. state courts, which increases the protection of firms (banks ) trade secrets by preventing workers who know a firm s trade secrets from leaking these to a rival firm. Specifically, the recognition of the IDD gives a former employer a stronger right to sue a former employee over leaked proprietary information. 5 Given such a legal liability, departing staffs of a bank are discouraged from leaking customers proprietary information to this bank s competitors, which would protect this bank s relationship with its existing clients. In other words, after the passage of the IDD, proprietary information obtained in the prior bank borrower relationship is expected to produce more surplus over a longer period. Our research question focuses on whether banks will share such surplus with their clients. On the one hand, a majority of relationship banking studies suggest that a bank would share a part of the surplus stemming from a long-term bank borrower relationship, resulting in better loan contract terms to relationship borrowers (e.g., Boot and Thakor, 1994; Bharath et al., 2007, 2011). Along with this line, the passage of the IDD, which would promote a bank s acquisition of proprietary information and facilitate long-term lending relationships, should further encourage value sharing between the bank and the borrower. We would, therefore, expect a reduction in the loan spread after the passage of the IDD. On the other hand, another stream of literature suggests that a strengthened bank borrower relationship need not improve borrowers welfare. Banks can leverage their superior 4 Justin Baer, The Wall Street Journal. January 5, The IDD also enhances the enforcement of existing non-competing agreements (Klasa et al. 2015). In this respect, the IDD complements rather than substitutes for a non-competing agreement. 2

4 information about clients to extract more rents. This is referred to as the hold-up problem (Sharpe, 1990; Rajan, 1992; Dell Ariccia et al., 1999; Degryse and van Cayseele, 2000; Santos and Winton, 2008). For example, Degryse and van Cayseele (2000) find that loan interest rate increases with a relationship s duration. In this respect, trade secret protection is likely to strengthen banks information monopoly about a borrower, which can exacerbate the hold-up problem. Therefore, how the protection of banks proprietary information affects loan contract terms offered to clients is an empirical question. To isolate the protection of proprietary information of a bank from that of its clients, our identification strategy exploits the passage of the IDD in the state in which the bank is located. We exclude borrowers located in the same state as their banks; 6 that is to say, we focus only on out-of-state lending loans made to borrowers located in a different state. 7 Since IDD applies to all local employees (and only local employees), it can prevent the bank s existing employees from hopping to both in-state and out-of-state banks. Therefore, in our research design, the passage of the IDD in the bank s state can reduce the risk of information leakage to two kinds of competitors: (i) banks located in the same state but that also engage in out-of-state lending and (ii) banks located in a different state that is the affected bank s target market. To further pin down the supply-side effect (information protection on banks), we include a large set of fixed effects to absorb borrower-specific fundamentals. In particular, we include the borrower s state interacted with year fixed effects. 8 We also saturate the 6 Hereafter, the term passage of the IDD is used interchangeably with passage of the IDD in the state in which the bank is located. 7 Note that most deals in Dealscan database are syndicated loans, which usually require coordination among multiple banks and approvals from top executives. In this respect, these loans are most likely initiated and coordinated by relationship managers in headquarters. 8 This makes sure that our estimate is net of the demand-side effects found in Klasa et al. (2016). However, some borrowers have segments located in a state other than the headquarters state. In a 3

5 specification with interactions of borrower industry and year fixed effects. By doing so, we can exhaustively control for unobserved time-varying shocks to the borrowers, such as their risks or growth opportunities. We find that, on average, banks in states that passed the IDD charge a spread that is to bps lower than do banks in other states. In addition, these banks also provide loan contracts of longer maturity. These findings are consistent with those of Bharath et al. (2011), in that banks would share part of the value stemming from an increase in proprietary information protection (or a more durable lending relationship). IDD adoptions are more valuable to relationship banks as the information acquired in prior transactions becomes better protected and can be reused for more periods. Therefore, we predict that the effects of the IDD on loan contract terms are more pronounced for bankborrower pairs with a prior lending relationship. For relationship loans, we find a reduction of 31.7 bps in the loan spread charged by banks in states that passed the IDD. This magnitude of reduction is larger than the whole-sample estimate. In contrast, for nonrelationship loans, the reduction in loan spread is 14.4 bps, which is statistically and economically smaller. Note that one prerequisite for relationship banks to be willing to share the surplus lies in that they are more likely to retain their clients after the IDD. In particular, if a relationship bank is willing to share part of the surplus stemming from a better protection of proprietary information, we would expect its clients to choose this bank at a higher probability, resulting in a reciprocal, long-lasting lending relationship. To verify this prerequisite, we follow the empirical design of Bharath et al. (2007), and show that IDD adoptions significantly enhance the likelihood of a borrower choosing prior relationship robustness test, we remove deals regarding borrowers that have multiple geographic segments and redo our tests. The result is qualitatively the same. 4

