Zicklin School of Business | Baruch College



Why Do Firms Withhold Loan Contract Disclosures?Edward Xuejun LiZicklin School of BusinessBaruch College - City University of New YorkNew York, NY 10010Edward.Li@baruch.cuny.eduMonica NeamtiuZicklin School of BusinessBaruch College - City University of New YorkNew York, NY 10010Monica.Neamtiu@baruch.cuny.eduTommy Zhiyuan TuZicklin School of BusinessBaruch College - City University of New YorkNew York, NY 10010Zhiyuan.Tu@baruch.cuny.eduPRELIMINARY – PLEASE DO NOT DISTRIBUTE WITHOUT THE AUTHORS’ PERMISSION. COMMENTS WELCOMEWhy Do Firms Withhold Loan Contract Disclosures?Abstract: This study investigates the determinants of borrows’ decision to withhold detailed loan contract disclosures. Our descriptive statistics indicate that almost one in four loan contracts (one in five for loans exceeding 5-percent-of-assets) are not filed with SEC under Regulation S-K’s disclosure mandates. We hypothesize that concessions of control rights to creditors (e.g., more and/or tighter covenants and other loan restrictions) due to frictions in debt contracting incentivize firms to suppress the details of their loan agreements. Using the firm-bank relationship to capture the unobservable probability of having tighter restrictions, we find evidence that borrowers are less likely to disclose more restrictive loan contracts. We also find that, when loans contracts are more likely to include restrictive terms due to borrowers’ lack of bargaining power, such loans are less likely to be filed with SEC. This study contributes to the literature on firms’ compliance with mandatory disclosure requirements and to the debt contracting literature. IntroductionPrivate debt is the largest source of external financing for U.S. firms. Unsurprisingly, contractual terms set forth in private debt agreements are critical to firms’ financial condition and business. Lenders rely on covenants and other restrictions to monitor and influence firms’ behavior (Tung, 2009; Christensen, Nikolaev, and Wittenberg-Morerman, 2016). Shareholders and other stakeholders also use these terms to assess firms’ financial health and act accordingly. For example, when firms are approaching or experience covenant violations, their stock price declines (Leftwich, 1981; Lys, 1984), external financing providers and trade creditors reduce funding (Nini, Smith, and Sufi, 2012; Zhang, 2018), and competitors take more aggressive strategies in product markets (Billett, Esmer, and Yu, 2018). Due to the importance of contractual term information, Regulation S-K of the Securities Act of 1933 mandates that firms file material loan contracts as exhibits with relevant Securities and Exchange Commission (SEC) filings. Nevertheless, prior research highlights the lack of detailed loan contract information (Drucker and Puri, 2008; Christensen and Nikolaev, 2012). For example, Christensen and Nikolaev (2012) find that about 50% of DealScan credit agreements lack covenant data. They argue that, since the great majority of loans contain covenants, this absence indicates the inability of data aggregators to access contract details. Practitioners also express dissatisfaction with available loan contract details (e.g., Morgenson, 2003) and complain against companies for keeping their loan covenants in secrecy (Stone Street Advisors, 2014). Despite a vast literature on firms’ accounting choices to circumvent loan covenants (Armstrong, Guay, and Weber, 2010), surprisingly little is known about firms’ decision to withhold information regarding covenants and other loan restrictions. Our study aims to fill this gap by examining the determinants of firms’ decision to withhold information about the detailed terms of their loan contracts. We argue that frictions in loan contracting (e.g., information asymmetry, conflicts of interest) can create incentives for borrowers to suppress the detailed terms of their loan agreements for several reasons. First, borrowers can withhold information about their covenants due to concerns about how shareholders and other stakeholders could react in anticipation of a covenant violation. When a loan contract is filed with the SEC, outsiders can better assess the ex-ante probability of a covenant violation, but they still face information asymmetry with respect to ex-post contract renegotiations (Sufi, 2015). Borrowers who anticipate a greater probability of future covenant violations may be more likely to withhold loan information to mitigate the possibility that outsiders act on a potential violation that does not ultimately materialize (i.e., a potential violation that is cured through renegotiation). Second, borrowers can suppress loan details when information asymmetry forces them to surrender control rights to lenders to an extent that could be viewed as excessive by shareholders. When they have a weaker relationship with the borrowers, lenders face greater information asymmetry and may find it optimal to set tighter restrictions than what would be warranted by the borrower’s financial condition under symmetric information (Garleanu and Zwiebel, 2009; Prilmeier, 2017). Third, firms can also suppress loan information when they face hold-up problems which give the banks the ability to extract rents (Boot, 2000; Rajan, 1992; Houston and James, 1996). Finally, firms can hold back loan details when there are greater conflicts of interests between lenders and shareholders (Smith and Warner, 1979). In the presence of greater conflicts of interest, lenders are likely to impose more restrictions on a firm and shareholders may perceive the loss of decision rights to creditors due to these restrictions as less favorable.In summary, the previous arguments suggest that, in circumstances when they face tighter restrictions in their lending agreements, borrowers can face disincentives to disclose the detailed terms of their loan agreements. We hypothesize that borrowers are less likely to disclose the detailed terms of a private debt contract when contract terms involve greater concessions of control rights to creditors (i.e., more restrictions on the borrower).We measure firms’ decision to withhold detailed loan contract information by focusing on whether firms comply with Regulation S-K’s requirements to file material loan contracts with their SEC filings. We begin with a comprehensive DealScan sample of loan contracts initiated between 1996 and 2016. Similar to prior research (Nini, Smith and Sufi, 2009), we use text-analysis tools to identify and collect all loan contracts included in 10-K, 10-Q, 8-K, registration statement and tender offer statement filings in the SEC’s Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system. For our final sample of 19,084 loan contracts with non-missing data, we are able to locate 77% of the actual credit agreements in EDGAR. The remaining 23% represent our sample of contracts not filed with SEC. We start our analysis by providing descriptive evidence that the information about contractual details is significantly more limited in those instances where the borrower chooses not to file the actual loan contract with the SEC. We find that information regarding the presence of at least one financial covenant is available from DealScan for 76% of loan contracts filed with SEC, as compared to about 20% of loan contracts not filed. Similarly, information about dividend restrictions (collateral) is present for about 78% and 19% (82% and 41%) of the filed and non-filed loan contracts, respectively. For loan contracts not filed with SEC, DealScan still provides information about the existence of the loan and some general terms (i.e., amount, lender, interest rate), but typically no detailed information about the type, definition or tightness of covenants and other restrictions. This evidence is consistent with our conjecture that a firm’s decision not to file a loan contract amounts to withholding information about contract details. To test our hypothesis, we need to identify firms that face more or tighter restrictions in their loan contracts. One challenge we face in our empirical analyses is that the detailed terms of a private loan contract are often not directly observable to outsiders when a borrower chooses to withhold this information. To overcome this challenge, we rely on the extensive prior theoretical and empirical literature that establishes a link between debt contracting terms and firm-bank lending relationships. Based on the incomplete contracting theory, Garleanu and Zwiebel (2009) suggest that in a setting with information asymmetry lenders impose tighter restrictions on borrowers when the lending relationship is weaker. As relationships deepen, monitoring of borrowers becomes more efficient which, in turn, improves contracting efficiency. Consequently, lenders relax restrictions and borrowers benefit from more favorable borrowing terms (Bharath, et al. 2009; Schenone, 2009; Prilmeier, 2017). Collectively, prior research suggests that firms with lower intensity lending relationships are likely to face more restrictive terms in their loan contracts; as these relationships intensify, borrowers are likely to experience less restrictive contract terms.We start our multivariate analyses by investigating whether firms are more likely to withhold loan contract disclosures when they have lower intensity lending relationships. Following prior research (e.g., Bharath et al. 2009; Prilmeier, 2017), we assess relationship intensity as the proportion of a borrower’s total loan financing provided by the current lender. We find that firms are less likely to file their loan agreements with SEC when they have weak relationships with current lenders. This result is robust to controlling for various firm and industry characteristics. The evidence is consistent with our prediction that firms with weaker lending relationships tend to have tighter contractual restrictions, and thus, are more likely to suppress the detailed terms of their loan contracts. To further support our hypothesis, we use cross-sectional tests based on variation in intensity of contractual restrictions. We expect that borrowers with lower negotiating power (greater hold-up potential) will face less favorable terms and tighter restrictions even when their lending relationships intensify and thereby maintain an incentive to withhold their loan details. In line with our expectation, we find that borrowers with less negotiating power (i.e., smaller, non-rated, higher information asymmetry borrowers) are more likely not to file their loan contracts, even when the lending relationships deepen. We also find that firms with favorable information about their prospects (as measured by changes in future performance) are more likely to disclose their contract terms. To the extent a borrower’s ability to communicate favorable projections to its lenders results in less restrictive lending terms, this evidence provides further evidence consistent with our hypothesis. In sensitivity analyses, we find that our results are robust when we re-estimate our regressions using only larger (i.e., more likely to be material) loan contracts. The results are also robust to using alternative measures of relationship intensity, alternative estimation methods for our probit regressions (including different types of fixed effects), and additional controls for agency costs and for the costs of engaging in non-filing behavior.Our study offers two contributions. First, we contribute to the literature studying how firms comply with mandatory disclosure requirements. Prior studies examine how managerial incentives compromise mandated disclosures in settings like environmental liabilities (Barth et.al, 1997), executive compensation (Robinson et al., 2011), internal control weakness (Rice and Weber, 2012), or permanently reinvested foreign earnings (Ayer et al., 2014). So far, the nature of disclosure compliance with material contract reporting has been little understood, perhaps due to the difficulty of obtaining an objective measure of inadequate contract disclosures. To our knowledge, ours is the first study to document how bank financing and frictions in bank debt contracting influence a firm’s decision to comply with mandatory contract disclosure requirements. Loan contracting has not typically been viewed as having an effect on disclosure; instead, a large body of research focuses on how the features of loan contracts impact managers’ accounting choices (see Armstrong, et al., 2010 for a review). We argue that it is important to understand borrowers’ decision to disclose information about these contract features in the first place.Second, we contribute to the debt contracting literature. A growing stream of research that relies on the detailed terms of the loan contracts to assess the determinants and consequences of firms’ contracting choices (Li, 2010; Li, 2016; Dyreng et al., 2017). Prior studies typically take the availability of detailed information about debt contract terms at face value and eliminate from their analyses observations with missing data. Our results indicate that this missing information may stem from borrowers’ strategic decision to withhold these details and that the intensity of firm–bank relationship is associated with the decision not to file the loan contract. This finding is important for future research because a firm’s relationship with its lenders has been often neglected in studies that rely on the detailed terms of loan contracts.Background and hypothesis development Information about the specific terms of a loan agreement is likely to be of interest to firm outsiders for reasons related to monitoring and private information production. First, lenders play an