THE ROLE OF ACCOUNTING IN DEBT CONTRACT RENEGOTIATIONS: EVIDENCE FROM POSITIVE SHOCKS by MASSACHUSETTS INSTITUTE OF TECHNOLOLGY Kexin Zheng MAR 112015 BA, Nankai University, 2004 MBA, Boston College, 2009 LIBRARIES SUBMITTED TO THE SLOAN SCHOOL OF MANAGEMENT IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN MANAGEMENT AT THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY SEPTEMBER 2014 0 2014 Massachusetts Institute of Technology. All rights reserved Signature redacted Signature of Author MIT Sloan School of Management August 8, 2014 Signature redacted Certified by Joseph Weber George Maverick Bunker Professor of Management Professor of Accounting Thesis Supervisor Signature redacted Accepted by Ezra Zuckerman Nanyang Technological University Professor Chair, MIT Sloan PhD Program I 2 THE ROLE OF ACCOUNTING IN DEBT CONTRACT RENEGOTIATIONS: EVIDENCE FROM POSITIVE SHOCKS by Kexin Zheng Submitted to the Sloan School of Management on August 8, 2014 in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Management ABSTRACT Using a hand-collected sample of private debt contracts between U.S. publicly traded firms and financial institutions, I examine the role of accounting in the renegotiation of debt contracts following a positive shock to the borrower's credit quality. I find that, following a positive shock to their credit quality, firms with more timely reporting of good news are more likely to renegotiate their loan contracts and they do so sooner than firms with less timely good news reporting. Further, these effects are more pronounced for firms whose positive shocks can be more credibly communicated through financial reporting. My paper contributes to the literature on the role of accounting information in debt contract renegotiations. Thesis Supervisor: Joseph Weber Title: George Maverick Bunker Professor of Management, Professor of Accounting 3 Acknowledgements The journey to this thesis represents an important chapter of my life. During the five years I have spent at MIT, I have been very fortunate to have received the help and support of numerous people. First and foremost, I would like to thank the members of my dissertation committee: S.P. Kothari, Rodrigo Verdi, and Joseph Weber (Chair) for their guidance, feedback, and suggestions on this paper. I appreciate the amount of time that they have spent in guiding me through the PhD coursework, thesis, and job market. Their support has been invaluable. I am also grateful to the other current and former faculty members of the Accounting Group at the MIT Sloan School of Management: John Core, Anna Costello, Joao Granja, Michelle Hanlon, Scott Keating, Mozaffar Khan, Jeffery Ng, Christopher Noe, Reining Petacchi, Sugata Roychowdhury, Nemit Shroff, Ewa Sletten, Eric So, and Ross Watts, as well as my fellow Accounting PhD students and friends: Brian Akins, Josh Anderson, Nick Guest, Derek Johnson, Zawadi Lemayian, Rebecca Lester, Lynn Lei Li, Mihir Mehta, Patricia L. Naranjo, Heidi Packard, Daniel Saavedra, Benjamin Yost, and Luo Zuo. I must also thank Nancy Leonelli and Hillary Ross for their kindness and assistance to me. I received many valuable comments from the workshop participants at the FDIC Center for Financial Research, University of Connecticut, University of Minnesota, University of Toronto, and Washington University in St. Louis. I am grateful that the professors and economists at these institutes thought enough of me to give me the opportunity to interview to be a potential colleague. My paper has also benefitted from my interaction with fellow graduate students and friends: Winston Wei Dou, Song Lin, Indrajit Mitra, Hong Ru, Yu Xu, and Yang Sun. I thank my cousin-in-law, Jia Li at Duke University, for our helpful discussions on econometrics. I could not have made it to MIT without the support of and guidance from Professors Peter Wilson and Mary Ellen Carter at Boston College. They have always believed in me and encouraged me to realize my potential. I am also deeply grateful to all my friends both in the U.S. and back in China (too many to mention by name), who have consistently supported me and given me a great deal of joy. Particularly, I want to thank Dongwei Hao, Lynn Lei Li, Qian Liu, Yang Sun, Miao Yan, and Qiujie Zheng. Their friendship, compassion, and support have helped me through many challenges. I feel truly blessed to have found them to share my passions and outlook in life. Most importantly, I wish to thank my family. I owe a debt of gratitude to my parents Xiaoguang Wang and Li Zheng, for their unconditional love and many sacrifices in raising me to believe in myself, to endeavor, and to persevere. I thank my in-laws for lending a hand whenever asked. I thank my son, Rubin Xiao, who has brightened every aspect of my life since he came into the world. He has been a source of endless joy and comfort for me, and for our whole family. My last acknowledgement goes to my beloved husband, Bin Xiao, who has supported me in more ways than I can count, and kept my life balanced by showing me the beauty of the world. Beyond all doubt, his love and optimism have helped make this moment unforgettable. I dedicate this thesis to my family. 4 Table of Contents 1. In tro d uctio n ................................................................................................................................. 7 2. Background Literature and Hypothesis Development........................................................... 12 2.1 Renegotiations after Positive Shocks............................................................................... 12 2.2 Timely Reporting of Good News and Renegotiations after Positive Shocks ................... 13 2.3 Credibility of Accounting Information ............................................................................ 15 3. Sample and Research Design ................................................................................................. 18 3.1 Data and Sample Selection ............................................................................................... 