6 banks in future borrowing. This finding suggests that the IDD prolongs existing lending relationship. Finally, we go beyond the primary loan market and examine if the participants in the secondary loan market take the effect of IDD adoptions into account when pricing the loans. Due to the enhanced proprietary information protection, banks are encouraged to build a long-term banking relationship with borrowers, which will result in better monitoring from the bank (Bharath et al., 2011; Srinivasan, 2014). Creditable bank monitoring is essential in the secondary loan market since it can mitigate buyers concern about the agency problem that loan sellers may not have incentives to monitor (Drucker and Puri, 2008). In this respect, the IDD alleviates this moral hazard problem and increases the demand for loan purchases, thus making it easier for banks to sell part of their loans. To test this conjecture, we investigate the effect of IDD adoptions on loan trade bid-ask spreads. We find that IDD adoptions reduce loan trade bid-ask spread and increase the bidding price. We also find that the probability of loan sale increases after the passage of the IDD, but the result is not statistically significant. Jointly with the evidence regarding the primary loan market, our findings on the secondary loan market suggest that the value shared by banks could be partially attributed to the beneficial effect of the IDD on banks secondary loan market transactions. Our paper is related to Klasa et al. (2015), who also examine the effect of the IDD passage on loan interest rates. However, they explore how the protection of a borrower s trade secrets affects the loan spread the borrower is charged. In other words, we examine the economic consequences of a shock on the supply side (i.e., the bank) of the capital market, whereas Klasa et al. (2015) focus on that on the demand side (i.e., the borrower). The banking literature has provided ample discussion about the importance of distinguishing supply-side effect from demand-side effect (e.g., Khwaja and Mian, 2008; 5

7 Leary, 2009; Chava and Purnanandam, 2011; Murfin, 2012). Moreover, the theoretical predictions are different. Klasa et al. (2015) focus on how borrowers risk affects loan terms. Given that creditors have an unambiguous preference for borrowers downside risk, they have a one-way prediction that the loan spread would decrease. However, our prediction, on how banks information risk shapes the equilibrium loan pricing, is two-sided loan spread can either go up or go down, depending on the optimal response of banks to the protection of their trade secrets. Therefore, our investigation is related to, and contributes to, the controversies over benefits versus hold-up costs associated with long-term lending relationships (e.g.,rajan, 1992; Boot and Thakor, 1993; Boot, 2000; Gorton and Winton, 2003; Santos and Winton, 2008; Ioannidou and Ongena, 2010; Srinivasan, 2014). The general approach of the prior literature is to develop a proxy to measure a firm s susceptibility to hold-up and then examine variations in bank behaviors under different scenarios. Such proxies include: the presence of credit ratings (Santos and Winton, 2008), firm size, analyst coverage (Bharath et al., 2011), distance between the borrower and lender (Dass and Massa, 2011), and duration of lending relationships (Peterson and Rajan, 1994; Berger and Udell, 1995). However, these proxies can be varied or controlled by the borrower firm in response to capital and product market conditions and, hence, cannot be exogenous to dependent variables (e.g., Srinivasan, 2014). 9 Our paper overcomes the endogeneity problem by utilizing the quasi-natural experiment instituted by the IDD in the state in which the bank is located. Passage of the IDD is unlikely to be driven by client-specific characteristics. In addition, it may not affect or be affected by banks financial conditions. In this respect, state-level adoption of the IDD would directly impact bank client proprietary information 9 For example, the positive association between the relationship variable and the loan spread could reflect the self-selection of high-risk borrowers into relationships. 6

8 protection and in turn, affect the distribution of surplus between banks and their clients. Then, the value sharing (or hold-up) between banks and their borrowers can be implied via examining loan contract terms. Our paper is also related to the large body of proprietary cost literature. Prior studies center on the discussion about how proprietary cost affects corporate disclosures in realsector firms (i.e., Harries, 1998; Botosan and Stanford, 2005; Berger and Hann, 2007; Dedman and Lennox, 2009; Berger, 2011; Bens et al., 2011; Ali et al., 2014). Our study extends this stream of literature in two dimensions. First, the focus of the literature up to this point has been proprietary information of real-sector firms, thus our work fills the void concerning the banking sector, wherein the proprietary information about clients is extremely valuable. Second, we link the protection of proprietary information to the secondary loan market, in which proprietary information plays a key role in determining the loan trade volume and pricing (Drucker and Puri, 2009; Wittenberg-Moerman, 2010). In this sense, we also contribute to the secondary loan market literature (Drucker and Puri, 2009; Wittenberg-Moerman, 2010; Gande and Saunders 2012; Parlour and Winton, 2013). The remainder of the article is organized as follows. Section 2 presents the institutional background for banking-sector trade secrets and the passage of the IDD. Section 3 discusses theoretical predictions and testable implications. Section 4 describes the data and sample selection procedure. The research methodology and results are discussed in Section 5. We conclude the paper in Section Institutional background 2.1. Banks trade secrets: A lawsuit A bank s proprietary information about its customers is a trade secret and perhaps the bank s most valuable intangible asset. American Banker reported that a bank s customer 7