important role in borrower monitoring (Christensen, Nikolaev and Wittenberg-Moerman, 2016; Tung, 2009) and the specific structure of a loan contract critically affects the extent of bank monitoring (Rajan and Winton, 1995). More and tighter covenants and other restrictions (collateral, borrowing base, performance pricing) mean less slack for borrowers, which leads to more regular lender monitoring and intervention. Through covenants and other loan terms, banks can significantly constrain managers’ discretion with respect to financing, investing and operating decisions and can also affect the composition of boards of directors and CEO retention (Chava and Roberts, 2008; Roberts and Sufi, 2009b; Nini et al., 2009 and 2012; Ferreira, et al., 2018; Akins et al. 2018). Therefore, the detailed information about the ex-ante restrictions included in a loan agreement may help firm outsiders anticipate and act on the important changes in corporate policies brought about by lender monitoring and intervention (Lys, 1984, Nini, Smith, and Sufi, 2012; Zhang, 2018, Billett, Esmer, and Yu, 2018). Second, loan terms are likely to be of interest to outsiders because they can reveal the private information obtained by banks in the lending process. Banks have access to borrower private information, they also have specialized screening and monitoring capabilities and thus, are better equipped than other stakeholders to overcome information asymmetry (Bharath, Sunder and Sunder, 2008). Prior financial intermediation research concludes that a key comparative advantage of banks is their ability to produce private information about borrowers (Boot, 2000). In line with this argument, prior research finds that the details of a loan’s terms can predict future changes in: covenant variables, a firm’s investing and financing policies, outcomes of covenant violations and credit ratings (Demiroglu and James, 2010; Manso et al., 2010; Li, Vasvari and Wittenberg-Moerman, 2016). The information in earnings-based loan covenants can also convey relevant information to equity analysts to improve the accuracy of their earnings forecasts (Gallimberti, Lee, and Lo, 2017).Consistent with the importance of loan term disclosures, Regulation S-K of the Securities Act of 1933 mandates that the SEC registrants file their material loan contracts with relevant SEC filings. The reporting requirements for loan contracts are described in Item 601 “Exhibits” of Regulation S-K (e.g., Item 601(b)(10) require the disclosure of all material contracts). In addition, SEC explicitly names financial covenants as an example of information that could be material to investors (SEC, 1997, 2001).The importance of loan disclosures notwithstanding, information about the detailed terms of a credit agreement is not always publicly available (Morganson, 2003; Christensen and Nikolaev, 2012; Drucker and Puri, 2008). This lack of information suggests that, in certain circumstances, firms may face disincentives to disclose their private loan contracts. We expect the probability of suppressing the details of a loan contract to be associated with the contracting frictions involved in the lending process. In particular, in the presence of incomplete contracts, borrowers may withhold information about their covenants due to concerns about how shareholders and other stakeholders may react in anticipation of potential covenant violations. Advancements in contract theory highlight the fundamental incompleteness of loan contracts (Roberts and Sufi, 2009a). As a result, loan contracts rely on both covenants ex ante and renegotiations ex post to allocate control rights between borrowers and lenders (Sufi, 2015; Nikolaev, 2018). Covenants and other contingencies assign control rights to lenders ex ante, but could lead to ex post inefficient outcomes. Upon realization of the future state of the world, borrowers and lenders engage in Pareto-improving renegotiations ex post to achieve more efficient outcomes. Empirical research provides evidence that lenders often set covenants tightly as “trip wires” (Dichev and Skinner, 2002). However, realized violations are much less frequent as renegotiations help borrowers cure potential violations. Roberts (2015) estimates that a typical private loan is renegotiated five times before its maturity and the majority of renegotiations are initiated by borrowers due to changing conditions, rather than lenders intervening after actual covenant violations. Lenders retain strong control rights when faced with an informational disadvantage and renegotiations can serve as an ex post solution to an ex ante restrictive contract. “It is precisely because borrowers are able to renegotiate the terms of their contracts that they are willing to accept such restrictive contracts in the first place” (Roberts, 2015 p.77).If a loan contract is publicly available (i.e., filed with SEC), shareholders and other stakeholders can rely on the ex-ante covenant information to assess ex post whether firms are approaching technical default. However, information asymmetry prevents outsiders from being fully informed about ex-post renegotiations in anticipation of a violation. Even in the presence of an actual covenant violation, how lenders choose to exercise their control rights at the time of the violation depends not only on the formal (i.e., explicit) terms of the contract, but also on the informal (i.e., implicit) relationship that has been established between lenders and borrowers (Armstrong et al. 2010). The implicit contract between a firm and its lender is less observable to third parties. Outsiders facing uncertainty about the probability of success or even the very existence of renegotiations, could act upon potential covenant violations that may not ultimately realize. These actions could bring extra costs to borrowers’ business and further jeopardize borrowers’ renegotiation prospects. As long as loan contracts cannot specify all contingencies ex ante and outsiders cannot perfectly predict renegotiations, there is always room for such inefficient outcomes. Besides concerns about outsiders’ reactions to potential covenant violations, several other factors could also provide disincentives for loan contract disclosures. Specifically, borrowers may suppress loan details when information asymmetry forces them to give up control rights to lenders to an extent that shareholders perceive as excessive. When they have a weaker relationship with the borrowers, lenders face greater adverse selection costs and may find it optimal to set tighter restrictions than what is warranted under information symmetry. As the lender-borrower relationship intensifies and the lender learns more about the borrower, the need for covenant protection decrease and the restrictions are relaxed (Garleanu and Zwiebel, 2009; Bharath, et al. 2009; Prilmeier, 2017). In addition, hold-up problems may give banks the ability to extract rents from borrowers. The information monopoly the banks generate in the process of lending discourages competition from non-lending banks and could force “locked-in” borrowers to accept non-competitive terms from incumbent banks (Boot, 2000; Rajan, 1992; Houston and James, 1996; Murfin, 2012). Thus, “hold-up” borrowers face incentives to withhold information about non-competitive loan terms. Lastly, there are long-recognized conflicts of interests between lenders and shareholders due to differences in their payoff structures (Smith and Warner, 1979). While bank monitoring can be beneficial when shareholder and lender interests overlap, in the presence of conflicts of interest, bank governance may not always benefit shareholders. For example, for firms with a weaker financial position, creditors would prefer less risk taking than shareholders would. Banks monitor for themselves; they do not have a fiduciary duty to monitor on behalf of shareholders (Tung, 2009). In the presence of larger conflicts of interest, banks could impose more and/or tighter restrictions on borrowers and shareholders could view these restrictions to be less favorable, creating incentives for managers to suppress detailed information about the terms of a loan agreement. To sum up, in the presence of information asymmetry and incomplete contracting, firms may face concerns about stakeholders’ premature actions in anticipation of loan contract restrictions. In addition, contracting frictions (e.g., information asymmetry and hold-up problems) could force borrowers to surrender control rights to lenders to an extent that can be viewed as excessive by shareholders. We hypothesize that borrowers are less likely to disclose the detailed terms of a private loan contract when the contract terms involve greater concessions of control rights to creditors (i.e., more restrictions on the borrower). Consequently, we state our hypothesis as follows (in the alternative form):H1: Borrowers that surrender more decision rights to lenders (via covenants and other loan contract restrictions) are less likely to file a loan contract under Regulation S-K’s requirements.3. Research design and sample selection3.1 Research design and variable measurementTo test our hypothesis, we start by identifying firms that are likely to face more and/or tighter restrictions in their loan contracts. Because the restrictions included in private loan contracts are not always observable to outsiders, we rely on prior theoretical and empirical research on the effect of lending relationships on loan terms (Garleanu and Zwiebel, 2009; Prilmeier, 2017; Bharath, et al. 2009; Schenone, 2009). For borrowers with weaker bank relationships, lenders can impose excessively tight restrictions due to higher information asymmetry. As the bank-firm relationship strengthens, relationship banks face less information asymmetry and loosen the restrictions imposed on borrowers. Therefore, we assess whether firms are more likely to withhold the details of a loan agreement when they have a lower intensity relationship with the bank offering the agreement. Specifically, we estimate the following probit regression (at loan level):Not_Filed = a0 + a1Relation_Intensity + a2DealAmount_Asset + a3Interest_Spread + a4Syndicated + a5#_Prior_Loans_Log + a6Log_Asset + a7Litigation + a8Inst_Own + a9Std_StockRet + a10MTB + a11ROA + a12Loss + a13Leverage + a14Liquidity + a15Not_Rated + a16Tangibility + a17Log_INDSZ + a18Log_ENTRY + Year_FE + Industry_FE + ε (1)where the dependent variable in equation (1), Not_Filed is an indicator variable equal to one for those DealScan loan deals where we cannot find the full-text corresponding loan agreement in borrower’s SEC filings and zero otherwise. Our main regressor of interest in equation (1) is Relation_Intensity.For any given loan, we determine the intensity of the lending relationship (Relation_Intensity) by focusing on the identity of the loan’s lead arrangers. The lead arranger acts as an intermediary between the borrower and the other loan participants and is responsible for negotiating the loan terms (Ivashina, 2009). We focus on all loans obtained by a borrower over a prior five-year window. Following prior research (Bharath et al. 2011; Prilmeier, 2017), we measure Relation_Intensity as the proportion of loan amounts over the previous five years that have been obtained from the current loan’s lead arranger to total loan amounts obtained by the borrower over the past five years. When a loan has more than one lead lender, we calculate a measure of relation intensity for each lead lender and retain only the highest value across all lead banks as our measure of relationship intensity as follows: Relation_Intensity = MAXi(jDeal_Amountj*I(i)jDeal_Amountj) (2)where I(i) is an indicator variable that equals 1 if a lead lender i participates in the deal j, and 0 otherwise. Low values of Relation_Intensity imply that the current loan arranger is one of the many banks financing the firm, whereas higher values imply a more exclusive lending relationship. Lender-borrower information asymmetry is likely to be lower in a more exclusive relation, because the lender has stronger incentives to invest in costly screening and monitoring activities when it does not have to share the benefits of private information production with other lenders. If borrowers have stronger incentives to withhold the details of a loan arrangement when it includes greater restrictions and if lower values of Relation_Intensity proxy for greater restrictions, we expect a negative coefficient on Relation_Intensity (i.e., a1 < 0).To ensure that the non-filing decision is not attributable to immaterial loan contracts, we control for loan materiality measured as the current loan amount scaled by borrower’s total assets (DealAmount_Asset). We also control for loan characteristics (Interest_Spread, Syndicated) that are broadly available for our sample loans regardless of whether a borrower chooses to file the actual loan contract with the SEC. As discussed in the “Descriptive Statistics” section, many other detailed characteristics are much more likely to be observable only when the loan contract is filed with the SEC. Equation (1) also includes firm-level controls intended to proxy for borrower’s size, profitability, credit risk, growth options and level of information asymmetry (#_Prior_Loans_Log, Log_Asset, Inst_Own, MTB, ROA, Leverage, Liquidity, Not_Rated, Tangibility). We also control for characteristics typically viewed as determinants of a firm’s disclosure decision such as litigation risk (Litigation), level of scrutiny (Loss) and proprietary costs (Log_INDSZ, Log_ENTRY) as well as industry and time fixed effects. Appendix A provides detailed variable definition.We also re-estimate equation (1) after replacing