18 3.2 M easure of Renegotiation.................................................................................................19 3.3 M easure of Timely Reporting of Good News.................................................................. 21 3.4 Test of Likelihood of Renegotiations .............................................................................. 22 3.5 Test of Time to Renegotiate............................................................................................. 24 3.6 Cross-Sectional Tests on the Credibility of Accounting Information.............................. 25 4. Empirical Results.......................................................................................................................27 4.1 Descriptive Statistics............................................................................................................27 4.2 Likelihood and Timeliness of Renegotiations after Positive Shocks............................... 28 4.3 Cross-Sectional Test Results........................................................................................... 29 4.4 Additional Tests ................................................................................................................... 30 4.4.1 Economic Impact of Timely Reporting of Good News on Renegotiations after Positive 30 S h o ck s .................................................................................................................................... 4.4.2 Implications for Accounting Conservatism................................................................33 4.4.3 Nature of Good News Received ............................................................................... 34 4.4.4 Alternative Measure of Timely Reporting of Good News ......................................... 35 4.4.5 Timely Reporting of Good News Interacted with the Size of Positive Shock..........36 4.4.6 Firm Loans without Earnings-Based Performance Pricing Provision............36 4.4.7 Switch of Lenders...................................................................................................... 37 4.4.8 Control for Changes in Firm Characteristics.................................................................37 5 . C on clu sion ................................................................................................................................. 37 Referen ces......................................................................................................................................3 9 Appendix I: Variable Definitions ............................................................................................... 45 Appendix II: Example of Amendment Report........................................................................... 49 5 List of Tables T able 1: Sam ple Selection ........................................................................................................ 53 Table 2: D escriptive Statistics ................................................................................................... 54 Table 3: C orrelation M atrix ...................................................................................................... 56 Table 4: The Likelihood and Timeliness of Renegotiations after Positive Shocks...................57 T ab le 5 ........................................................................................................................................... 58 Panel A: Cross-Sectional Tests on the Likelihood of Renegotiations ................................... 58 Panel B: Cross-Sectional Tests on the Timeliness of Renegotiations ................................... 59 T able 6: E conomic Impact.............................................................................................................6 1 Table 7: Likelihood and Timeliness of Renegotiations on Conservatism.................................62 Table 8: Nature of Good News Received ................................................................................. 63 Table 9: Alternative Measure of Timely Reporting of Good News .......................................... 64 Table 10: Interact Timely Reporting of Good News with the Size of Positive Shock..............65 Table 11: Firm Loans Without Earnings-Based Performance Pricing Provision..................... 66 Table 12: Refinancing from Outside Lenders .......................................................................... Table 13: Control for Changes in Firm Characteristics.............................................................68 List of Figures Figure 1: Timeline of the Measurement Period for the Variables.............................................52 6 67 1. Introduction Renegotiations play a critical role in corporate financial contracting (Aghion and Bolton 1992; Bolton and Scharfstein 1996; Hart and Moore 1998; Roberts and Sufi 2009a). While the majority of the previous literature focuses on renegotiations in the context of corporate default, recent empirical studies show that debt contracts are frequently renegotiated even when a default has not occurred (Roberts and Sufi 2009b; Roberts 2012). In fact, numerous renegotiations are voluntarily initiated by borrowers, especially when the borrower's credit quality improves. But while recent research has provided evidence of the pervasiveness of such renegotiations, the role of accounting in such a setting is not well understood. In my paper, I examine whether and how the timely reporting of good news influences contract renegotiations following a positive shock to the borrower's credit quality (hereafter, 'renegotiations after positive shocks'). I first hypothesize that, following a positive shock to their credit quality, firms with more timely reporting of good news are more likely to renegotiate their loan contracts and they do so sooner than firms with less timely good news reporting. This hypothesis is motivated by the incomplete contracting literature. When a positive shock to credit quality occurs after