9 list and knowledge of a specific bank s risk tolerance and risk management are the bank s trade secrets. 10 However, such information, possessed by high-ranking employees, is also extraordinarily valuable to competitors (if it can be acquired), which results in a serious concern for information leakage. The importance of customers proprietary information is illustrated in the lawsuit between TD Bank and a former employee (Business Law News, 2014). 11 A former loan officer of TD Bank was accused of misappropriating sensitive customer information, such as tax returns and credit approvals, in the weeks before leaving his position in New Jersey in The officer then sent the information to Kearny Federal Savings Bank, his new employer and a competitor of TD Bank. TD Bank argues that this information is highly sensitive and that Kearny Federal Savings Bank could use it to solicit the clients of TD Bank. Based on the statement of the complaint, Kearny Federal s executives planned for the loan officer s departure from TD Bank and his new position in Kearny. The plan involved discussing TD Bank customers and deals and a competitive strategy to steal business from TD Bank. Undoubtedly, leaking important customer-specific information to a rival greatly jeopardized TD Bank. In fact, the problem of information leakage is not unusual in the banking industry (American Banker, 2014). Most banks have to rely on a non-disclosure agreement (NDA) and/or a covenant not to compete (CNC) to protect their business secrets. However, the protection provided by NDAs is limited, since violations must be detected and proven before the bank can initiate legal action. The protection of a CNC is also limited, since it is not effective when a former employee seeks to switch to a new job in another state (Malsberger, 2004; Klasa et al., 2015). 10 Andy Peters, American Banker, May 5,

10 2.2. The Inevitable Disclosure Doctrine (IDD) Compared with NDAs and CNCs, the IDD provides incremental protection of firms trade secrets (Klasa et al., 2015). First, a firm could sue a former employee who leaks its trade secrets to competitors even if the employee did not sign a CNC or an NDA with the firm. Second, the law is applicable even if a firm does not have direct evidence of a departing employee s bad faith or actual wrongdoing. Third, a firm could sue an employee hopping to a rival firm located in another state. Three conditions are necessary for the IDD to apply: (i) The employee had access to the firm s trade secrets, (ii) the employee s duties at the new employer would be so similar to those at the previous employer that, in performing them, the employee will inevitably use or disclose the trade secrets, and (iii) disclosure of the trade secrets would inflict irreparable economic harm on the firm s business. Regarding the reasons for passing the IDD state by state rather than nationally (in one fell swoop), Harris (2000) and Godfrey (2004) show that changes in the IDD provide a balance between employers interests in protecting trade secrets and employee mobility freedom. We use the time of a precedent case, the first time a law is brought to court by a plaintiff, as the effective date for the passage/rejection of IDD, which is grounded by legal academia. The doctrine of stare decisis ( stand by decided matters ) is the tradition of Anglo-American common law. It states that a court looks to past decisions that is, precedents for guidance on how to respond to a new case. To use a precedent case as a reference has been justified as providing predictability, stability, fairness, and efficiency in the law (Lehman and Phelps, 2005). In a precedent case, the court will explain how, when, and where the law can be applied. For example, the IDD was passed in a precedent case in 9

11 the state of Illinois on February 9, Therefore, banks in Illinois were not affected by this law change until February 9, Since February 9, 1989, banks have been able to use this law to protect their trade secrets by suing former employees who could potentially leak their customer lists and solicit their customers. 3. Hypotheses development 3.1. IDD and contract terms in primary loan market Proprietary information about borrowers is critical for a bank to produce a surplus from a lending relationship and compete in the financial market (Srinivasan, 2014). Banks thus spend resources in acquiring private information about the prospects of their customers and, meanwhile, are subject to the risk of the leakage of such information. A primary channel of information leakage is employee mobility. Passage of the IDD is aimed at preventing employees who have access to proprietary information from moving to a competitor. While the passage of the IDD increases the information advantage of banks engaged in the lending business, it is not clear how this shapes banks lending behaviors, which in turn has implications for the economic welfare of borrowing firms. Enhanced information protection reduces the cost of maintaining proprietary information and generates an extra surplus for the lender. Sharing this surplus with clients can be a Pareto improvement (Boot and Thakor, 1994; Drucker and Puri, 2005; Bharath et al., 2007, 2011). For example, banks can benefit from future mandates, such as to conduct more underwriting business and, meanwhile, offer to lower underwriting fees (or lower interest rates) to borrowers. If the benefit-sharing effect dominates, we would expect a reduction in the loan spread when the bank can better protect its proprietary information via the IDD. 10