the continuous relationship intensity measure with two indicator variables (Relation_High, Relation_Low), to capture any potential non-linearity, as prior research documents a non-linear effect of firm-bank relationship on loan terms (Prilmeier, 2017; Schenone, 2009). Relationship intensity is categorized as “Low” if Relation_Intensity is less than 0.30 and “High” if it is greater than 0.70 (Prilmeier, 2017).3.2. Sample Selection To assess whether registrants’ file their loan contracts with SEC under Regulation S-K’s requirements, we need an independent and relatively comprehensive data source to identify incomplete filing behavior. We take advantage of the loan information collected by LPC, the primary aggregator and disseminator of global private loan market data and use the LPC’s DealScan database as a benchmark against which to assess the completeness of a firm’s loan contract filing. Prior research (Dichev and Skinner, 2002; Chava and Roberts, 2008) argues that only about 60% of the DealScan data come from SEC filings. In addition to collecting loan information included by borrowers in their SEC filings, LPC relies on information self-reported by lenders as well as information from credit industry contacts and from a staff of reporters. We start by collecting all observations with non-missing active date, loan amount and interest spread information (Nini et al. 2009) available in DealScan for the time period 1996-2016. Our sample starts in 1996, because the availability of loan contracts through SEC's EDGAR system is generally scarce prior to 1996. Prior research argues that after 1995, DealScan covers the large majority of sizable commercial loans (Dichev and Skinner, 2002). The sample includes both sole-lender and syndicated loans. DealScan refers to loan agreement as “deals” or “packages”; one “deal” or “package” could consist of one or more “facilities” (e.g., a package could consist of two term loans and one revolving loan). We conduct our empirical analyses at deal (i.e., package) level. This data collection step yields 121,046 deals from the DealScan database. Next, we merge these DealScan observations with the Compustat database using the DealScan-Compustat link data developed by Chava and Roberts (2008) (as updated through the end of 2017). We require each loan deal to have available total asset, long term liability, and book value of equity information in Compustat at the end of the most recent fiscal year prior to the loan initiation date. We only keep the observations that have a valid EDGAR unique firm identifier (i.e., CIK), available from Compustat. This step results in a sample of 43,232 loan deals. We also eliminate loan deals issued to non-US firms and to firms in the financial industries. We also require non-missing market value information from CRSP at the end of the most recent fiscal year prior to deal initiation date. We also exclude all loan deals that do not have at least two years of daily trading records in CRSP following the loan active date. We impose this condition to ensure that we allow firms sufficient time to file their material loan contracts with SEC. Finally, we exclude observations that do not have the financial data necessary to calculate the control variables included in our multivariate analyses. This process results in a baseline sample of 19,084 DealScan loan deals from 4,514 unique companies. We then use the EDGAR electronic filing system to obtain full-text credit agreements for the DealScan loans in our baseline sample. First, we download the 10-K, 10-Q, 8-K, registration statement and tender offer statement filings for the 4,514 CIK corresponding to the unique firms in our DealScan sample. Usually, full-text credit agreements appear as exhibits to a 10-K, 10-Q, 8-K, registration statement (e.g., form S-3, S-4) or tender offer statement (e.g., SC 14D1) filing. Second, we identify the credit agreements from these EDGAR fillings using a text-search approach similar to that used by Nini et al. (2009). Our search terms include the phrases used by Nini et. al. (2009). These search terms are frequently included in the title of credit agreements.After identifying credit agreements from SEC filings, we matched them with the DealScan observations following several steps. First, we match DealScan observations with EDGAR credit agreements by borrower CIK and date of the credit agreement. In addition, we also require the EDGAR credit agreements matched in this first step to have the capitalized term “TABLE OF CONTENTS” within 100 words after the contract title. In this step, we are able to find full-text credit agreements for 7,840 of our 19,084 DealScan observations. Second, for the remaining DealScan contracts, we match by CIK and loan date, but no longer require a “TABLE OF CONTENTS” term to be present in the EDGAR credit agreement. We eliminate this requirement because our reviewing of contracts reveals that a fair number of full-text contracts do not contain a table of contents. In this step, we are able to find credit agreements from EDGAR for 5,340 additional DealScan deals. To validate the quality of matching in the first and second steps, we manually check 200 randomly selected matched observations (100 from the first step plus 100 from the second step) by comparing the deal amount and the name of participating banks in the credit agreements, and we do not find incorrectly matched pairs. Our reviewing of loan contracts further revealed that some loan contracts cannot be matched in the first and second steps because of slight differences between DealScan and EDGAR with respect to the loan initiation date. To account for this issue, for the remaining 5,904 observations that are yet unmatched after the first two rounds, we employ a third step. We first match these remaining DealScan deals with all EDGAR credit agreements that have the same CIK and a deal initiation date within the (-45, +45) day window relative to the loan initiation date as reported in DealScan. Following this broad criterion, we find one or more EDGAR credit agreements for 3,743 out of the 5904 previously unmatched DealScan deals. For this subsample of 3,743 contracts, we manually compare the deal amount and the names of the participating banks in the DealScan deals and credit agreements. We consider a DealScan – EDGAR contract pair to be matched if both the deal amount and participating banks are the same. In this third step, we are able to match 1,311 additional DealScan observations with EDGAR credit agreements. Finally, we match 172 additional loans with full-text agreements by relaxing the assumption that the first date mentioned after the contract title represents the loan initiation date.In summary, across all these rounds, we are able to match 14,663 out of 19,084 deals in our DealScan sample with full-text credit agreements from SEC filings, and we regard the rest of the not-matched deals (4,421 observations) as the credit agreements that companies choose to withhold. Consequently, we label the 14,663 matched deals as the “filed” credit agreements and the remaining 4,421 deals as the “not-filed” contracts. Thus, the “non-filing” rate of credit loans in our sample equals about 23%. Table 1 presents information about our sample selection process.We note that our 77% match rate between DealScan loan deals and actual loan contracts from EDGAR is higher than the 40% match rate reported in prior research (e.g., Nini, Smith and Sufi, 2009). This higher match rate is likely attributable to the fact that our matching methodology is biased in favor of false positives (i.e., instances where a firm’s SEC filings reference a loan contract without including an exhibit for the actual loan agreement). While this approach makes data collection more labor intensive (due to the need for manual checks), it has the advantage that it reduces the probability of false negatives (i.e., instances where a borrower files a loan contract and we fail to identify the contract in EDGAR).4. Empirical findings4.1 Descriptive statisticsWe present descriptive statistics for the variables used in our multivariate analyses in Table 2. Panel A of Table 2 includes descriptive information for the overall sample, while Panel B separately presents information on filed versus not-filed loan contracts. As presented in Table 2, the average loan in our sample amounts to 30% of borrower total assets, with a median value of 19% of total assets. These values suggest that our sample loans are economically significant for the borrowing firms. In cases where borrowers file the loan contract with SEC, the average and median loan amounts are 33% and 23% of total assets, respectively. By comparison, the loans represent a smaller proportion of total assets in instances where borrowers choose not to file the contract; with a mean and median value of 21% and 10%, respectively. Not surprisingly, this evidence suggests that materiality plays a role in the contract filing decision and highlights the importance of controlling for loan size in our multivariate analyses. Table 2 also shows that, for the subsample of loan contracts not filed with SEC, borrowers are larger, slightly more leveraged and more profitable (see Panel B). The continuous variables included in Table 2 are winsorized at 1% and 99% of their respective distributions.Panels C and D of Table 2 present the percentage of not-filed contracts by year and industry, respectively. The yearly non-filing rate in our sample varies from low of 15.57% of all loans in 2015 to a high of 35.39% on 2000. In Table 3, we present descriptive information about non-missing loan terms, separated by subsamples of filed and not-filed loan contracts. We find that, for 76.3% of the observations where the loan contract is filed with SEC, DealScan includes information about at least one financial covenant; this proportion drops to 20.1% for the not-filed contracts. The differences in non-missing information between filed and non-filed contracts are similarly large for other types of loan terms such as: nonfinancial covenants (78.3% vs. 19.5%), performance pricing (57.3% vs. 14.4%) and collateral (81.9% vs. 41.0%). The evidence from this initial descriptive analysis corroborates our conjecture that the availability of information about detailed loan terms is significantly reduced when borrowers choose not to file a loan contract.We note that, even if a borrower chooses not to file the full-text loan agreement as an exhibit under Regulation S-K, the firm could still provide disclosure about covenants and other loan terms in other parts of its SEC filings. For example, a borrower may provide some loan covenant information in its 10-K filings as part of the “Risk factors” discussion, the “Liquidity and capital resources” discussion in the MD&A or a “Long-Term Debt” footnote. While these disclosures can vary from generic to more specific numerical information; most often borrowers include a statement that various loan covenants are in effect, rather than providing details regarding each covenant (see Appendix B for some examples of 10-K covenant disclosures). Even numerical covenant threshold disclosures are not as informative as the full-text loan contract because they lack details about the specific definition of the covenant variables. Prior debt contracting research (Dichev and Skinner, 2002; Li, 2010; Dyreng et al., 2017) documents substantial tailoring of contractual definitions based on borrower-specific circumstances. In this study, we do not rely on the assumption that there is no disclosure of loan term information when the borrower chooses not to file the loan contract. Instead, we rely on the assumption that the information available to firm outsiders is less precise when the loan contract is not filed, and thus, outsiders have a harder time ex-ante assessing when the contract restrictions are binding (e.g., ex-ante assessing the probability of a covenant violation). Table 4 presents Pearson correlations for the variables used in our multivariate analyses. As shown in Table 4, the correlation between the firm-bank relationship intensity and the decision not to file a loan contract is negative and statistically significant (p-value < 0.01). This univariate evidence is consistent with our expectation that borrowers have stronger incentives to withhold the details of a loan at lower levels of relationship intensity, when the loan agreement is more likely to impose greater restrictions on the borrower.4.2. Multivariate regression results 4.2.1. Main resultsIn this section, we explore whether borrowers that face more and/or tighter restrictions in a loan contract are more likely to withhold the detailed terms of the contract. We rely on the intensity of the firm-bank relationship to capture the probability of having more/tighter restrictions. In Table 5, we present probit regressions of the decision not to file a loan contract (Not_Filed) on our proxy of relationship intensity and control variables. As presented in Panel A of Table 5, the coefficient on our main variable of interest (Relation_Intensity) is negative and statistically significant (p-value <0.01). To the extent loan agreements have more restrictive terms when there is a weaker firm-bank relationship, this evidence is consistent with our hypothesis that borrowers are less likely to disclose contracts with more/tighter restrictions. This finding is robust to controlling for the materiality of loan contracts. The coefficient on DealAmount_Asset is also negative and statistically significant at 1% level, suggesting that the greater a loan amount relative to borrower total assets (our proxy for loan materiality) the larger the probability that the contract will be filed. The