the origination of loan contracts, borrowers are motivated to renegotiate any constraining contract terms that do not reflect the current credit quality (Hart and Moore 1998; Maskin and Moore 1999). However, lenders may not agree to renegotiate with borrowers if the information concerning the borrower's credit quality is not verifiable (Petersen 2004). In that case, the borrower's accounting system is expected to influence renegotiations because accounting information is scrutinized by auditors and regulators, which makes the recognition of the good news verifiable. Therefore, it can serve as a key source of verifiable information for updating the lender's prior belief about the borrower's credit quality (Armstrong, Guay, and Weber 2010; 7 Minnis 2011). Further, by disclosing good news in their financial statements in a more timely manner, borrowers can provide lenders with more timely and verifiable data on improvements in credit quality. This allows lenders to better re-assess a borrower's credit quality, improving the probability and the timeliness of renegotiations. In my second hypothesis, I predict that the effects of the timely reporting of good news on debt contract renegotiations after positive shocks are more pronounced for firms whose positive shocks can be more credibly communicated through financial reporting. While firms can report good news in a timely manner, such news can influence renegotiations only to the extent that lenders find it credible (Maines and Wahlen 2006; Minnis 2011). To test this prediction, I build on prior research that shows that reported information is perceived to be more credible when the firm does not have a history of financial misreporting, proxied by financial restatements (Graham et al. 2008; Dechow et al. 2010; Kravet and Shevlin 2010); when the firm's auditor provides more effective audits, proxied by industry specialist (Levitt 2000; Dechow et al. 2010; Defond and Zhang 2013); and when the improvement in performance is less transitory (Ramakrishnan and Thomas 1998; Baber et al. 1998; Li 2010). I then hypothesize that the effects of the timely reporting of good news on renegotiations after positive shocks are more pronounced when the firm does not have prior financial restatements, when the firm is audited by an industry specialist, and when the improvement in performance is less transitory. I test the above hypotheses on a sample of 466 firm loans for the years 1996 through 2005. Sample firms must have experienced a significant positive shock to credit quality in the two quarters following contract origination. I proxy for such a shock by requiring that a firm experience a 1% decrease in the probability of bankruptcy, measured by the Black-ScholesMerton option-pricing model (BSM-Prob) (Hillegeist et al. 2004). After obtaining a sample of 8 firms that experienced a positive shock, I hand-collect information on renegotiation by examining each borrower's SEC filings (10-Ks, 10-Qs and 8-Ks) for the six quarters after the positive shock. I use the probit model and the hazard model to test respectively the probability and the timeliness of renegotiations. In the probit model, the dependent variable is an indicator variable that takes the value of one if the original contract is amended within the six quarters subsequent to the positive shock. In the hazard model, the dependent variable is a continuous variable measuring the time that elapses between a positive shock to credit quality and a later renegotiation. The mean probability of renegotiation in my sample is 50%, and the mean time between positive shock and renegotiation is 1.58 quarters. As my hypothesis relates to the timely reporting of good news, I use the Khan and Watts (2009) G-Score measure, which captures how quickly good news is reflected in accounting earnings. The existing debt contracting literature focuses largely on cross-sectional variation in the degree of total timely loss recognition and in asymmetric timeliness of loss recognition. However, as Guay and Verrecchia (2006) note, under the assumption that the timely reporting of information is a key property of accounting reports, the timely reporting of gains should also play an important role in corporate behavior. This is particularly the case for my study, as I focus on the timeliness with which firms can credibly recognize a positive shock to credit quality. Consistent with my first hypothesis, G-Score is positively associated with the probability of renegotiations after positive shocks and negatively associated with the time to renegotiate. The results are also economically significant. A one standard deviation increase in G-Score is associated with a 6.2% increase in the predicted probability of renegotiation compared to 50% unconditional probability of renegotiation. In addition, a one standard deviation higher G-Score 9 translates into a 2.7% higher renegotiation hazard rate, which is equivalent to 1.06 quarters shorter time to renegotiate. To put this in perspective, among firms that report a renegotiation after a positive shock (50% of my sample), 30% of firms receive a reduction in interest rates, 38% of firms receive an increase in loan amount, and 34% of firms extend their loan maturity. On average, firms can receive a reduction of 86 bps in interest rates (relative to the average initial interest rate of 265.71 bps), an increase of 53% of original loan amount, or an increase of 92% of original loan maturity during renegotiations. Thus, my results suggest that by increasing the chance of a renegotiation, firms with a higher G-Score are more likely to obtain better contract terms. Consistent with my second hypothesis, the timely reporting of good news plays a more important role in renegotiations after positive shocks when the firm does not have prior restatements (my proxy for prior misreporting), when it is audited by industry specialists (my proxy for auditing quality), and when the reported income is less affected by one-time items (my proxy for transitory earnings). In terms of economic significance, a one standard deviation increase in G-Score is associated with a 7.6% increase in the predicted probability of renegotiation (Reneg) when firms do not have prior restatements, versus a 3.5% decrease when firms have restated their financial statements. Similarly, a one standard deviation increase in GScore is associated with a 7.7% increase in Reneg for firms that are audited by an industry specialist, as opposed to a 5.0% increase for firms audited by non-specialists. Lastly, a one standard deviation increase in G-Score is associated with an 8.3% increase in Reneg when the proportion of one-time items is below the sample median, in contrast to a 3.7% increase when it is above the sample median. 