12 On the other hand, the IDD can exacerbate the hold-up problem. Since the increased protection of customer information also enhances a bank s information monopoly, the bank can leverage this monopolistic power to hold up the borrower strategically and extract a higher rent (Santos and Winton, 2008; Hale and Santos, 2009). In this respect, we expect trade secret protection to lead to higher loan spreads. The passage of the IDD enables us to directly test benefit sharing versus the hold-up problem associated with the lending relationship. We develop the following hypotheses: H1a (benefit-sharing hypothesis): Banks protected by the IDD offer a lower loan spread to customers. H1b (hold-up hypothesis): Banks protected by the IDD charge a higher loan spread to customers IDD and loan trades in the secondary loan market The effect of IDD adoptions may go above and beyond the changes in loan contract terms in the primary loan market, and further impose effect on loan trades in the secondary loan market. In fact, this angle is becoming more and more important since the size of secondary loan market is increasing in the United States. For example, the market volume reached around $160 billion in 2008 from around $8 billion traded in Therefore, examining the effect of IDD adoptions on loan trades in the secondary loan market can help assess the importance of trade secret protection for the banking sector. An important concern of secondary-loan-market investors is the moral hazard problem: After selling the loan and shedding credit risk, the seller may not have the incentive to engage in costly screening and monitoring (Pennacchi, 1988; Gorton and Pennacchi, 1995). However, bank monitoring is particularly important in the secondary loan market since the majority of loan trading involves leveraged loans which are issued by riskier borrowers

13 (Diamond, 1991; Wittenberg-Moerman, 2010). Protection of proprietary information can help resolve this problem, since it enhances the value of customer information in the long run and thus encourages a relationship bank to monitor the customer (Srinivasan, 2014). Such a mitigation on moral hazard problem can be further priced in the loan transactions in the form of a lower bid-ask spread (Wittenberg-Moerman, 2010) or a higher bidding price. Based on these arguments, we develop the following hypotheses: H2: Protection on banks proprietary information via IDD reduces the loan trade bid-ask spreads and increases the bidding prices in the secondary loan market. 4. Data and sample selection Information about loan contract terms was obtained from LPC Dealscan. We use the Dealscan Compustat link file provided by Chava and Roberts (2008) to merge Dealscan data with borrowing firms whose accounting variables are available in Compustat. The loans in Dealscan are mainly syndicated loans. To identify the lead lending bank in each syndication, we follow Sufi (2007) to use the Lead Arranger Credit variable in the Dealscan database. Specifically, if the variable indicates yes, we define this bank to be a lead bank. Banks that are classified as lead arrangers usually hold a large fraction of a loan and are often the single arranger of the loan (Bharath et al., 2009). After this step, nearly 90% of loans in our sample have a single lead bank. 13 For some loans, the syndication involves more than one lead bank. These lead banks may be from different states and may have different timings of the passage of the IDD. To 13 Bharath et al. (2011) also include banks with the roles of agent, administrative agent, arranger, and lead bank as lead banks if their retained shares of the loan were greater than 25%. Our results are robust to using this alternative definition of lead banks. 12

14 ensure that we capture the unique effect of trade-secret protection, we exclude loans on which different lead banks have different status of IDD. 14 All money variables are deflated to the constant year 2000 U.S. dollars. Borrowing firms involved in major mergers (Compustat footnote code AB) are excluded. We also delete the following borrowing firms: Non-U.S. firms; firms with missing data on stock prices, shares outstanding, headquarters state, and the book value of assets; firms that reported format codes 4 to 6; financial firms (SIC code between 6000 to 6999); and firms with book assets below 10 million. We further exclude loans from banks headquartered outside the United States and banks with missing information on headquarters states. 15 After requiring non-missing borrower characteristics (used in our regressions), the final sample contains 10,777 facility lender borrower observations from 1989 to The information on the IDD passage dates comes from Klasa et al. (2015). Our sample period is , which starts on the year Illinois adopted the IDD, in 1989, and ends five years after Kansas adopted the IDD, in During our sample period, courts in 13 states adopted the IDD and courts in three states rejected the IDD they had adopted in prior years. Our sample period excludes events associated with the adoption of the IDD by a few states in earlier years because Dealscan s coverage of earlier years is sparser. 16 The precedent-setting legal cases adopting or rejecting the IDD are shown in Appendix B. Besides the primary loan market, we also examine the effect of IDD adoptions on the secondary loan market. Information about loans sold in the secondary market is collected 14 The inclusion of these loans and their definition as having received the treatment of the IDD does not affect our inferences. 15 In our sample observations, 0.4% consist of loans from foreign incorporated banks with U.S. headquarters (e.g., NatWest Bank NA, HSBC Bank USA NA). These banks registered their headquarters in New York as separate legal entities and thus are treated as New York located banks in our sample. Excluding these observations does not affect our results. 16 In the 1960s, Delaware, Florida, and Michigan adopted the IDD. In 1919, New York adopted the IDD. 13