coefficient on borrower’s size is positive and statistically significant (p-value <0.01). This result may also be explained by materiality considerations, to the extent a new loan is less likely to be material for larger borrowers that have other available sources of financing and a well-established credit history. We also find that for riskier loan deals (as proxied by higher loan spreads), the loan contract is less likely to be filed under Regulation S-K’s requirements. Loans issued to higher market-to-book borrowers are also less likely to be filed. To the extent lenders perceive higher growth borrowers as riskier, these borrowers may face tighter restrictions in their loan agreements and thus, face stronger disincentives to disclose. Loans issued to borrowers with a larger proportion of institutional shareholders and more uncertain information environments (as proxied by greater stock return volatility) are more likely to be disclosed, suggesting that greater demand for information may also play a role in borrowers’ loan contract disclosure decisions. We do not find evidence that the contract non-disclosure decision is associated with proprietary costs as defined in the traditional sense of conveying relevant information to a firm’s product market competitors. The coefficients on our proprietary cost proxies (Log_INDSZ and Log_ENTRY) are not statistically significant at conventional significance levels. This finding is perhaps not surprising in light of prior research arguing that proprietary costs can vary significantly across different types of material contracts and that debt contract disclosures may have a lower level of proprietary cost relative to other types such as purchase/sale agreements or licensing contracts (Verrecchia and Weber, 2006; Li 2013).To help assess the economic significance of our empirical findings, in Table 5, we present regression coefficients in column 1 and marginal effects in column 2. In column 2 of Panel A, the marginal effect of our main regressor of interest on the contract non-disclosure decision is -0.094. Thus, an increase of 0.61 (i.e., 1 - 0.39) from the bottom to the top quartile of the distribution of firm-bank relationship intensity (see Table 2) is associated with a 5.7% (i.e., 0.094*0.61) lower probability of non-disclosure. Relative to the average non-disclosure probability in our sample of 23% (Table 2), this effect represents a reduction of about 25%, suggesting that our finding is also economically significant. In Panel B of Table 5, we present the results of re-estimating equation (1) using a non-linear specification as prior research documents a non-linear effect of firm-bank relationship on loan terms (Prilmeier, 2017; Schenone, 2009). Following Prilmeier (2017), in the specification presented in Panel B, we categorize relationship intensity as “Low” if Relation_Intensity is less than 0.30 and “High” if it is greater than 0.70. Loans with medium relationship intensity represent the base group. For this specification, the coefficient on the Relation_Intensity_Low variable is positive and statistically significant at 1% level. In contrast, the coefficient on the Relation_Intensity_High indicator variable is not statistically significant at conventional levels. These results suggest that loan contracts with low lending relationship intensity have a significantly larger probability of non-disclosure relative to loans with medium relationship intensity, while loans from banks that are borrowers’ main lenders (Relation_Intensity_High=1) are not significantly less likely to be withheld compared with medium relationship intensity loans.4.2.2. Cross-sectional analysesTo provide further support for our hypothesis, we employ the variation in the extent of borrower bargaining power and examine whether borrowers with more negotiating power (i.e., those firms less likely to concede control rights to creditors) are more likely to disclose the details of their loan agreements. While relationship lending can be beneficial to borrowers, a firm in a stronger lending relationship may also face a greater potential for “hold-up” problems where the incumbent bank extracts rents and holds up the customer from receiving competitive loan terms from someone else (Rajan, 1992). On average, borrowers may be able to retain more control rights as the lending relationship intensifies; however, this favorable effect may be mitigated for borrowers that have less bargaining power and face a greater hold-up risk. We expect these hold-up borrowers to maintain an incentive to withhold their loan terms even when the lending relationship intensifies.We use four proxies to identify borrowers with greater hold-up potential. In particular, we conjecture that smaller firms, firms that do not have access to bond markets (as proxied by the absence of a long-term issuer credit rating), firms with a less transparent information environment (as measured by lower financial analyst coverage) and firms with a less well-established reputation in the credit markets (as proxied by shorter credit history) are more likely to experience hold-up problems. As shown in Table 6, when we interact our main variable of interest (Relation_Intensity) with these hold-up proxies, we find a positive and statistically significant coefficient on all four interaction terms. For three of the four measures, the coefficients are significant at 1% level and the last measure is significant at 5% level. We also continue to find a negative and statistically significant coefficient on the main Relation_Intensity term (p-value < 0.01). These results suggest that borrowers experiencing greater hold-up risk (i.e., smaller, non-rated, higher information asymmetry borrowers) are less likely to benefit from more relaxed lending restrictions as the firm-bank relationship intensifies. Thus, these borrowers maintain their incentives to suppress their loan details, even when the lending relationship deepens.In supplemental cross-sectional analyses, we also investigate whether borrowers with less favorable information about their future prospects (as measured by changes in future performance) are less likely to disclose their contract terms. Following prior research on debt contracting (Demiroglu and James, 2010), we use borrower’s changes in sales and return on assets (ROA) from the year before to the year after the loan to identify borrowers with less favorable information. To the extent a borrower’s with less favorable projections about its future prospects is forced to accept more restrictive lending terms or anticipates a greater probability of future covenant violations, the borrower may be more likely to withhold the details of a loan agreement. We present our results in Table 7. In columns 1 and 3, we include our proxies for information about unfavorable prospects (i.e., low future sale growth and changes in ROA) as stand-alone variables. We find some weak evidence that, when borrowers anticipate low sale growth at the date of a loan contract, they are less likely to file such a contract under Regulation S-K requirements. In column 1, the coefficient on sale growth is positive and statistically significant at 10% level. We do not find a statistically significant coefficient on future changes in ROA. This result is in line with Demiroglu and James (2010). They show that the tightness of a loan’s covenant structure is predictive of future sale growth but not of changes in ROA.In columns 2 and 4, we include these future sale growth and changes in ROA terms interacted with Relation_Intensity. As before, the coefficient on the main Relation_Intensity term is negative and statistically significant (p-value < 0.01). The coefficients on the future sale growth and changes in ROA interaction are positive and statistically significant at 1% and 5% level, respectively. These findings are consistent with the notion that, as the lending relationship intensifies, borrowers may benefit from less restrictive terms and thus, be less inclined to suppress the details of the contract. However, this effect of more intense relationships is mitigated for loans issued to borrowers with unfavorable future prospects (lower sale growth and ROA changes). Taken together, these findings provide further support for the hypothesis that contracts with more restrictive terms are less likely to be disclosed. 4.2.3. Sensitivity analyses4.2.3.1 The effect of loan materiality To further understand the impact of loan materiality on our main results, we re-estimate equation (1) using several subsample of deals where we retain progressively larger loans (i.e., the loan amount represents at least 5%, 10%, 15%, 20%, 25% and 30% of the borrower’s total assets). When we impose this relative loan size filter, the sample size drops from 16,317 observations for loans where Dealamoun_Asset > 5% to 6,548 observations for loans where Dealamoun_Asset > 30% (see Panel A of Table 8). The mean value of our dependent variable (Not_Filed), drops from 0.23 for the overall sample to about 0.19 for the subsample where Dealamoun_Asset > 5% to about 14% for the subsample where Dealamoun_Asset > 30%. The fact that the proportion of non-filed contracts decreases when we retain larger loans suggests that the loan amount relative to a borrower’s total assets plays a role in the non-filing decision. This evidence extends our understanding on how managers assess materiality in a setting where auditors are less likely to be involved. Prior research (Gleason and Mills, 2002) finds that 5% of total assets is an important materiality threshold in a context where auditors are involved (i.e., contingent tax liability reporting). In contrast, we find that about 19% of the loans larger than 5% of total assets are not filed by borrowers under Regulation S-K’s requirements to file material contracts. This result suggests that: either managers use substantially different materiality thresholds when auditors are not involved in the decision process or the size of the item under consideration does not play an important role in gauging materiality for a number of loan contracts.Table 8 (Panels B and C) presents the results of our multivariate analysis based on subsamples of progressively larger loans. Consistent with our main results, even for larger loans, we still find that borrowers are less likely to withhold the details of a loan contract as the intensity of the relationship with the lender increases. Similar to the results presented in Table 5 (Panel B), we still find that this result is mostly attributable to loans with low intensity relationships (see Panel C of Table 8). Only when we increase the relative size of loans to greater than 30% of total assets, does the coefficient on Relation_Intensity_Low drop below conventional significance levels. Taken together, these results provide evidence about the strength of our findings in the presence of materiality considerations.4.2.3.2 Alternative relationship intensity measuresIn Table 9 (panel A), we present results from estimating equation (1) after replacing our main Relation_Intensity measure with two alternative proxies. First, as an alternative to our main measure which retains only the lead arranger with the maximum relationship intensity value, we also calculate Relation_Intensity_Weighted as follows (Prilmeier, 2017):Relation_Intensity_Weighted = ∑i(jDeal_Amountj/Nj*I(i)jDeal_Amountj) where Nj denotes the number of lead arrangers for each loan j. Since it is defined as the sum of relationship intensities among all lead lenders, this measure gives equal weight to each lead lender involved with the current loan.Second, we measure the firm-bank relationship as Relation (Duration_log) - the duration of time between the current loan date and the borrower’s first interaction with the current lead arranger (as documented in the DealScan database). Unlike the previous two relationship intensity measures, this variable focuses on a bank’s knowledge of the firm in absolute terms instead of focusing on the current bank’s knowledge relative to that of other banks.As presented in Panel A of Table 9, we find a negative and statistically significant coefficient on relationship when this variable is defined as Relation_Intensity_Weighted and Relation (Duration_log). Thus, when we rely on alternative relationship measures to capture the probability of a loan having more restrictive lending terms, we continue to find that loans contracts with a weaker intensity relationship are less likely to be filed with the SEC consistent with our hypothesis.4.2.3.3 Correlated random effects and loan fixed-effectsWhile in our main tests we estimate the regressions using pool probit models for consistency with prior research, in additional tests, we also assess the sensitivity of our findings to using correlated random effects (CRE) probit estimators. Columns 1 and 3 in Panel B of Table 9 present regression coefficients from specifications that exclude and respectively include loan-type and loan-purpose fixed effects. Columns 2 and 4 include marginal effects. Compared to the pooled probit results in Table 5, employing the CRE estimators leads to somewhat lower p-values suggesting the presence of correlated unobservable firm-specific heterogeneity. This unobservable firm-specific heterogeneity has an effect on some of our control variables (i.e., Size and Inst_Own). More importantly, however, our main variable of interest (Relation_Intensity) remains statistically significant at 1% level using both types of estimators.As shown in column 2 of Panel B, the marginal effect of Relation_Intensity on the contract non-disclosure decision is -0.071. For this specification, an increase of 0.61 (1 - 0.39) from the bottom to the