10 My paper makes two main contributions. First, it enhances our understanding of the role of accounting information in incomplete contracting. Incomplete contract theory posits that contract renegotiation is often triggered by the arrival of new information. While the majority of the accounting literature has focused on renegotiation triggered by negative news such as covenant violations (Healy and Palepu 1990; Sweeney 1994; DeFond and Jiambalvo 1994; Dichev and Skinner 2002; Zhang 2008; Tan 2013), there is little evidence on the role of accounting information in renegotiations voluntarily initiated by borrowers when their credit quality improves.' My paper fills this gap, suggesting that accounting information, proxied by the timely reporting of good news, plays an important role in such a setting. Second, by providing evidence on the role of accounting in ex post contract renegotiations after positive shocks, my paper also provides a rationale for prior findings on the ex ante benefits of accounting conservatism (i.e., conservative firms receive better contract terms at the time of the loan initiation) (Zhang 2008; Wittenberg-Moerman 2008). Specifically, I find that firms that accelerate disclosing good news (i.e., less conservative firms) are more likely to renegotiate their contracts after a positive shock to credit quality.2 This represents a risk to the lenders because renegotiations are associated with better contract terms from the borrower's perspective (e.g., lower interest rates). As a result, lenders anticipate and price protect against My paper is related to Dou (2013), which investigates the impact of accounting information on the likelihood of all renegotiations. However, since Dou (2013) does not distinguish between renegotiations after positive shocks and those after negative shocks, it focuses on other attributes of accounting, such as the debt contracting value of accounting information. My paper, in contrast, focuses on renegotiations after positive shocks and on the role of the timely reporting of good news, which is (arguably) more suitable for this setting. 2 Technically speaking, the timely reporting of good news is not necessarily inversely related to conservatism. For example, the standard estimation of conservatism (i.e., asymmetric timeliness as in Basu 1997) is measured incrementally to good news. So it is possible that firms could increase their reporting timeliness of good news without affecting their asymmetric timeliness. Empirically and in my study, however, these constructs are negatively correlated, so the results related to the timely reporting of good news are likely to have implications for accounting conservatism. In fact, in robustness tests, I show that for more conservative firms (measured as asymmetric timeliness), renegotiations after positive shocks are both less likely and less timely. 11 this risk by charging higher interest rates for less conservative firms at the time of the loan initiation. The rest of the paper proceeds as follows. Section 2 reviews the literature and develops my hypotheses. In Section 3, 1 introduce the sample and research design, and Section 4 presents empirical evidence. Section 5 summarizes my findings. 2. Background Literature and Hypothesis Development 2.1 Renegotiations after Positive Shocks Renegotiation occurs when parties to a contract are unable to commit to the initial terms of their agreement after the relationship has been initiated. The theory of incomplete contracts posits that renegotiation is driven mainly by exogenous uncertainty.3 When unanticipated (i.e., non-contractible) states of the world occur, the initial terms of the contract are no longer optimal, and renegotiations will be required (Hart and Moore 1998; Maskin and Moore 1999; Gorton and Kahn 2000; Garleanu and Zwiebel 2009).4 While early studies focus on renegotiations in financial distress or corporate default such as covenant violations, Roberts and Sufi (2009b) study all renegotiations.5 They find that 90% of long-term loan contracts are renegotiated before maturity, and fewer than 18% of renegotiations are directly or indirectly associated with a covenant violation or payment default. Consistent with the theoretical model, their analysis of what triggers renegotiation indicates that new information concerning credit quality is a strong 3 When drawing up a contract, it is not viable to specify all the relevant contingencies. Hence, the contract will often be incomplete. 4 In addition to exogenous uncertainty, the original contract design plays an important role in renegotiation (Aghion et al. 1994). Empirically, therefore, I control for the original contract tenns. 5 Prior literature on renegotiations triggered by corporate default studies the outcome and implications of covenant violations. For example, creditors could either waive the covenant violations or demand certain conditions such as an increase in the interest rate, a reduction in the borrowing base, or a shift of control rights to the creditors (Beneish and Press 1993; Chen and Wei 1993; Smith 1993; Chava and Roberts 2008; Nini et al. 2009). 