15 from Markit Loan Pricing database. This database contains daily bid-offer price, numbers of bids, bid-ask spread, and maturity of loans sold in the secondary market in the United States since There are 3.1 million observations between January 2000 and December To identify loans that are sold in the secondary market from our Dealscan- Compustat sample, we match the sample with Markit Loan Pricing data based on two dimensions: the start year of the loan and borrower CUSIP. We define a loan as sold in a year if it has a quotation record in the Markit Loan Pricing database during the loan s lifetime. After the matching process, we obtain a loan-year panel in which each loan survives in the sample for a few years until the loan is sold based on the information from Markit. This step screened our sample to 7,613 loans, in which 339 loan facilities are sold in the secondary market from 2000 to Table 1 shows the summary statistics of various loan and borrower characteristics. Panel A reports key borrower characteristics, Panel B reports loan characteristics, and Panel C presents the status for the IDD passage. In Panel A and B, we report the summary statistics for our baseline sample and secondary loan market sample (from 2000 to 2011) separately. Our baseline sample contains 10,777 facility lender borrower cross-state loan observations from 1989 to 2011, in which the average borrower size is $28.81 million (the median level is $5.95 million). Borrowers are highly levered, with an average debt ratio of 43% (median 44%). The average loan spread is bps (median 175 bps), and the average maturity is 44 months (median 47 months). The average loan size is $2.43 million, while the median level is much smaller ($1.03 million). In the secondary loan market sample, 3% of the loan-year observations have a loan sale, and 96% of the observations are 17 Some earlier studies use Loan Syndications and Trading Association (LSTA) Mark to Market Pricing database to identify loans that are sold in the secondary market (e.g., Drucker and Puri, 2009). It is also a data set of daily secondary market loan quotations gathered by third-party providers (LPC and LSTA). There are 2.5 million observations between May 1998 and September 2005 in the LSTA Mark to Market Pricing database. 14

16 syndicated loans. Conditional on having a loan sale, the yearly average bid-ask spread is 0.98 and the yearly average bidding price is Empirical strategy and results 5.1. Empirical strategy We use a difference-in-differences method to exploit the staggered passage of the IDD across different states. After the recognition of the IDD by the state court, banks headquartered in that state have their proprietary information better protected, since employees will face a higher litigation cost if they leak information to rival banks. One important assumption here lies in that most of the syndication loan offers are originated from the state of bank s headquarters. 18 This assumption is reasonable since the size of a syndication loan is large and the syndication structure can be very complex, which requires multiple parties to cooperate. Locating the relationship managers in the headquarters could better facilitate not only the coordination within a lead bank but also that among participating banks. To justify this assumption, we manually check the job openings for syndication relationship managers in Unite State during November 12, 2016 to December 12, We use a search engine provide by We require that the scope of the job include the developing or maintaining a lending relationship. In addition, we require the employer banks to be included in our bank sample. This process yields 61 vacancies out of which 88.5% (54) are in the state in which the bank headquarters is located. 18 Banks may have branches outside the headquarters state, which are not affected by the IDD. If the loan is issued by a branch in a state other than the headquarters state, our identification would be undermined. However, since Dealscan mainly includes syndicated loan deals, and the majority of syndicated loans are originated by loan officers in head offices, this concern is less likely to affect our identification. 15