top quartile of the distribution of firm-bank relationship intensity (Table 2) is associated with a 4.33% (0.071*0.61) lower probability of contract non-disclosure. Relative to the average non-disclosure probability in our sample of 23% (Table 2), this effect represents a reduction of 18.8%, suggesting that our results remain economically significant after controlling for unobservable firm-specific heterogeneity.4.2.3.3 Variation in agency costsIn panel C of Table 9, we include supplemental analyses meant to assess to what extent our findings are affected by variation borrower agency problems. We use two proxies for the extent of managerial entrenchment: an indicator variable for CEOs who also hold the President of the board position and an indicator variable for firms with a low proportion of independent directors on the board (calculated based on data from Boardex). After merging our original sample with the Boardex database, the sample size decreases to 11,875 observations.As shown in Panel C of Table 9, we do not find statistically significant coefficients on the managerial entrenchment proxies, when these variables are included in our model either as stand-alone terms or interacted with Relation_Intensity. More importantly, our variable of interest (Relation_Intensity) continues to load at the 1% level, suggesting that our findings are robust to controlling for managerial entrenchment.4.2.3.4 The costs of non-filling loan contractsTo better understand some of the costs associated with contract non-disclosure behavior, we further probe the role of litigation risk and external capital needs as determinants of the decision to withhold a loan contract. In some of our previous analyses (see for Example Table 5), the indicator variable Litigation does not load at conventional significance levels. Given that the Litigation variable heavily relies on industry membership, it is possible that the effect of Litigation is subsumed by industry fixed effects. In Panel D of Table 9, we present results from re-estimating equation (1) after replacing the Litigation variable with an alternative proxy for litigation risk (KS_Litigation_Risk) that is affected less by industry membership. KS_Litigation_Risk defined as the probability of litigation for each firm-year estimated using the litigation risk model (3) in Table 7 of Kim and Skinner (2012). We also add additional controls for future external financing needs (Post_Equity_Issuance and Post_Bond_Issuance). Prior disclosure studies (Barth et al. 1997, Verrecchia and Weber, 2006) find that firms expecting to access external capital markets are more likely to disclose information to avoid the adverse selection costs associated with lack of transparency. Our main results continue to hold when we include these additional controls. In addition, we find that the coefficients on KS_Litigation_Risk and Post_Equity_Issuance are negative and statistically significant (at the 5% level or better). The results in Panels D support the conclusion that litigation risk and the need to access external financing can have a deterrent effect on the decision to withhold loan contract fillings.5. ConclusionAnecdotal evidence and prior academic research highlight the lack of available loan contract details. This evidence is puzzling in light of the importance of loan terms for assessing the extent of lender monitoring (Tung, 2009; Christensen et al., 2016) and in light of the SEC requirements for registrants to file their material contracts under the provisions of Regulation S-K of the Securities Act of 1933. We investigate the determinants of borrowers’ decision to withhold loan contract disclosures and in particular, how firms comply with Regulation S-K’s requirements to file material loan contracts with their SEC filings. We predict that frictions in loan contracting can result in greater concessions of control rights to creditors (e.g., more and tighter covenants and other loan restrictions), thereby creating incentives for borrowers to suppress the detailed terms of their loan agreements.Using the DealScan database as a benchmark against which to evaluate the completeness of contract filings, we are unable to identify a corresponding loan contract filing for about 23% of loan deals reported in DealScan,. Even when we retain only loans that represent at least 5% of a borrower’s total assets, we still find that about 19% of loan deals does not have a corresponding loan contract filed with SEC. Our study provides descriptive evidence that information about contractual details is significantly more limited in those instances where the borrower chooses not to file the actual loan contract with the SEC. We find that information regarding the details of at least one financial covenant is available on DealScan for about three quarters of the contracts filed with SEC vesus about one-fifth of the non-filed contracts.In our multivariate analyses, to capture the unobservable probability that a loan has tighter restrictions, we rely on an extensive theoretical and empirical literature that establishes a link between debt contracting terms and firm-bank lending relationships. We find that firms are less likely to file their loan agreements with SEC when they have weak relationships with the lenders offering the agreement. To the extent loan agreements have more restrictive terms when there is a weaker firm-bank relationship (Garleanu and Zwiebel, 2009; Prilmeier, 2017; Bharath, et al. 2009; Schenone, 2009), this finding suggests that borrowers are less likely to disclose contracts with more/tighter restrictions. We contribute to the literature on firms’ compliance with mandatory disclosure requirements and also to the debt contracting literature. ReferencesAkins, B., De Angelis, D., & Gaulin, M. (2018). Debt contracting on management, Working paper, Rice University. Armstrong, C. S., Guay, W. R., & Weber, J. P. (2010). The role of information and financial reporting in corporate governance and debt contracting. Journal of Accounting and Economics, 50(2-3), 179-234.Ayers, B. C., Schwab, C. M., & Utke, S. (2014). Noncompliance with mandatory disclosure requirements: The magnitude and determinants of undisclosed permanently reinvested earnings. The Accounting Review, 90(1), 59-93.Ball, R., Bushman, R. M., & Vasvari, F. P. (2008). The debt‐contracting value of accounting information and loan syndicate structure. Journal of accounting research, 46(2), 247-287.Barth, M. E., McNichols, M. F., & Wilson, G. P. (1997). Factors influencing firms' disclosures about environmental liabilities. Review of Accounting Studies, 2(1), 35-64.Bharath, S. T., Sunder, J., & Sunder, S. V. (2008). Accounting quality and debt contracting. The Accounting Review, 83(1), 1-28.Bharath, S. T., Dahiya, S., Saunders, A., & Srinivasan, A. (2009). Lending relationships and loan contract terms. The Review of Financial Studies, 24(4), 1141-1203.Billett, M. T., Esmer, B., & Yu, M. (2018). Creditor control and product-market competition. Journal of Banking & Finance, 86, 87-100.Bird, A., Karolyi, S. A., & Ruchti, T. G. (2018). Information Sharing, Holdup, and External Finance: Evidence from Private Firms. The Review of Financial Studies, forthcoming.Boot, A. W. (2000). Relationship banking: What do we know? Journal of financial intermediation, 9(1), 7-25.Chava, S., & Roberts, M. R. (2008). How does financing impact investment? The role of debt covenants. The Journal of Finance, 63(5), 2085-2121.Christensen, H. B., & Nikolaev, V. V. (2012). Capital versus performance covenants in debt contracts. Journal of Accounting Research, 50(1), 75-116.Christensen, H. B., Nikolaev, V. V., & Wittenberg-Moerman R. (2016). Accounting information in financial contracting: The incomplete contract theory perspective. Journal of accounting research, 54(2), 397-435.Colonnello, S., Koetter, M., & Stieglitz, M. (2019). Benign neglect of covenant violations: Blissful banking or ignorant monitoring? Working paper.Demiroglu, C., & James, C. M. (2010). The information content of bank loan covenants. The Review of Financial Studies, 23(10), 3700-3737.Dichev, I. D., & Skinner, D. J. (2002). Large–sample evidence on the debt covenant hypothesis. Journal of accounting research, 40(4), 1091-1123.Drucker, S., & Puri, M. (2008). On loan sales, loan contracting, and lending relationships. The Review of Financial Studies, 22(7), 2835-2872.Dyreng, S. D., Vashishtha, R., & Weber, J. (2017). Direct evidence on the informational properties of earnings in loan contracts. Journal of Accounting Research, 55(2), 371-406.Ferreira, D., Ferreira, M. A., & Mariano, B. (2018). Creditor control rights and board independence. The Journal of Finance, 73(5), 2385-2423.Gallimberti, C., L.F. Lee, and A. Lo. (2017). Equity market use of loan market information: Evidence from analysts’ forecast revisions around loan disclosures. Working paper, Boston College.Garleanu, N., & Zwiebel, J. (2008). Design and renegotiation of debt covenants. The Review of Financial Studies, 22(2), 749-781.Gleason, C. A., & Mills, L. F. (2002). Materiality and contingent tax liability reporting. The Accounting Review, 77(2), 317-342.Gorton, G., & Winton, A., 2003. Financial intermediation. In: Constantinides, G., Harris, M., Stulz, R. (Eds.), Handbook of the Economics of Finance 1., Elsevier, North Holland, Amsterdam, 431–552.Gomes, A., & G. Phillips. 2012. Why do public firms issue private and public securities? Journal of Financial Intermediation 21 (4), 619–658. Houston, J., & James, C. (1996). Bank information monopolies and the mix of private and public debt claims. The Journal of Finance, 51(5), 1863-1889.Ivashina, V. (2009). Asymmetric information effects on loan spreads. Journal of financial Economics, 92(2), 300-319.Karuna, C. (2007). Industry product market competition and managerial incentives. Journal of accounting and economics, 43(2-3), 275-297.Kim, I., & Skinner, D. J. (2012). Measuring securities litigation risk. Journal of Accounting and Economics, 53(1-2), 290-310.Kysucky, V., & Norden, L. (2015). The benefits of relationship lending in a cross-country context: A meta-analysis. Management Science, 62(1), 90-110.Leftwich, R. (1981). Evidence of the impact of mandatory changes in accounting principles on corporate loan agreements. Journal of Accounting and Economics, 3(1), 3-36.Li, N. (2010). Negotiated measurement rules in debt contracts. Journal of Accounting Research, 48(5), 1103-1144.Li, E. X. (2013). Revealing future prospects without forecasts: The case of accelerating material contract filings. The Accounting Review, 88(5), 1769-1804.Li, N. (2016). Performance Measures in Earnings‐Based Financial Covenants in Debt Contracts. Journal of Accounting Research, 54(4), 1149-1186.Lys, Thomas. “Mandated Accounting Changes and Debt Covenants: The Case of Oil and Gas Accounting.” Journal of Accounting and Economics 6 (1984): 39–65.Manso, G., Strulovici, B., & Tchistyi, A. (2010). Performance-sensitive debt. The Review of Financial Studies, 23(5), 1819-1854.Morganson, G. “Why the Secrecy About Financial Covenants?” Market Watch - New York Times, October 12, 2003.Murfin, J. (2012). The supply‐side determinants of loan contract strictness. The Journal of Finance, 67(5), 1565-1601.Nikolaev, V. (2018). Scope for renegotiation in private debt contracts. Journal of Accounting and Economics, 65(2-3), 270-301.Nini, G., Smith, D. C., & Sufi, A. (2009). Creditor control rights and firm investment policy. Journal of Financial Economics, 92(3), 400-420.Nini, G., Smith, D. C., & Sufi, A. (2012). Creditor control rights, corporate governance, and firm value. The Review of Financial Studies, 25(6), 1713-1761.Press, E., & Weintrop, J. (1991). Financial statement disclosure of accounting-based debt covenants. Accounting Horizons, 5(1), 64.Prilmeier, R. (2017). Why do loans contain covenants? Evidence from lending relationships. Journal of Financial Economics, 123(3), 558-579.Rajan, R. G. (1992). Insiders and outsiders: The choice between informed and arm's‐length debt. The Journal of finance, 47(4), 1367-1400.Rajan, R., & Winton, A. (1995). Covenants and collateral as incentives to monitor. The Journal of Finance, 50(4), 1113-1146.Rice, S. C., & Weber, D. P. (2012). How effective is internal control reporting under SOX 404? Determinants of the (non‐) disclosure of existing material weaknesses. Journal of Accounting Research, 50(3), 811-843.Roberts, M. R., & Sufi, A. (2009a). Financial contracting: A survey of empirical research and future directions. Annual Review Financial Economics, 1(1), 207-226.Roberts, M.R., Sufi,A., (2009b). Control rights and capital structure: an empirical investigation. Journal of Finance 64, 1657–1695.Robinson, J. R., Xue, Y., & Yu, Y. (2011). Determinants of disclosure noncompliance and the effect of the SEC review: Evidence from the 2006 mandated compensation disclosure regulations. The Accounting Review, 86(4), 1415-1444.Schenone, C. (2009). Lending relationships and information rents: Do banks exploit their information advantages? The Review of Financial Studies, 23(3), 1149-1199.SEC’s Division of Corporation Finance. (1997). Staff Legal Bulletin No. 1, dated February 28,1997 as amended by the Addendum, dated July 11, 2001. Accessed at: , A., & Vishny, R. W. (1997). A survey of corporate governance. Journal of Finance, 52(2), 737-783.Smith, C. W., & Warner, J. B. (1979). On financial contracting: An analysis of bond covenants. Journal of financial economics, 7(2), 117-161.Stone Street Advisors. (2014). Unilife Investors Kept in The Dark. Accessed at: , F. (2009). Leverage in the board room: The unsung influence of private lenders in corporate governance. UCLA Law Review, 57, 115.Verrecchia, R. E., & Weber, J. (2006). Redacted