12 predictor of the incidence and outcomes of renegotiation. In particular, numerous renegotiations are voluntarily initiated by borrowers, especially when the borrower's credit quality improves. Extending the work of Roberts and Sufi (2009b), which focuses on the probability of renegotiation, Roberts (2012) and Nikolaev (2013) further investigate the dynamics of renegotiation, i.e., the timing (scope) of renegotiation. While recent finance research has provided evidence of renegotiation outside of default, the extant accounting literature focuses on the role of accounting in the context of covenant violations. For example, several studies provide evidence that managers make accounting choices or take real actions to avoid covenant violations (Healy and Palepu 1990; Defond and Jiambalvo 1994; Sweeny 1994; Dichev and Skinner 2002; Daniel et al. 2008). In addition, Zhang (2008) shows that accounting conservatism benefits lenders by accelerating covenant violations, and Tan (2013) finds that firms' financial reporting becomes more conservative immediately after covenant violations. My paper contributes to the accounting literature by examining the role of accounting information in renegotiations after positive shocks. 2.2 Timely Reporting of Good News and Renegotiations after Positive Shocks While previous literature emphasizes that lenders bear downside risk but no upside potential, improvement in the borrower's credit quality is not inconsequential to them. With a credible threat to leave current lenders in favor of alternative sources of financing, borrowers can voluntarily initiate renegotiations with lenders when their credit quality improves (Hart and Moore 1998; Gorton and Kahn 2000; Garleanu and Zwiebel 2009; Roberts and Sufi 2009b). 6 In other words, a positive shock to credit quality can shift bargaining power in the borrower's 6 In practice, many lenders refer to borrowers' wishing to "go outside the deal" as a motivation for an amendment. Roberts and Sufi (2009b) suggest that competition among lenders and the existence of alternative sources of financing, such as external equity, increase borrowers' bargaining power by increasing the credibility of any threat to switch lenders. 13 favor, so the borrower may be able to renegotiate any constraining contract terms that do not reflect the firm's current credit quality (e.g., reduce interest rates, increase loan amounts, extend loan maturity, and loosen the restrictiveness of loan covenants). While new information concerning the borrower's credit quality is an important determinant of renegotiations, lenders may not agree to renegotiate with borrowers if this information is not verifiable. Petersen (2004) suggests that decision makers place more weight on verifiable information (i.e., "hard" information) because it is more objective and precise; in contrast, "soft" information is subjective and usually cannot be verified at low cost, which makes it less reliable. The borrower's accounting system is expected to significantly influence renegotiations because accounting information can serve as a key source of verifiable information to update a lender's prior belief regarding a borrower's credit quality (Armstrong, Guay, and Weber 2010; Minnis 201 1).7 The reason is that accounting information is scrutinized by auditors and regulators, which makes the recognition of the good news verifiable. Further, by disclosing good news in their financial statements in a more timely manner, borrowers can provide lenders with more timely and verifiable data on improvements in credit quality. This allows lenders to better re-assess the credit quality of such firms in a timely manner, which improves the probability and timeliness of renegotiations (Ball and Shivakumar 2005; Wittenberg-Moerman 2008; Cassar 2011; Kim et al. 2011; Minnis 2011).8 While borrowers that do not report good news in a timely fashion may voluntarily disclose private information (i.e., soft information) 7 Financial statement verification is considered by regulators and exchanges to be so important that all firms with publicly available filings are required to have their financial statements audited. The role of an audit is to assure financial statement users that the statements are compiled and presented according to Generally Accepted Accounting Principles (GAAP) and faithfully represent the client's underlying economic activities (Ball 2001; Ball, Jayarman, and Shivakumar 2009). Per FASB's Statement of Financial Accounting Concepts No. 2: "the purpose of verification is to provide a significant degree of assurance that accounting measures represent what they purport to represent" (FASB 1980, paragraph 8 1). 8 I focus on the probability that firms renegotiate their loan contracts within a given time window (e.g., six quarters after positive shocks). 14 about gains to lenders, such voluntary disclosures are usually more costly to verify, leading to higher verification cost. Given the same magnitude of positive change in credit quality, higher verification cost may prevent borrowers to renegotiate their loans with lenders in time. Therefore, firms with less timely good news reporting may have to wait and renegotiate their contracts at a later point, when the good news becomes hard information for lenders. This leads to my first set of hypotheses: H1(a): Ceteris paribus, following a positive shock to credit quality, firms with more timely reporting of good news are more likely to renegotiate their loan contracts. H1(b): Ceteris paribus, following a positive shock to credit quality, firms with more timely reporting of good news are more timely in renegotiating their loan contracts. 