17 We define loans issued by treated banks (i.e., banks whose headquarters states are affected by IDD) as affected loans and compare them with loans issued by banks located in other states. In other words, what we are interested in is the supply-side effect of loan terms. To ensure that the demand-side effect does not drive our results (e.g., borrowers located in the same state and thus simultaneously affected by the IDD), we exclude loans in which the borrower and lender are located in the same state. Therefore, our primary focus is banks that are involved in cross-state lending. However, the results in this paper are robust to the inclusion of loans for which the borrower and the lender are from the same state. More specifically, our difference-in-differences approach employs the following regression model: Loan term i,j,s,s,k,t = β 0 + β 1 IDD s,t + β j Borrower characteristics j,t + β i Loan characteristics i,t + ω s + μ s t + φ k + ε i,j,s,s,k,t (1) where i denotes the loan, s denotes the state of the lender, s denotes the state of the borrower, t denotes the year in which the loan was issued, j denotes the borrower of the loan, k denotes the industry of the borrower, and IDD is a dummy variable that equals 1 if the state in which the headquarters of the lender is located passed the IDD in a certain year and 0 otherwise. If the IDD was rejected in a precedent lawsuit case, we define the IDD as reverting back to 0 in the state. Therefore, in some states the dummy variable may vary from 0 to 1 and then from 1 to 0, while in other states the dummy variable may remain at 0 or 1 during the whole time period. The outcome of interest, Loan term, includes the 16

18 loan spread (all-in spread drawn, or AISD), 19 the log of the loan maturity, and other terms. The variable ω s is lender state fixed effects, μ s,t is borrower state year fixed effects, φ k is borrower industry fixed effects, and ε i,j,s,s,k,t is the residual term. Note that borrower state year fixed effects control for the effect of IDD passage on the borrower s side. Borrower characteristics includes borrower firm size, book leverage, profitability, tangibility, the current ratio, the market-to-book ratio, and invest grade rating dummy. 20 Loan characteristics includes the loan spread, maturity, loan size, and collateral. The lender and facility uniquely determine each observation. Other variable definitions are listed in Appendix A. The coefficient of interest is β 1, which is the difference-in-differences estimator that gauges the impact of the adoption or rejection of the IDD on Loan term. β 1 captures the change in the loan term in adopting or rejecting states in excess of unaffected states. With this identification strategy, the staggered adoption or rejection of the IDD allows a lender in a given state to be both treatment and control at different times. The effect of IDD adoption on bank loan should go through the channel of trade secrets protection. In other words, residual variations in the IDD (after controlling for all variables) should be uncorrelated with unobservable factors that affect bank lending. This assumption will be undermined if unobserved local economic conditions can simultaneously drive the passage of the IDD and local banks incentives of offering better loan terms. We control for lender state year fixed effects (in addition to borrower state year fixed effects) 19 The AISD is the bank s interest rate spread over the U.S. prime rate or the London Inter-Bank Offered Rate (LIBOR). It is considered drawn because it includes all costs of the loan, such as annual bank fees. Another variable to proxy for loan costs is the all-in spread (undrawn), which does not include costs such as annual bank fees. We choose the AISD as a proper proxy since this variable is available for more loan contract observations. 20 We also reproduced the results in Tables 2 to 6 using the fixed effects of 23 long term rating levels from S&P as a more refined rating control. The results are qualitatively similar to our original results. 17

19 to mitigate this concern. To further control for unobserved local economic conditions, we conduct alternative tests in which we find for each treated state an adjacent state that works as the control group and re-estimate the difference-in-differences regression. One concern is that a law passed in one state will affect people s expectations in another state (possibly the state in which borrowers are located), which in turn affects our outcome variables. However, this argument needs not be applicable in our setting. Tradesecret laws are state laws, which means that only if the local state court accepts the law and sets an effective date will a precedent case exist and can we assume the law is applicable in that state. In addition, different states have very distinct legal systems, which we assume do not converge. Even if such expectations can be formed, it would only lead to an underestimation of the effect, as banks would take action and reduce the loan price prior to the passage of IDD The effect of the IDD on loan contract terms in the primary market Baseline results To test the hypothesis, we implement the difference-in-differences strategy as specified in equation (1). The dependent variable in Table 2 is the loan spread (i.e., AISD). Column (1) shows results of controlling for borrower characteristics, lender fixed effects and year fixed effects. The coefficient on IDD is negative and significant. The magnitude of the coefficient suggests that the passage of the IDD reduces the loan spread by bps for affected banks as compared to unaffected banks. The second column shows results when adding controls for other loan terms (maturity, loan size, and collateral), which yields a smaller coefficient on IDD ( 24.87). For column (3), we further add borrower state fixed effects, and the result remains qualitatively similar ( 18