disclosure. Journal of Accounting Research, 44(4), 791-814.Wooldridge, J.M., (2002). Econometric Analysis of Cross Section and Panel Data. Cambridge: MIT Press.Zhang, Z. (2018). Bank interventions and trade credit: evidence from debt covenant violations. Journal of Financial and Quantitative Analysis, 1-59.Appendix AVariable Definitions:Variable NameDefinitionData SourceNot_FiledAn indicator variable equal to 1 if a Dealscan contract is not matched by a loan contract filed in EDGAR and 0 otherwise,DealScan, EDGARRelation_IntensityProportion of a borrower’s total loan financing provided by the current loan’s lead lender, defined as the total dollar amount borrowed from the current loan’s lead arranger in the past 5 years (including the current loan amount) divided by the total dollar amount of all loans the borrower has obtained over the past 5 years (including the current loan). When a loan has more than one lead lender, we calculate a measure of relation intensity for each lead lender and retain only the highest value across all lead banks, DealScan Relation_Intensity_HighAn indicator variable equal to 1 if the value of Relation_Intensity is greater than 0.7, and 0 otherwise (Prilmeier, 2017), DealScan Relation_Intensity_LowAn indicator variable equal to 1 if the value of Relation_Intensity is lower than 0.3, and 0 otherwise (Prilmeier, 2017), DealScan Relation_Intensity_WeightedSum of weighted relation intensities across all of the current loan’s lead lenders, defined as the total dollar amount (divided by the number of lead banks) borrowed from all of the current loan’s lead arrangers in the past 5 years (including the current loan amount) divided by the total dollar amount of all loans the borrower has obtained over the past 5 years (including the current loan), DealScan Relation_Intensity_Duration_LogLog of the number of years elapsed since the borrower obtained its first loan from the current loan’s lead lender (calculated based on the borrower’s entire history available in DealScan), DealScan First_LoanAn indicator variable equal to 1 if the current the loan is the first loan a company has in DealScan during the most recent 5 years, and 0 otherwise, DealScan SyndicatedAn indicator variable equal to 1 for syndicated loans, and 0 otherwise, DealScanDealamount_AssetCurrent loan amount divided by total assets measured at the end of the most recent fiscal year before the loan date,DealScan, CompustatInterest_Spread"All-in-spread drawn”, defined as the spread over LIBOR on the drawn amount plus fees, divided by 100, DealScan #_Prior_Loans_LogNatural log of 1 plus the number of previous loans obtained by the borrower (calculated based on the borrower’s entire history available in DealScan) that are still outstanding,DealScan Log_AssetNatural log of total assets for the most recent fiscal year ended before the current loan date, CompustatLitigationAn indicator variable equal to 1 if the borrower’s SIC code falls in the following groups: 3570-3577, 7370-7374, 3600-3674, 2833-2836, 8731-8734, or 5200-5961, and 0 otherwise, CompustatKS_Litigation_RiskThe probability of litigation for each firm-year estimated using the litigation risk model (3) in Table 7 of Kim and Skinner (2012)Compustat/CRSPInst_OwnPercentage of shares outstanding that are owned by institutional investors, measured based on the most recent data available before the current loan date, Thomson Reuters (13f) HoldingsStd_stockretStandard deviation of monthly stock returns measured over the past 12 months before the current loan date, CRSPMTBMarket to book ratio, calculated as the market value of the borrower's shares outstanding plus the book value of debt divided by the book value of total assets, CompustatROAReturn on assets, calculated as income before extraordinary items divided by total assets, CompustatLossAn indicator variable equal to 1 if company's income before extraordinary items is negative, and 0 otherwise,CompustatLeverageLong-term debt plus debt in current liabilities divided by total assets, CompustatLiquidityCash and cash equivalents divided by total assets,CompustatNot_RatedAn indicator variable equal to 1 if a borrower does not have a S&P long-term issuer credit rating, and 0 otherwise, CompustatTangibilityProperty, plant, equipment divided by total assets, CompustatLog_INDSZMarket size in the borrower’s industry (four-digit SIC code), defined following Karuna, (2007) as the natural log of industry sales in millions, CompustatLog_ENTRYCost of entering borrower’s industry (four-digit SIC code), defined following Karuna (2007) as the natural log of the weighted average of property, plant and equipment in millions (where the weight is each firm's market share in the industry), CompustatSmall_BorrowerAn indicator variable equal to 1 if the borrower's total assets are below the median value of all sample companies during the same year, and 0 otherwise, CompustatAnalyst_LowAn indicator variable equal to 1 if the number of financial analysts following the borrower is below the median value of all sample companies during the same year, and 0 otherwise, I/B/E/SShort_Credit_HistoryAn indicator variable equal to 1 if the borrower’s number years available in the DealScan history is below the median value of all sample companies during the same year, and 0 otherwise, DealScan Sales_Growth_p1n1_LowAn indicator variable equal to 1 if the borrower’s percentage change in sales from year -1 to year +1 relative to the loan year (year 0) is below the median value of all sample companies during the same year, and 0 otherwise, CompustatOI_AT_Change_p1n1_Low An indicator variable equal to 1 if the borrower’s change in operating income over total assets from year -1 to year +1 relative to the loan year (year 0) is below the median value of all sample companies during the same year, and 0 otherwise, CompustatCEO_Dual An indicator variable equal to 1 if the borrower’s CEO also serves as the president of the board of directors, and 0 otherwise, BoardExBoard_Indep_LowAn indicator variable equal to 1 if borrower’s proportion of independent directors is below the median value of all sample companies during the same year, and 0 otherwise. BoardExPost_Equity_IssuanceAn indicator variable equal to 1 if the borrower issues new equity in the year after the current loan date, and 0 otherwise.SDC Post_Bond_IssuanceAn indicator variable equal to 1 if the borrower issues new bonds in the year after the current loan date, and 0 otherwise.SDC Appendix B – Examples of covenant disclosures in SEC filingsDONNKENNY INC (CIK=29693) - 10-K Form filed on 1999-03-31 6. LONG-TERM DEBTThe Credit Facility contains numerous financial and operational covenants, including limitations on additional indebtedness, liens, dividends, stock repurchases and capital expenditures. In addition, the Company is required to maintain specified levels of tangible net worth and comply with a maximum cumulative net loss test.EMPIRE RESOURCES INC (CIK=1019272) - 10-K Form filed on 2006-03-30MANAGEMENT’S DISCUSSION AND ANALYSIS OF FINANCIAL CONDITION AND RESULTS OF OPERATIONSLiquidity and Capital Resources (in thousands)?We currently operate under a revolving line of credit, including a commitment to issue letters of credit, with four commercial banks. The maximum availability of this facility is $90,000. Borrowings under these lines of credit are collateralized by security interests in substantially all of our assets. Under these credit agreements, we are required to maintain working capital and net worth ratios. These facilities expire on June 30, 2006. As of December 31, 2005, the amount outstanding under our revolving lines of credit was $87,677 (including letters of credit of approximately $3,177). Management is in discussions to extend the maturity, increase the value of the overall facility, and revise some of the covenants and terms of the Credit Facility. We expect to complete this transaction prior to the expiration of the current facility. ATP OIL & GAS CORP (CIK=1123647) - 10-K Form filed on 2005-03-31Item?1A. Risk FactorsOur debt instruments impose restrictions on us that may affect our ability to successfully operate our business. ?In March 2004, we entered into a new term loan, which was subsequently amended in September 2004 (the “Term Loan”), consisting of a $185.0 million Senior Secured First Lien Term Loan Facility and a $35.0 million Senior Secured Second Lien Term Loan Facility. The Term Loan matures in March 2009 and is secured by substantially all of our oil and gas assets in the Gulf of Mexico and the U.K. Sector – North Sea and is guaranteed by our wholly owned subsidiaries ATP Energy and ATP Oil & Gas (U.K.) Limited. As of December 31, 2004, we had $218.4 million principal amount outstanding under the Term Loan. The Term Loan contains customary restrictions, including covenants limiting our ability to incur additional debt, grant liens, make investments, consolidate, merge or acquire other businesses, sell assets, pay dividends and other distributions and enter into transactions with affiliates. We also are required to maintain specified financial requirements under the terms of our Term Loan including the following, as defined in the Term Loan: ????Current Ratio of 1.0/1.0; ??Consolidated Net Debt to EBITDAX coverage ratio which is not greater than 3.25/1.0 through June 30, 2004 and 3.0/1.0 at each of the quarters ending thereafter; ??Consolidated EBITDAX to Interest Expense which is not less than 2.5/1.0 for any four consecutive fiscal quarters commencing with the quarter ended June 30, 2004 and at each of the quarters ending thereafter; ??PV10 of our Total Proved Developed Producing Oil and Gas Reserves to Net Debt of at least 0.5/1.0 at June 30 and December 31 of any fiscal year; ??PV10 of our Total Proved Oil and Gas Reserves to Net Debt of at least 2.5/1.0 at June 30 and December 31 of any fiscal year; ??Net Debt to Proved Developed Oil and Gas Reserves of less than $2.50/Mcfe at December 31, 2004 and at each of the years ending thereafter, and ??the requirement to maintain hedges on no less than 40% of the next twelve months of forecasted production attributable to our proved producing reserves. Table 1. Sample Selection　Description# of Credit Agreements1Start with all DealScan loan agreements initiated during 1996-2016 that have available loan amount and interest spread information.121,0462Merge with Compustat using the DealScan-Compustat link file provided by Micheal Roberts (Chava and Roberts, 2008), require each loan contract to have non-missing total assets, long-term liabilities and book value of equity available from Compustat for the most recent fiscal year before the loan initiation date.(77,814)3Retain loan agreements only for US non-financial companies; require each loan contract to have non-missing market value information available from CRSP at the end of the most recent fiscal year before the loan initiation date. (22,209)4Require each loan observation to have non-missing data for all the control variables used in our multivariate analyses. (1,939)　Baseline Sample:19,084Table 2. Descriptive StatisticsPanel - A Full Sample:MeanStd. Dev.P25MedianP75NLoan-Level Variables:Not_Filed0.23 0.42 0.00 0.00 0.00 19,084 Relation_Intensity0.66 0.32 0.39 0.71 1.00 19,084 Dealamount_Asset0.30 0.33 0.08 0.19 0.39 19,084 Interest_Spread1.95 1.40 1.00 1.75 2.75 19,084 Syndicated0.79 0.41 1.00 1.00 1.00 19,084 Company-Level Variables:#_Prior_Loans2.41 2.60 1.00 2.00 3.00 19,084 Asset5,291.33 11,322.26 267.60 1,040.02 4,107.39 19,084 Litigation0.22 0.41 0.00 0.00 0.00 19,084 Inst_Own0.58 0.29 0.36 0.63 0.82 19,084 Std_Stockret0.12 0.07 0.07 0.10 0.15 19,084 MTB1.81 1.08 1.15 1.47 2.06 19,084 ROA0.03 0.11 0.01 0.04 0.07 19,084 Loss0.21 0.41 0.00 0.00 0.00 19,084 Leverage_Ratio0.28 0.20 0.14 0.27 0.40 19,084 Liquidity_Ratio0.10 0.13 0.01 0.05 0.13 19,084 Not_Rated0.50 0.50 0.00 1.00 1.00 19,084 Tangibility0.34 0.25 0.13 0.27 0.51 19,084 Log_INDSZ9.63 2.02 8.17 9.72 10.96 19,084 Log_ENTRY7.56 2.06 6.02 7.48 9.14 19,084 Panel A of this table presents descriptive statistics for the variables used in our multivariate analysis. The loan level variables are measured at the initiation date of each DealScan loan agreement. The company-level variables are measured at the end of the most recent fiscal year prior to the loan agreement initiation date. See Appendix A for variable definitions.Table 2. Descriptive Statistics (continued)Panel B - Sub-sampling by Not-Filed vs. Filed ContractsNot-Filed Contracts (4,421 Contracts)Filed Contracts (14,663 Contracts)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)MeanStd. Dev.P25MedianP75MeanStd. Dev.P25MedianP75Diff in MeansRelation_Intensity0.58 0.35 0.25 0.59 1.00 0.69 0.31 0.44 0.74 1.00 -0.11***Dealamount_Asset0.21 0.31 0.04 0.10 0.25 0.33 0.33 0.11 0.23 0.43 -0.12***Interest_Spread1.80 1.56 0.50 1.43 2.55 2.00 1.34 1.00 1.75 2.75 -0.21***Syndicated0.83 0.38 1.00 1.00 1.00 0.78 0.41 1.00 1.00 1.00 0.05***#_Prior_Loans3.05 3.45 1.00 2.00 4.00 2.22 2.24 1.00 2.00 3.00 0.83***Asset9,557.77 15,409.16 596.90 2,557.48 11,165.30 4,004.97 9,387.49 226.00 818.80 2,950.69 5552.80***Litigation0.20 0.40 0.00 0.00 0.00 0.22 0.41 0.00 0.00 0.00 -0.02**Inst_Own0.58 0.26 0.41 0.63 0.78 0.58 0.30 0.34 0.64 0.83 0Std_Stockret0.11 0.07 0.06 0.09 0.14 0.13 0.08 0.07 0.11 0.16 -0.02***MTB1.84 1.08 1.18 1.49 2.10 1.80 1.07 1.14 1.47 2.05 0.04*ROA0.04 0.09 0.02 0.04 0.08 0.02 0.11 0.00 0.04 0.07 0.01***Loss0.16 0.37 0.00 0.00 0.00 0.23 0.42 0.00 0.00 0.00 -0.07***Leverage_Ratio0.30 0.19 0.17 0.29 0.41 0.28 0.20 0.12 0.26 0.40 0.02***Liquidity_Ratio0.08 0.12 0.01 0.04 0.10 0.10 0.13 0.02 0.05 0.13 -0.01***Not_Rated0.36 0.48 0.00 0.00 1.00 0.55 0.50 0.00 1.00 1.00 -0.19***Tangibility0.35 