2.3 CredibilityofAccounting Information My second hypothesis exploits the extent to which a good news shock can be credibly communicated. While firms can report good news in a timely manner, such news will influence the contract renegotiation only to the extent that lenders find it credible (Maines and Wahlen 2006; Minnis 2011). In this section, I discuss three characteristics - financial misreporting, audit quality, and transitory earnings - that may affect the credibility of accounting information and thereby affect the impact of the timely reporting of good news in renegotiations after positive shocks. First, I argue that reported financial information is perceived by lenders to be more credible when the firm does not have a history of financial misreporting (e.g., restatements of financial reports). Prior literature has shown that restatements of financial reports are associated with more earnings quality problems (i.e., both intentional misstatements and unintentional errors) and affect the credibility of financial reporting (Hribar and Jenkins 2004; Bums and 15 Kedia 2006; Hennes et al. 2010; Dechow et al. 2010; Kravet and Shevlin 2010). Although a restatement might increase the uncertainty about one particular accounting item, it also causes lenders to question other aspects of the firm's operations and reported performance, increasing the overall uncertainty about a company's financial information. As a result, the perceived information asymmetry between borrowers and lenders increases after restatement (Graham et al. 2008; Costello and Wittenberg-Moerman 2011).9 The increased information asymmetry reduces financial statements' perceived credibility and makes any gains promptly reported therein seem less credible to lenders, leading to fewer renegotiations after positive shocks. In addition, the effectiveness of an audit in identifying misstatements is also positively associated with reporting credibility (Levitt 2000; Dechow et al. 2010).10 Prior literature suggests that industry specialists provide more effective audits than non-specialists do, leading to enhanced financial statement credibility (Wright and Wright 1997; Balsam et al. 2003; Krishnan 2003; Dunn and Mayhew 2004)." This is because industry specialists' expertise, which is derived from serving other clients in the same industry and learning and sharing best practices across the industry, enables them to more effectively identify misstatements. In addition, to protect their market shares, specialists also have more of an incentive to correct or report identified misstatements. Therefore, lenders are more likely to trust the gains specialist auditors have verified and are thus more likely to agree on the borrower-initiated renegotiations. Finally, I examine the extent to which the improvement in performance is transitory. While not directly capturing the credibility of financial information, the performance's transitory Costello and Wittenberg-Moerman (2011) find that when a firm reports an internal control weakness, lenders decrease their reliance on financial-ratio-based performance pricing provisions and financial covenants. That is, lenders place less weight on questionable financial data. 10 As noted by Levitt (2000), investor perceptions of audit quality play a critical role in maintaining systemic confidence in the integrity of financial reporting. The higher perceived audit quality, the more credible the auditee's financial statements. " Big 4 auditors represent another widely used proxy for audit quality. However, I find that since the Big 4 audit approximately 90% of my sample, there is very little cross-sectional variation in this variable. 9 16 components relate to the extent to which good news can be used for re-contracting purposes. Specifically, debt contracts are more likely to use core earnings (e.g., EBITDA) which are reported earnings minus transitory earnings (Li 2010; Kothari et al. 2010; Demerjian 2011). One important reason is that core earnings are more credible for measuring firm performance and less susceptible to manipulation (Leftwich 1983; Watts and Zimmerman 1986; Ohlson 1999; Collins et al. 1997; McVay 2006).13 Thus, to the extent that good news is transitory (e.g., a gain from asset sales), while it will affect earnings in a timely manner, it might be excluded for recontracting purposes, leading to a lower probability of renegotiations. Therefore, all else equal, the higher proportion of core earnings, the more likely it is that the gains promptly recognized in financial statements will trigger renegotiations. Taken together, this argument motivates the following hypotheses:' 4 H2(a): Ceteris paribus, the effects of the timely reporting of good news on renegotiations after positive shocks are more pronounced when the firm does not have prior financial restatements. H2(b): Ceteris paribus, the effects of the timely reporting of good news on renegotiations after positive shocks are more pronounced when the firm is audited by an industry specialist. 1 Transitory earnings measure certain earnings components that are generally viewed as non-recurring in standard financial statement analysis (e.g., extraordinary items and income from asset sales, asset revaluation, discontinued operations, accounting changes); and more than half of debt contracts exclude transitory earnings when defining net income (Li 2010). 1 Transitory earnings are less informative for forecasting firms' future performance and are usually used as an income smoothing device (Collins et al. 1997). In addition, Leftwich argues that GAAP numbers are adjusted (e.g., transitory earnings are excluded) to reduce managers' ability to circumvent covenant violation through accounting manipulation. " When financial reporting is less credible, firms with more timely reporting of good news may not trigger renegotiations, leading to longer time elapsed until a renegotiation occurs (i.e., infinite time). Therefore, the second hypothesis applies to both the likelihood and the timeliness of renegotiations. 17 H2(c): Ceteris paribus, the effects of the timely reporting of good news on renegotiations after positive shocks are more pronounced when the improvement in performance is less transitory. 3. Sample and Research Design 3.1 Data and Sample Selection Table 1 summarizes the sample selection process. I begin with a sample of loan deals from Reuters Loan Pricing Corporations' DealScan database that are matched with firms' financial characteristics from Standard and Poor's Compustat database.' 