20 24.48). To mitigate the concern that our result can be affected by a demand-side (borrower) effect, we explicitly control for borrower industry year fixed effects in column (4), and borrower state year fixed effects and borrower industry fixed effects in column (5). The coefficients on IDD remain statistically and economically significant ( in column (4), in column (5)). We next examine whether the passage of the IDD affects loan maturity. The results in Table 3 show that passage of the IDD significantly increases loan maturity. After controlling for lender state fixed effects, borrower industry fixed effects, and borrower state year fixed effects, we find that the maturity of loans from affected banks increases by 11.5%, compared to those from banks unaffected by the IDD in column (5). Since loan terms are simultaneously determined, we also examine the changes in other non-pricing terms, such as loan amounts, covenant strictness (Murfin, 2012), and collateral requirements. Based on untabulated results, changes in these loan terms are all economically and statistically insignificant. In sum, the results reported in Tables 2 to 3 suggest that banks that receive a favorable shock in proprietary information protection are willing to offer better loan terms to borrowers. However, our findings so far do not necessarily rule out the dynamic hold-up in repeated lending, in which banks adopt a low-balling strategy. We will discuss this argument in the next section Incremental effect of relationship lending Note that a reduction in loan spread needs not go against the dynamic hold-up argument. More specifically, banks could adopt a low-balling strategy charging a lower interest rate at the start of the relationship while gradually increasing the rate as the 19

21 relationship continues (or as the lending transactions repeat). This is especially possible for relationship banks. As the IDD protects borrowers soft information acquired in prior lending transactions, a relationship bank gains a higher degree of information monopoly, which can translate into a stronger ability to hold up customers and extract more rent in the future. We test this possibility by investigating the incremental effect of an existing lending relationship. If the reduction in the loan spread is driven by the low-balling strategy of banks, we would expect an increase in loan spreads in relationship loans, since loan spreads should be increasing with the duration or exclusiveness of a lending relationship when hold-up problems exist. 21 Following Bharath et al. (2007), we define a loan as a relationship loan (Rel(dummy) = 1) if two parties of a loan have been engaged in at least one lending deal during the past 5 years. We then split our sample into two subsamples (relationship loans and nonrelationship loans) and re-estimate equation (1) using these two subsamples separately. Panel A of Table 4 presents the results. The dependent variables in columns (1) and (2) are loan spreads. We find that the reduction in the loan spread is larger for relationship loans. More specifically, the passage of the IDD reduces the loan spread by bps for relationship loans, while the reduction in loan spread for non-relationship loans is 14.4 bps. The chi-squared test shows that the pairwise difference in coefficients is statistically significant, with a p-value of 8%. This evidence goes against the dynamic hold-up argument. To the contrary, it implies that the relationship banks, which anticipate a larger benefit 21 If the proprietary information the bank acquires is negative (e.g., zombie borrowers during crisis), we may observe a negative relation between relationship duration and loan spreads (Srinivasan, 2014). In general, being held up by banks would suggest that loan spreads are positively related to the duration of the relationship as we can reasonably assume a small number of zombie borrowers in our sample. 20

22 from an enhanced protection of the acquired information, are willing to share more surplus with the borrower than other kinds of banks. We also examine loan maturities in columns (3) and (4). The coefficients on IDD dummy are positive for both subsamples, but the difference between the two coefficients are statistically similar (with p-value equal to 59%). This implies that the increase in loan maturity is not sensitive to past lending relationships. To substantiate our inference, we further examine the effect of the exclusiveness of a lending relationship. In particular, if a borrower s information is acquired or shared by several banks (not exclusive), the value of surplus stemming from the lending relationship would be diluted. In this respect, banks would have a weaker incentive to offer better loan terms. Thus, we expect a stronger result when the lending relationship is more exclusive. We define the level of exclusiveness in a lending relationship as the loan amount provided by the lender divided by the total loan amount that the borrower received during the past 5 years. This variable also reflects the dependence of a borrower on a particular bank. Then, we split the relationship loans into two groups, based on the median value of this measure, and re-estimate equation (1) using these two subsamples, respectively. The results, reported in Panel B of Table 4, show that the reduction in the loan spread and the increase in maturity are more pronounced for the more exclusive subsample. The chisquared test shows that the coefficients for the two subsamples are significantly different (at the 6% level on spread regressions; at the 0.1% level on maturity regressions). In Panel C, we use an alternative measure of exclusiveness the number of loans provided by the lender divided by the total number of loans that the borrower received during the past 5 years. The results from this alternative specification are qualitatively similar to those in Panel B. 21