0.24 0.15 0.30 0.54 0.33 0.25 0.12 0.26 0.50 0.02***Log_INDSZ10.03 1.92 8.76 10.16 11.29 9.51 2.03 8.04 9.59 10.89 0.52***Log_ENTRY8.04 1.96 6.60 8.22 9.56 　7.42 2.06 5.89 7.30 8.95 　0.62***Panel B of this table presents the descriptive statistics for two sub-samples: not-filed and filed loan agreements. The descriptive statistics of the not-filed loan agreements are tabulated in column 1-6, and the statistics for the filed loan agreements are tabulated in column 6-10. Column 11 presents the difference in mean values of each variable calculated as the not-filed minus the filed loan agreement values. ***, **, and * denote statistical significance at 1%, 5% and 10% levels, respectively. See Appendix A for variable definitions.Table 2. Descriptive Statistics (continued)Panel C - Percentage of Not-filed Credit Agreements by Year:(1)(2)Year% of Not-filed Credit Agreements # of Credit Agreements 199624.49%1,021 199724.65%1,448 199825.22%1,142 199928.57%1,071 200035.39%1,085 200130.50%1,128 200226.18%1,119 200324.15%1,064 200420.66%1,118 200521.73%1,031 200623.87%909 200722.42%883 200818.50%573 200917.87%414 201017.09%597 201116.76%877 201219.58%669 201320.55%793 201417.46%773 201515.57%732 201615.70%637Panel C of this table presents the percentage of loan contracts that are not filed with the SEC by year (column 1) and the total number of loan contracts (column 2) initiated each year during our sample period. Table 2. Descriptive Statistics (continued)Panel D - Percentage of Not-Filed Contracts by Industry:(1)(2)　Industry % of Not-filed Credit Agreements # of Credit Agreements1Consumer Non-Durables25.37%1,289 2Consumer Durables21.97%569 3Manufacturing21.80%2,647 4Oil, Gas, and Coal Extraction and Products20.41%1,514 5Chemicals and Allied Products28.92%619 6Business Equipment21.04%2,500 7 Telephone and Television Transmission27.43%773 8Utilities37.56%1,539 9Wholesale, Retail, and Some Services20.55%3,153 10Healthcare, Medical Equipment, and Drugs23.87%1,416 12Other Industries19.64%3,065 Panel D of this table presents the percentage of loan contracts that are not filed (column 1) and the total number of loan contracts (column 2) initiated in each of the industries included in our sample. The industry definitions are based on the Fama-French 12 industry classifications. Table 3. Frequency of Non-Missing Contract Terms in Not-Filed versus Filed Loan Contracts　　(1)(2)(3)　Not-Filed ContractsFiled ContractsDiffFinancial Covenants:Performance Covenant16.20%65.40%-0.49***Capital Covenant11.60%41.10%-0.30***Any Financial Covenants:20.10%76.30%-0.56***Non-Financial Covenants:Asset Sales Sweep7.89%31.10%-0.23***Debt Issuance Sweep7.17%27.40%-0.20***Equity Issuance Sweep6.38%24.30%-0.18***Excess Cash Flow Sweep5.97%20.70%-0.15***Capital Expenditure Restrictions3.19%15.00%-0.12***Dividend Restrictions19.00%77.60%-0.59***Any Non-Financial Covenants:19.50%78.30%-0.59***Other Contract Terms:Performance Pricing14.40%57.30%-0.43***Collateral Requirement41.00%81.90%-0.41***Borrowing Base4.89%17.70%-0.13***Maturity90.30%98.10%-0.08***Lead Lender Share17.40%40.20%-0.23***　Any Other Contract Terms:99.80%100.00%-0.00***Observations:4,42114,663　This table presents the frequency of non-missing contract terms for the not-filed and the filed credit agreements. Column 1 presents the percentage of the not-filed contracts with non-missing contract term information, and Column 2 presents the percentage of the filed contracts that have non-missing contract term information. Column 3 presents the difference between the percentages of the not-filed and the filed contracts. ***, **, and * denote statistical significance at 1%, 5% and 10% level, respectively. Contract terms are defined as non-missing if the corresponding cell in DealScan is not marked as missing. Following Christensen and Nikolaev (2012), financial covenants are classified into performance covenants and capital covenants. Performance covenants are classified as non-missing for a contract if at least one of the following terms is not missing in DealScan: (1) Cash interest coverage ratio, (2) Debt service coverage ratio, (3) Level of EBITDA, (4) Fixed-charge coverage ratio, (5) Interest coverage ratio, (6) Ratio of debt to EBITDA, and (7) Ratio of senior debt to EBITDA. Capital covenants are classified as non-missing if at least one of the following terms is not missing: (1) Quick ratio, (2) Current ratio, (3) Debt-to-equity ratio, (4) Loan-to-value ratio, (5) Ratio of debt to tangible net worth ratio, (6) Leverage ratio, (7) Senior leverage ratio, (8) Net Worth, and (9) Tangible Net Worth. Following Prilmeier (2017), the non-financial covenants include: (1) Asset Sales Sweep, (2) Debt Issuance Sweep, (3) Equity Issuance Sweep, (4) Excess Cash Flow Sweep, (5) Capital Expenditure Restriction, and (6) Dividend restriction. The Lead lender share variable, defined as the fraction of the loan amount retained by the lead lender in a loan syndicate (Ball, Bushman, and Vasvari, 2008), is classified as non-missing if this variable can be calculated with DealScan data.Table 4. Pearson Correlations\sThis table presents Pearson correlations for the 19,084 DealScan loan contracts included in our baseline sample. The sample consists of 4,421 not-filed credit agreements and 14,663 filed credit agreements. Correlations that are statistically significant at 1% level are presented in bold font. See Appendix A for variable definitions.Table 5. Determinants of the Not-filing DecisionPanel A - Relation_Intensity – Continuous specification(1)(2)Relation_Intensity-0.345***-0.094***　(-7.08)(-7.13)Dealamount_Asset-0.243***-0.066***(-3.73)(-3.74)Interest_Spread0.104***0.028***(8.13)(8.22)Syndicated-0.046-0.013(-1.19)(-1.19)#_Prior_Loans_Log-0.048-0.013(-1.52)(-1.52)Log_Asset0.207***0.056***(11.67)(11.85)Litigation-0.010-0.003(-0.15)(-0.15)Inst_Own-0.270***-0.074***(-4.24)(-4.25)Std_Stockret-1.567***-0.428***(-6.09)(-6.10)MTB0.064***0.017***(4.20)(4.22)ROA0.1210.033(0.72)(0.72)Loss-0.109**-0.030**(-2.49)(-2.49)Leverage_Ratio-0.036-0.010(-0.38)(-0.38)Liquidity_Ratio0.1710.047(1.22)(1.22)Not_Rated0.0120.003(0.29)(0.29)Tangibility-0.150-0.041(-1.46)(-1.46)Log_INDSZ-0.013-0.003(-0.64)(-0.64)Log_ENTRY0.0320.009(1.57)(1.57)Constant-1.727***(-5.13)Year Fixed EffectYESYESIndustry Fixed EffectYESYESN19,084 19,084 pseudo R-sq0.1070.107Panel A of this table presents results from regressions of borrowers’ decision to withhold a loan agreement by not filing it with the SEC (Not-filed indicator) on the firm-bank relationship intensity (Relation_Intensity) and control variables. Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. The coefficients of this probit model are presented in column 1, and the marginal effects are reported in column 2. In parentheses, we present z-statistics. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 5. Determinants of the Not-filing Decision (continued) Panel B - Relation_Intensity_Low vs. Relation_Intensity_High(1)(2)Relation_Intensity_Low0.313***0.085***　(8.87)(8.91)Relation_Intensity_High-0.009-0.002　(-0.30)(-0.30)Dealamount_Asset-0.226***-0.061***(-3.49)(-3.51)Interest_Spread0.102***0.028***(7.95)(8.03)Syndicated-0.035-0.009(-0.90)(-0.90)#_Prior_Loans_Log-0.032-0.009(-1.04)(-1.04)Log_Asset0.205***0.056***(11.54)(11.73)Litigation-0.006-0.002(-0.09)(-0.09)Inst_Own-0.273***-0.074***(-4.28)(-4.28)Std_Stockret-1.586***-0.431***(-6.14)(-6.15)MTB0.062***0.017***(4.09)(4.11)ROA0.1050.029(0.63)(0.63)Loss-0.113**-0.031**(-2.56)(-2.56)Leverage_Ratio-0.050-0.014(-0.53)(-0.53)Liquidity_Ratio0.1510.041(1.08)(1.08)Not_Rated0.0160.004(0.37)(0.37)Tangibility-0.146-0.040(-1.42)(-1.42)Log_INDSZ-0.012-0.003(-0.60)(-0.60)Log_ENTRY0.0320.009(1.55)(1.55)Constant-2.016***(-5.93)Year Fixed EffectYESYESIndustry Fixed EffectYESYESN19,084 19,084 pseudo R-sq0.1090.109Panel B of this table presents the results of regressions of borrowers’ decision to withhold a loan agreement by not filing it with the SEC (Not-filed indicator) on the firm-bank relationship intensity measured using two indicator variables: Relation_Intensity_Low and Relation_Intensity_High. Probit models are estimated for these analyses. The regressions include year and Fama-French 48 industry fixed effects. The coefficients of this Probit model are presented in column (1), and the marginal effects are reported in column (2). In parentheses, we present z-statistics. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 6. The Effect of Relationship Intensity on the Not-filing Decision Depending on Borrower Bargaining Power　(1)(2)(3)(4)Relation_Intensity-0.449***-0.472***-0.456***-0.427***(-7.18)(-7.55)(-7.47)(-6.70)Relation_Intensity * Small_Borrower0.280***　　　　(3.29)　　　Relation_Intensity * Not_Rated　0.317***　　　　(3.71)　　Relation_Intensity * Analyst_Low　　0.282***　　　　(3.51)　Relation_Intensity * Short_Credit_History　　　0.181**　　　　(2.15)Small_Borrower-0.036(-0.48) Analyst_Low-0.114*(-1.80)Short_Credit_History0.009(0.13)Dealamount_Asset-0.259***-0.252***-0.250***-0.252***(-3.96)(-3.87)(-3.86)(-3.86)Interest_Spread0.102***0.102***0.102***0.102***(7.93)(7.96)(7.93)(7.96)Syndicated-0.039-0.043-0.042-0.044(-1.01)(-1.12)(-1.08)(-1.15)#_Prior_Loans_Log-0.041-0.039-0.044-0.034(-1.31)(-1.25)(-1.40)(-1.08)Log_Asset0.233***0.207***0.216***0.212***(10.89)(11.68)(11.33)(11.89)Litigation-0.007-0.012-0.005-0.002(-0.10)(-0.18)(-0.07)(-0.03)Inst_Own-0.269***-0.271***-0.250***-0.255***(-4.24)(-4.25)(-3.89)(-3.99)Std_Stockret-1.582***-1.565***-1.550***-1.596***(-6.12)(-6.05)(-6.01)(-6.19)MTB0.063***0.063***0.069***0.059***(4.17)(4.11)(4.45)(3.88)ROA0.1140.1140.1330.141(0.68)(0.68)(0.79)(0.83)Loss-0.106**-0.110**-0.108**-0.107**(-2.42)(-2.52)(-2.46)(-2.44)Leverage_Ratio-0.024-0.030-0.029-0.040(-0.25)(-0.31)(-0.30)(-0.42)Liquidity_Ratio0.1530.1430.1710.140(1.09)(1.02)(1.23)(1.00)Not_Rated-0.009-0.182***0.0110.001(-0.20)(-2.66)(0.25)(0.03)Tangibility-0.143-0.146-0.146-0.154(-1.41)(-1.43)(-1.43)(-1.52)Log_INDSZ-0.013-0.014-0.012-0.016(-0.67)(-0.71)(-0.60)(-0.83)Log_ENTRY0.0320.0330.0310.034*(1.56)(1.61)(1.51)(1.66)Constant-1.898***-1.660***-1.766***-1.752***(-5.39)(-4.90)(-5.14)(-5.09)Year Fixed EffectYESYESYESYESIndustry Fixed EffectYESYESYESYESN19,08419,08419,08419,084pseudo R-sq0.1080.1080.1080.108This table presents regressions of borrowers’ decision to withhold a loan contract (Not-filed indicator) on the firm-bank relationship intensity (Relation_Intensity), conditioning on the borrowers’ bargaining power (i.e., Small_Borrower, Not_Rated, Analyst_Low, and Short_Credit_History). Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. Estimated coefficients are presented in column 1 to 4 and z-statistics are included in parentheses. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 7. The Effect of Bank Relationship Intensity on the Not-filing Decision Depending on Future Performance Changes　(1)(2)(3)(4)Relation_Intensity-0.344***-0.471***-0.358***-0.416***(-7.03)(-7.26)(-7.15)(-7.47)Sales_Growth_p1n1_Low0.049*-0.094*(1.84)(-1.66)OI_AT_Change_p1n1_Low-0.008-0.142**(-0.25)(-2.23)Relation_Intensity * Sales_Growth_p1n1_Low　0.228***　　　　(2.91)　　Relation_Intensity * OI_AT_Change_p1n1_Low 　　　0.222**　　　　(2.51)Dealamount_Asset-0.240***-0.229***-0.250***-0.244***(-3.59)(-3.43)(-3.72)(-3.63)Interest_Spread0.106***0.106***0.104***0.105***(8.15)(8.20)(7.89)(7.94)Syndicated-0.047-0.050-0.057-0.057(-1.23)(-1.29)(-1.43)(-1.43)#_Prior_Loans_Log-0.041-0.043-0.042-0.044(-1.30)(-1.37)(-1.31)(-1.37)Log_Asset0.205***0.206***0.207***0.207***(11.51)(11.58)(11.38)(11.43)Litigation-0.004-0.0040.0040.003(-0.05)(-0.05)(0.06)(0.04)Inst_Own-0.269***-0.270***-0.262***-0.262***(-4.19)(-4.20)(-4.00)(-4.00)Std_Stockret-1.608***-1.602***-1.666***-1.663***(-6.18)(-6.15)(-6.15)(-6.15)MTB0.070***0.072***0.064***0.065***(4.50)(4.61)(3.88)(3.92)ROA0.1330.1280.1510.151(0.78)(0.75)(0.83)(0.83)Loss-0.120***-0.118***-0.109**-0.106**(-2.71)(-2.66)(-2.38)(-2.32)Leverage_Ratio-0.025-0.029-0.049-0.052(-0.26)(-0.30)(-0.50)(-0.53)Liquidity_Ratio0.1500.1520.1450.142(1.04)(1.06)(0.95)(0.93)Not_Rated0.0180.0190.0230.023(0.42)(0.43)(0.51)(0.51)Tangibility-0.143-0.142-0.160-0.159(-1.39)(-1.38)(-1.53)(-1.52)Log_INDSZ-0.010-0.010-0.016-0.016(-0.51)(-0.52)(-0.79)(-0.78)Log_ENTRY0.0290.0300.038*0.037*(1.43)(1.45)(1.78)(1.77)Constant-1.775***-1.693***-1.660***-1.626***(-5.25)(-5.00)(-4.64)(-4.53)Year Fixed EffectYESYESYESYESIndustry EffectYESYESYESYESN18,948 18,948 18,041 18,041 pseudo R-sq0.1070.1080.1090.110This table presents regressions of borrowers’ decision to withhold a loan agreement (Not-filed indicator) on the firm-bank relationship intensity (Relation_Intensity), conditioning on the borrower’s change in future operating performances (Sales_Growth_p1n1_Low and OI_AT_Change_p1n1_Low). Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. Estimated coefficients are presented in columns 1 to 4 and z-statistics are included in parentheses. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 8. Not-File Contracts by Deal Amount to Total Assets Panel A - Descriptive ResultsDeal_Amount <=5% AssetDeal_Amount >5% AssetDeal_Amount >10% AssetDeal_Amount >15% AssetDeal_Amount >20% AssetDeal_Amount >25% AssetDeal_Amount >30% AssetTotal Number of Contracts2,767 16,317 13,468 11,160 9,363 7,835 6,548 % of Not-Filed Contracts49.87%18.64%16.27%14.76%14.27%14.07%13.72%Pearson Corr. b/w Not_filed & Relation_Intensity-0.12***-0.07***-0.04***-0.04***-0.04***-0.04***-0.03***Panel A of this table presents: 1) the total number of loan contracts for various subsamples of Deal Amount to Total Assets, 2) the percentage of not-filed contracts for each subsample, and 3) the Pearson correlation between Not_Filed and Relation_Intensity for various subsamples of Deal Amount to Total Assets. **, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 8. Not-File Contracts by Deal Amount to Total Assets (continued)Panel B - Determinants of the Decision not to File a Loan Contract by Deal Amount to Total Assets – Continuous SpecificationAll ObservationsDeal_Amount >5% AssetDeal_Amount >10% AssetDeal_Amount >15% AssetDeal_Amount >20% AssetDeal_Amount >25% AssetDeal_Amount >30% Asset(1)(2)(3)(4)(5)(6)(7)Relation_Intensity-0.345***-0.202***-0.175***-0.214***-0.274***-0.294***-0.205**　(-7.08)(-3.77)(-2.83)(-3.08)(-3.52)(-3.43)(-2.14)Dealamount_Asset-0.243***-0.122**-0.0370.0590.105*0.139**0.168**(-3.73)(-2.01)(-0.61)(0.97)(1.68)(2.12)(2.48)Interest_Spread0.104***0.065***0.059***0.054***0.062***0.052***0.041**(8.13)(4.67)(3.84)(3.31)(3.61)(2.85)(2.01)Syndicated-0.046-0.090**-0.111**-0.140***-0.153***-0.176***-0.195***(-1.19)(-2.09)(-2.30)(-2.66)(-2.66)(-2.82)(-2.83)#_Prior_Loans_Log-0.048-0.018-0.013-0.027-0.058-0.062-0.029(-1.52)(-0.55)(-0.37)(-0.70)(-1.38)(-1.34)(-0.57)Log_Asset0.207***0.151***0.130***0.113***0.110***0.106***0.090***(11.67)(7.88)(6.18)(4.94)(4.38)(3.96)(3.07)Litigation-0.010-0.032-0.030-0.046-0.069-0.085-0.167*(-0.15)(-0.48)(-0.43)(-0.62)(-0.88)(-0.98)(-1.73)Inst_Own-0.270***-0.207***-0.225***-0.208***-0.197**-0.147-0.134(-4.24)(-3.14)(-3.19)(-2.69)(-2.32)(-1.62)(-1.32)Std_Stockret-1.567***-1.110***-0.758***-0.639**-0.493-0.439-0.288(-6.09)(-4.33)(-2.81)(-2.19)(-1.57)(-1.31)(-0.78)MTB0.064***0.050***0.030*0.004-0.012-0.034-0.035(4.20)(3.20)(1.78)(0.25)(-0.63)(-1.60)(-1.49)ROA0.1210.1560.2040.1860.3180.3510.200(0.72)(0.91)(1.13)(0.96)(1.51)(1.51)(0.77)Loss-0.109**-0.061-0.0220.0070.0340.0380.056(-2.49)(-1.35)(-0.46)(0.14)(0.60)(0.60)(0.80)Leverage_Ratio-0.0360.0720.0630.030-0.047-0.051-0.113(-0.38)(0.75)(0.62)(0.27)(-0.40)(-0.40)(-0.83)Liquidity_Ratio0.1710.1400.294**0.2270.2630.337*0.377*(1.22)(0.99)(1.98)(1.41)(1.54)(1.80)(1.88)Not_Rated0.0120.0220.0190.0450.0290.0140.017(0.29)(0.48)(0.39)(0.86)(0.52)(0.23)(0.26)Tangibility-0.150-0.158-0.145-0.117-0.0200.0120.041(-1.46)(-1.54)(-1.35)(-1.04)(-0.17)(0.09)(0.30)Log_INDSZ-0.013-0.003-0.012-0.0120.010-0.0020.003(-0.64)(-0.15)(-0.56)(-0.52)(0.42)(-0.08)(0.10)Log_ENTRY0.0320.0260.0340.0280.0090.0230.015(1.57)(1.29)(1.54)(1.19)(0.36)(0.85)(0.51)Constant-1.727***-1.590***-1.338***-1.239***-1.291***-1.337***-1.265***(-5.13)(-4.79)(-4.07)(-3.47)(-3.33)(-3.10)(-2.76)Year Fixed EffectYESYESYESYESYESYESYESIndustry Fixed EffectYESYESYESYESYESYESYESN19,084 16,317 13,468 11,160 9,363 7,835 6,548 pseudo R-sq0.1070.0590.0420.0370.0370.0370.038Panel B of this table presents results from regressions of borrowers’ non-filing decision (Not_Filed) on the firm-bank relationship intensity (Relation_Intensity) and control variables estimated using various subsamples of Deal Amount to Total Assets. Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. In parentheses, we present z-statistics. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 8. Not-File Contracts by Deal Amount to Total Assets (continued)Panel C - Determinants of the Decision not to File a Loan Contract by Deal Amount to Total Assets – Indicator Variable SpecificationAll ObservationsDeal_Amount >5% AssetDeal_Amount >10% AssetDeal_Amount >15% AssetDeal_Amount >20% AssetDeal_Amount >25% AssetDeal_Amount >30% Asset(1)(2)(3)(4)(5)(6)(7)Relation_Intensity_Low0.313***0.222***0.178***0.190***0.211***0.243***0.136(8.87)(5.49)(3.76)(3.42)(3.25)(3.29)(1.57)Relation_Intensity_High-0.009-0.013-0.030-0.053-0.088**-0.092**-0.076*(-0.30)(-0.45)(-0.89)(-1.50)(-2.27)(-2.18)(-1.65)Dealamount_Asset-0.226***-0.105*-0.0250.0690.115*0.148**0.172**(-3.49)(-1.73)(-0.42)(1.13)(1.84)(2.28)(2.54)Interest_Spread0.102***0.064***0.058***0.054***0.062***0.052***0.041**(7.95)(4.55)(3.77)(3.27)(3.58)(2.82)(2.00)Syndicated-0.035-0.087**-0.110**-0.139***-0.153***-0.178***-0.196***(-0.90)(-2.00)(-2.27)(-2.64)(-2.67)(-2.86)(-2.86)#_Prior_Loans_Log-0.032-0.015-0.012-0.026-0.057-0.061-0.027(-1.04)(-0.46)(-0.35)(-0.68)(-1.37)(-1.36)(-0.54)Log_Asset0.205***0.153***0.132***0.115***0.112***0.109***0.091***(11.54)(7.94)(6.25)(5.01)(4.48)(4.06)(3.11)Litigation-0.006-0.031-0.028-0.045-0.067-0.083-0.166*(-0.09)(-0.45)(-0.40)(-0.60)(-0.85)(-0.96)(-1.72)Inst_Own-0.273***-0.209***-0.227***-0.208***-0.197**-0.144-0.132(-4.28)(-3.17)(-3.21)(-2.69)(-2.33)(-1.59)(-1.30)Std_Stockret-1.586***-1.124***-0.776***-0.654**-0.509-0.456-0.293(-6.14)(-4.39)(-2.88)(-2.24)(-1.63)(-1.36)(-0.79)MTB0.062***0.049***0.030*0.004-0.013-0.035-0.035(4.09)(3.14)(1.73)(0.21)(-0.66)(-1.64)(-1.50)ROA0.1050.1460.1960.1800.3160.3490.204(0.63)(0.86)(1.09)(0.93)(1.51)(1.50)(0.79)Loss-0.113**-0.064-0.0250.0050.0310.0350.055(-2.56)(-1.43)(-0.52)(0.09)(0.55)(0.56)(0.79)Leverage_Ratio-0.0500.0610.0520.019-0.058-0.063-0.118(-0.53)(0.63)(0.51)(0.18)(-0.50)(-0.50)(-0.86)Liquidity_Ratio0.1510.1280.282*0.2150.2490.318*0.371*(1.08)(0.90)(1.89)(1.33)(1.46)(1.70)(1.84)Not_Rated0.0160.0250.0210.0470.0310.0160.018(0.37)(0.56)(0.43)(0.90)(0.55)(0.26)(0.27)Tangibility-0.146-0.153-0.143-0.114-0.0170.0140.043(-1.42)(-1.49)(-1.33)(-1.02)(-0.14)(0.11)(0.31)Log_INDSZ-0.012-0.002-0.011-0.0110.010-0.0020.002(-0.60)(-0.11)(-0.54)(-0.50)(0.42)(-0.08)(0.09)Log_ENTRY0.0320.0260.0340.0280.0090.0230.015(1.55)(1.27)(1.54)(1.18)(0.35)(0.85)(0.51)Constant-2.016***-1.769***-1.473***-1.391***-1.475***-1.522***-1.382***(-5.93)(-5.30)(-4.50)(-3.93)(-3.86)(-3.59)(-3.07)Year Fixed EffectYESYESYESYESYESYESYESIndustry Fixed EffectYESYESYESYESYESYESYESN19,084 16,317 13,468 11,160 9,363 7,835 6,548 pseudo R-sq0.1090.0610.0430.0370.0380.0380.038Panel C of this table presents results from regressions of borrowers’ decision not to file a loan contract (Not_Filed) on relationship intensity (Relation_Intensity_Low vs. Relation_Intensity_High) and control variables estimated using various subsamples of Deal Amount to Total Assets. Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. In parentheses, we present z-statistics. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 9. Robustness AnalysesPanel A - Alternative Proxies for Relationship Intensity(1)(2)Relation_Intensity_Weighted-0.364***　　(-7.07)　Relation_Intensity_Duration_Log　-0.076***　　(-4.67)Dealamount_Asset-0.240***-0.314***(-3.67)(-4.78)Interest_Spread0.104***0.109***(8.11)(8.37)Syndicated-0.043-0.054(-1.12)(-1.40)#_Prior_Loans_Log-0.057*0.036(-1.77)(1.17)Log_Asset0.203***0.209***(11.44)(11.83)Litigious-0.010-0.017(-0.14)(-0.25)Inst_Own-0.267***-0.262***(-4.17)(-4.08)Std_Stockret-1.565***-1.557***(-6.08)(-6.04)MTB0.064***0.061***(4.21)(4.02)ROA0.1280.132(0.77)(0.78)Loss-0.109**-0.109**(-2.48)(-2.50)Leverage_Ratio-0.044-0.021(-0.46)(-0.21)Liquidity_Ratio0.1800.117(1.29)(0.83)Not_Rated0.0150.004(0.34)(0.08)Tangibility-0.142-0.168(-1.39)(-1.64)Log_INDSZ-0.012-0.012(-0.58)(-0.63)Log_ENTRY0.0310.030(1.51)(1.49)Constant-1.697***-1.944***(-5.02)(-5.90)Year Fixed EffectYESYESIndustry Fixed EffectYESYESN19,084 19,084 pseudo R-sq0.1070.104Panel A of this table presents regressions of borrowers’ decision to withhold a loan agreement (Not-filed indicator) on alternative proxies for firm-bank relationship intensity: Relation_Intensity_Weighted and Relation_Intensity_Duration_Log and control variables. Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. Estimated regression coefficients are presented in columns 1-3 and z-statistics are included in parentheses. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10% level, respectively. See Appendix A for variable definitions.Table 9. Robustness Analyses (continued)Panel B - Results with Correlated Random Effects, Loan Purpose and Loan Type Fixed Effects(1)(2)(3)(4)Relation_Intensity-0.326***-0.071***-0.264***-0.055***(-5.39)(-5.41)(-4.34)(-4.35)Dealamount_Asset-0.615***-0.135***-0.480***-0.101***(-6.87)(-6.92)(-5.31)(-5.34)Interest_Spread0.158***0.035***0.125***0.026***(9.51)(9.63)(7.33)(7.38)Syndicated-0.090*-0.020*-0.056-0.012(-1.67)(-1.68)(-1.02)(-1.02)#_Prior_Loans_Log-0.079**-0.017**-0.087**-0.018**(-2.25)(-2.25)(-2.44)(-2.45)Log_Asset-0.013-0.003-0.038-0.008(-0.32)(-0.32)(-0.93)(-0.93)Litigation0.2380.052-0.048-0.010(0.57)(0.57)(-0.08)(-0.08)Inst_Own0.1150.0250.1360.029(1.16)(1.16)(1.35)(1.35)Std_Stockret-1.091***-0.239***-1.052***-0.221***(-3.20)(-3.20)(-2.97)(-2.98)MTB0.062**0.014**0.0360.008(2.38)(2.38)(1.37)(1.38)ROA-0.032-0.007-0.130-0.027(-0.12)(-0.12)(-0.47)(-0.47)Loss-0.107*-0.023*-0.114*-0.024*(-1.82)(-1.82)(-1.88)(-1.88)Leverage_Ratio0.300*0.066*0.285*0.060*(1.88)(1.88)(1.77)(1.77)Liquidity_Ratio-0.379-0.083-0.361-0.076(-1.56)(-1.56)(-1.43)(-1.43)Not_Rated-0.017-0.004-0.013-0.003(-0.26)(-0.26)(-0.19)(-0.19)Tangibility-0.083-0.0180.0280.006(-0.36)(-0.36)(0.12)(0.12)Log_INDSZ0.0830.0180.095*0.020*(1.54)(1.54)(1.73)(1.74)Log_ENTRY-0.053-0.012-0.069-0.015(-1.11)(-1.11)(-1.42)(-1.42)Constant-1.948***-1.439***(-8.70)(-6.35)Year Fixed EffectYESYESYESYESCorrelated Random EffectYESYESYESYESLoan Purpose Fixed EffectNONOYESYESLoan Type Fixed EffectNONOYESYESN19,08419,08419,08419,084pseudo R-sq0.119 　0.1190.162 　0.162Panel B of this table presents regressions of borrowers’ decision to withhold a credit agreement (Not-filed indicator) on firm-bank relationship intensity (Relation_Intensity) and control variables. Correlated Random Effect Probit models are estimated for these analyses. Column 1 reports the regression coefficients from a model which includes year fixed effects and random correlated effects. The marginal effects of this model are reported in column 2. Column 3 reports the regression coefficients from a probit model which includes year fixed effects, random correlated effects, loan-purpose and loan-type fixed effects. The marginal effects of this model are reported in column 4. z-statistics are included in parentheses. Standard errors are clustered at firm (borrower) level. ***, **, and * denote statistical significance at 1%, 5%, and 10% level, respectively. See Appendix A for variable definitions.Table 9. Robustness Analyses (continued)Panel C - The Effect of Relation Intensity on the Not-filing Decision Depending on Agency Costs(1)(2)(3)(4)Relation_Intensity-0.355***-0.413***-0.357***-0.332***(-5.62)(-5.74)(-5.65)(-4.16)CEO_Dual0.035-0.071　　　(0.77)(-0.82)　　Board_Indep_Low　　0.0030.033　　　(0.07)(0.44)Relation_Intensity * CEO_Dual　0.178　　　　(1.50)　　Relation_Intensity * Board_Indep_Low　　　-0.048　　　　(-0.47)Dealamount_Asset-0.470***-0.466***-0.470***-0.470***(-4.04)(-4.01)(-4.05)(-4.05)Interest_Spread0.145***0.145***0.145***0.145***(9.07)(9.10)(9.07)(9.07)Syndicated0.208***0.206***0.208***0.207***(3.50)(3.47)(3.51)(3.50)#_Prior_Loan_Log0.088**0.087**0.087**0.088**(2.14)(2.10)(2.12)(2.13)Log_Asset0.166***0.166***0.168***0.168***(7.14)(7.16)(7.17)(7.17)Litigation-0.018-0.016-0.021-0.022(-0.21)(-0.19)(-0.25)(-0.25)Inst_Own-0.374***-0.371***-0.373***-0.373***(-4.86)(-4.84)(-4.85)(-4.85)Std_Stockret-1.748***-1.749***-1.749***-1.756***(-4.61)(-4.64)(-4.62)(-4.63)MTB0.085***0.085***0.085***0.085***(3.88)(3.87)(3.86)(3.86)ROA0.2040.2100.2010.200(0.76)(0.79)(0.75)(0.75)Loss-0.106*-0.104*-0.108*-0.108*(-1.76)(-1.73)(-1.78)(-1.79)Leverage_Ratio-0.047-0.048-0.049-0.050(-0.37)(-0.38)(-0.39)(-0.39)Liquidity_Ratio0.1070.1110.1060.106(0.51)(0.53)(0.51)(0.51)Not_Rated0.0700.0710.0680.067(1.22)(1.24)(1.19)(1.18)Tangibility-0.043-0.045-0.043-0.043(-0.32)(-0.34)(-0.33)(-0.32)Log_INDSZ-0.020-0.020-0.020-0.020(-0.75)(-0.76)(-0.76)(-0.76)Log_ENTRY0.0120.0130.0130.012(0.43)(0.46)(0.44)(0.44)Constant-2.026***-1.986***-2.027***-2.041***(-6.86)(-6.72)(-6.84)(-6.79)Year Fixed EffectYESYESYESYESIndustry Fixed EffectYESYESYESYESN11,875 11,875 11,875 11,875 pseudo R-sq0.1140.1140.1140.114Panel C of this table presents regressions of borrowers’ decision to withhold a loan agreement (Not-filed indicator) on the firm-bank relationship intensity (Relation_Intensity), after controlling for managerial entrenchment (CEO_Dual and Board_Indep_Low). Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. Estimated coefficients are presented in columns 1 to 4 and z-statistics are included in parentheses. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions.Table 9. Robustness Analyses (continued)Panel D - Additional Controls for the Costs of Not-filing Loan Contracts(1)(2)Relation_Intensity-0.330***(-6.28)Relation_Intensity_Low0.312***(8.27)Relation_Intensity_High0.000(0.01)KS_litigation_Risk-1.415**-1.360**　(-2.34)(-2.26)Post_Equity_Issuance-0.111***-0.111***　(-2.89)(-2.89)Post_Bond_Issuance0.0200.017　(0.62)(0.52)Dealamount_Asset-0.289***-0.268***(-3.67)(-3.45)Interest_Spread0.114***0.111***(8.18)(7.98)Syndicated-0.0020.011(-0.04)(0.25)#_Prior_Loans_Log-0.039-0.025(-1.11)(-0.74)Log_Asset0.218***0.215***(10.74)(10.61)Inst_Own-0.302***-0.305***(-4.36)(-4.39)Std_Stockret-1.400***-1.434***(-4.89)(-5.00)MTB0.082***0.080***(4.68)(4.61)ROA0.1590.128(0.83)(0.67)Loss-0.130***-0.135***(-2.71)(-2.81)Leverage_Ratio-0.062-0.078(-0.58)(-0.72)Liquidity_Ratio0.1920.177(1.18)(1.09)Not_Rated0.0170.021(0.36)(0.43)Tangibility-0.120-0.119(-1.03)(-1.01)Log_INDSZ0.0030.005(0.16)(0.22)Log_ENTRY0.0150.015(0.66)(0.64)Constant-1.940***-2.214***(-4.61)(-5.23)Year Fixed EffectYESYESIndustry Fixed EffectYESYESN16,510 16,510 pseudo R-sq0.1130.116Panel D of this table presents regressions of borrowers’ decision to withhold a loan agreement (Not-filed indicator) on the firm-bank relationship intensity (Relation_Intensity or Relation_Intensity_Low vs. Relation_Intensity_High) and control variables. In these regressions, litigation risk is captured using the litigation risk model (3) in Table 7 of Kim and Skinner (2012). Probit models are estimated for these analyses. All regressions include year and Fama-French 48 industry fixed effects. Z-statistics are included in parentheses. Standard errors are clustered at firm level. ***, **, and * denote statistical significance at 1%, 5%, and 10%, respectively. See Appendix A for variable definitions. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download