5 The sample is restricted to deals initiated during the years 1996 - 2005. My starting year corresponds to the initial SEC-required electronic filing year; prior to 1996, electronic filings are only sparsely available on EDGAR. 16 This sample period is also consistent with that in Roberts and Sufi (2009b). As previously mentioned, I require the sample firms experience a significant positive shock to credit quality two quarters subsequent to the origination of loan contracts. The credit quality change is calculated as the change in the probability of bankruptcy (i.e., BSM-Prob), measured by the Black-Scholes-Merton option-pricing model (Hillegeist et al. 2004).7 The rationale for this measure is that the shock to credit quality is calculated by a market model, and The extracted loan deals consist of both original credit agreements and the agreements' major amendments and restatements. I perceive an amended/restated loan agreement as a new loan. This method is consistent with the DealScan data collection principle. 16 The sample selection procedure balances the need for a sufficiently large sample to yield reasonable power (and to ensure that the results are somewhat generalizable) against the costs in time and effort to manually obtain renegotiation information. Due to time constraints, the sample period is temporarily restricted to deals initiated during the years 1996 to 2005. This sample period will be extended in future work. 17 Hillegeist et al. (2004) estimate a market-based measure of the probability of bankruptcy based on the BlackScholes-Merton option-pricing model, BSM-Prob. They show that BSM-Prob provides considerably more information than either of the two popular accounting-based measures, Altman's (1986) Z-Score and Ohlson's (1980) O-Score. They estimate the default distance at the finn-year level. I compute this measure at the firm-quarter level by replacing their yearly variables with quarterly variables (i.e., equity volatility, the risk-free rate, market value, long-term debt, and dividend yield). I choose a two-quarter window to measure the credit quality improvement because BSM-Prob does not change significantly within a shorter window. 15 18 I assess whether accounting systems can disclose such good news in financial statements in a timely manner. I assume that lenders do not base contracts exclusively on market-based measures and argue that financial reporting provides incremental, verifiable information to lenders (Sloan 1996; Ball et al. 2008).18 1 include only firm deals that have experienced at least a 1% decrease in BSM-Prob, yielding a sample of 1,195 firm deals. 19 I require that each deal have information about the interest spread and the loan amount of all tranches in the deal. Deals must also have a maturity greater than 12 months, allowing enough time for firms to renegotiate their loans before maturity. From Compustat, I construct financial statistics (Appendix I) as the average over the four quarters prior to the signing of a loan agreement. I include only deals for which these borrower-level variables are present. In addition, data must be available to calculate the measure of the timely reporting of good news. These requirements yield the final sample of 466 firm loans. To obtain the restatement data, I use the GAO Financial Statement Restatement Database (e.g., Desai et al. 2006). This database was constructed using a Lexis-Nexis text search based on variations of the word "restate" and contains approximately 2,309 restatements between January 1997 and December 2007. Therefore, my sample period for cross-sectional tests on restatements is from 1998, leading to a sample of 409 firm loans. 3.2 Measure of Renegotiation After obtaining a sample of firms that experienced a positive shock, I then hand-collect the information on renegotiation by examining the SEC filings (10-Ks, 10-Qs, and 8-Ks) of each 18 By collecting and summarizing the financial effects of firms' investment, operating, and financing activities, accounting systems convey detailed information about the underlying sources of changes in credit quality. '9 The average probability of bankruptcy (BSM-Prob) of all compustat firms is 5%. Similar to the findings in Hillegeist et al. (2004), 1 find that approximately 60% of firm deals did not experience any change in the BSM-Prob (i.e., firms continue to have a zero probability of bankruptcy) two quarters after loan origination. Among firms that experienced a decrease in the probability of bankruptcy, only 15% of firms experienced not less than 1% decrease. Therefore, this 1% cut-off rate is economically significant, yielding reasonable power. 19 borrower for the six quarters after the positive shock. 2()A six-quarter window comprises a period sufficient for testing the timeliness of renegotiation as accounting systems that report good news ' less promptly may renegotiate at a later point when the good news becomes verifiable.2 To investigate the probability of renegotiation, I identify the ending date of the positive shock to credit quality (hereafter, 4the end of the positive shock'). The indicator variable Reneg takes the value of one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. In case the credit quality changes again after the positive shock, I also calculate the BSM-Prob at the quarter of renegotiation and delete those renegotiations that are not related to positive shocks in credit quality. 22 Further, to analyze the time to renegotiate, I identify the date of the first renegotiation after the positive shock. If the borrower reports a renegotiation, I define the dependent variable TimetoReneg as the number of quarters between the end of the positive shock and the renegotiation.2 3 If the firm does not report a renegotiation within the search period, I define TimetoReneg as the number of quarters between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter.2 4 The SEC requires disclosure of any material changes to the debt agreements in the exhibits of company filings. By following the explanations of debt agreements through time. I can detect whether the original loan contracts are amended. 