23 IDD adoptions and prolonged lending relationship The incentives of relationship banks to share more surplus have to be justified. We argue that, if a relationship bank is willing to share part of the surplus stemming from a better protection of proprietary information, we would expect its clients to choose this bank in future loans at a higher probability, resulting in a reciprocal, long-lasting lending relationship. One approach to examine this assumption is to investigate the effect of IDD on the the likelihood of retaining prior relationship bank in future lending businesses. To this end, we modify the specification of Bharath et al. (2007) by introducing the IDD effect into the model. In particular, we test how IDD affects the likelihood of retaining prior relationship lenders as a lead lender in future syndication loans. When borrowing syndicated loans, firms usually have a choice set that consists of potential lead lenders. Bharath et al. (2007) propose that the lead lenders are mostly from the top 40 biggest syndication loan suppliers in the United States. Therefore, we follow their study and define the potential lead banks, for each loan, as the top 40 biggest syndication loan suppliers in the previous year. Then, we run the following logit regression model: Chosen i,l,j,s,s,k,t = β 0 + β 1 IDD s,t + β 2 Relation j,l,t + β 3 PIDD l,t Relation j,l,t + β j Borrower characteristics j,t + β i Loan characteristics i,t + ω s + μ s t + φ k + ε i,j,s,s,k,t (2) where i indexes the loan, l indexes the potential lender (the 40 biggest loan supplier banks), s indexes the state of the lender, s indexes the state of the borrower, t indexes the year in which the loan was issued, j indexes the borrower of the loan, k indexes the industry of the borrower, PIDD is a dummy variable that equals 1 if the state in which the headquarters of the potential lender is located passed the IDD in year t and 0 otherwise, and the measure of prior lending relationships, Relation, is proxied by either Rel(number) or Rel(amount). We 22

24 are interested in the interaction term between the PIDD and the measures of prior lending relationships. The sample starts with our sets of 10,777 loan observations and 40 potential lenders for each loan. After screening out the observations of potential lenders with missing headquarters state information, we get 137,509 observations for our logit regression. The results reported in Table 5 show that the IDD increases the probability of retaining prior relationship banks as a lead lender in future syndication loans, suggesting that the IDD leads to a prolonged lending relationship The effect of IDD on loan trades in the secondary market We have shown that an enhanced protection on banks proprietary information facilitates relationship lending and therefore reduces the loan price in the primary loan market. However, banks may sell their loans in the secondary market as to improve the asset liquidity. Therefore, as to provide a comprehensive assessment of the importance of protections on banks trade secrets, we examine how IDD adoptions affect the transactions in the secondary loan market. By doing this, we also seek to provide evidence on some specific benefits that banks can obtain via the protection of proprietary information. To this end, we empirically test the effects of IDD adoptions on the loan trade bid-ask spreads and bidding prices in the secondary loan market (Hypotheses 2). Arguably, the moral hazard problem is alleviated via better protection of proprietary information. To the extent that concern over moral hazard in the secondary loan market is resolved, one would expect an increase in bidding prices and a decline in bid-ask spreads in loan transactions. Following Wittenberg-Moerman (2010), we use the average annual bid- 23

25 ask spread and average annual bidding price to test this conjecture. 22 Since we are conditioning the sample on having a loan sale, the number of observations is reduced significantly. Column (1) of Table 6 shows that after the adoption of IDD, loan bid-ask spread is 1.53 cents lower (economic magnitudes are reported as the cents/dollars per $100 of par value). This effect is substantial since it represents a 25.1% (15.6%) decline when benchmarked to the median (mean) bid-ask spread of the full sample. In column (2), we further show that annual bid price is 3.7% higher after the adoption of IDD. 23 In column (3), we find that the bidding frequency increases by 35.4% after the adoption of IDD. Although this result is not significant, it complements our conjecture that the adoption of IDD affects both the loan sale pricing and the volume. These findings are consistent with the notion that better protection of proprietary information, which enables banks to engage in more durable lending relationships, mitigates the moral hazard problem in the secondary loan market, which in turn reduces the bid-ask spread. Buyers are also willing to raise higher bids after the passage of the IDD. This evidence that banks are able to sell loans at higher prices is also indicative of the benefit that banks can obtain via enhanced protection of trade secrets. A related question is whetherbanks are more likely to sell their loans. Following Drucker and Puri s (2009) specification, we test the effect of IDD adoption on loan sales. Using an OLS model, we regress an indicator variable Loan Sold on IDD dummy and various determinants of loan 22 The detailed loan sale information is available after 2000 when the changes in IDD status are only concentrated on two states. The rest of the states are all seen as control groups. 23 In untabulated results, we further employ an event study to pinpoint the effect of IDD adoption on loan trade spread. We defined abnormal bid-ask spread as the difference between the average spread of the event periods (-1 month, +1 month) and that of (-5 month, -1 month). We find that around the announcement months of IDD, abnormal loan spread is cents lower (higher) after IDD adoptions (rejections). In addition, the abnormal bidding price is 9.4% higher (lower) after IDD adoptions (rejections). 24