21 Results are robust if I use a shorter window of four quarters or a longer window of eight quarters. 22 Specifically, if the reduction in BSM-Prob by the quarter of renegotiation drops more than half of the reduction in BSM-Prob within the first two quarters, then I do not consider it as renegotiation triggered by a positive shock in credit quality. My data shows that only 26 out of 234 renegotiations are this case, not affecting my results. 2' Time to Reneg is measured in quarters to be consistent with the quarterly measure of positive shock. However, the results are robust to monthly measure. 2 Nikolaev (2013) was among the first to explore the time to renegotiation, but his setting focuses on contract renegotiations that require lender majority consent, as opposed to unanimous consent. However, renegotiations that 2 20 3.3 Measure of Timely Reporting of Good News I use the Khan and Watts (2009) G-Score to measure the timely reporting of good news. Khan and Watts (2009) develop their firm-year measure of good news timeliness based on the Basu (1997) cross-sectional model: X =,81 + /q2Di + /33R+ P4D1R + ej, (1) where X is earnings, R is returns (measuring news), and D is a dummy variable equal to I when R<0 (bad news), zero otherwise. The good news timeliness is represented by 83, and the incremental timeliness for bad news over good news is represented by #4. To estimate the timeliness of good news at the firm-year level, they further define G-Score as a linear function of firm-specific characteristics (firm size, market-to-book ratio, and leverage) each year: G-Score, =3= pi + u2Sizei + p3 M/Bi + p 4Levi. (2) Substituting G-Score into equation (1), an annual cross-sectional regression model is derived to estimate the parameters (tI-P4) in equation (2). X =1 +/8 2D, + Ri(pi + p2Sizei + p 3M/Bi +(6jSizei + 6 2M/B, + 63 Levi + P 4Levi) + DiRi( + A2Size, + )M/Bi + X4Lev) + 6 4DiSizei + 5DiM/Bi + 6 6DLevi) + Ei. (3) The estimated parameters (u-p4) can be used to calculate the G-Score in equation (3). G-Scorei shows how good news is reflected in net income, where a larger coefficient indicates that earnings exhibit a greater response for a given amount of good news in returns. require unanimous consent are more likely to be due to significant events (e.g., a sufficient change in credit quality), which are important to my study. In general, Nikolaev (2013) provides evidence on the link between time to renegotiation and contracting frictions (as characterized by agency conflicts and information problems). 25 The discretion in accounting earnings that relates to the timeliness of good news recognition has not been discussed sufficiently in previous literature. The earnings recognition could leave managers discretion. For example, some products are combined with services. When firms already sold this kind of products, firms that report good news in a more timely manner may choose to recognize both the sale of products and the sale of services, but firms that report good news in a less timely manner may simply recognize the revenue from the sale of products and recognize the services revenue later. 21 Figure 1 provides a timeline of the measures of the main variables. As shown in Figure 1, I measure the timeliness of good news reporting G-Score at the year-end before the loan origination. 26 This method assumes that borrowers can commit to a certain level of reporting timeliness of good news and that they do not subsequently deviate due to either a lack of accounting slack or reputation concerns. 3.4 Test of Likelihood of Renegotiations Hi(a) hypothesizes that following a positive shock to credit quality, firms with more timely reporting of good news are more likely to renegotiate their loan contracts. I test this hypothesis by estimating the following probit model: Renegi =,8o + 1iG-Score +X'1 ( + ei. (4) The dependent variable Reneg measures the probability of renegotiations after positive shocks and G-Score captures how timely borrowers reflect good news in their financial statements. H 1(a) predicts that P > 0. In terms of control variables X, I first include the size of a positive shock to credit quality (ZBSM Prob), measured by the absolute value of the change in BSM-Prob two quarters after the loan origination. The larger the positive shocks to credit quality, the more likely firms are to initiate renegotiation with their lenders. Then, drawing from the quarterly data before the loan origination, I use the following firm-specific characteristics to proxy for the original credit quality of each firm. The natural log of the total assets (LNASSET) captures the firm's ability to secure or collateralize its debt, and proxies for the liquidation value in distress. A larger firm usually has more bargaining power, leading to a higher likelihood of renegotiations after positive shocks. I use the ratio of debt to EBITDA (DTE) and the ratio of debt to total book assets (LEV) One important reason for measuring G-Score at the loan origination is to partially alleviate endogeneity concerns. For robustness, I also measure G-Score at the year-end before the positive shock; the results are robust. 26 22 to proxy for the financial health of each firm. The market-to-book ratio (MTB) measures future investment opportunities and the ratio of EBITDA to assets (ROA) captures the short-term liquidity necessary for repayment and firm profitability. Stronger financial health and better future opportunities can provide firms with more bargaining power, leading to more renegotiations after positive shocks. In addition, I include the standard deviation of EBITDA/Assets (STDROA), which captures profitability uncertainty. Finally, I include Altman's Z-score (ZSCORE) to proxy for credit risk. Previous literature has shown that riskier firms renegotiate their contracts sooner and more frequently (e.g., Berlin and Mester 1992; Garleanu and Zwiebel 2009; Nikolaev 2013). The terms of the initial contract are also included as controls and consist of the average interest rate spread weighted by the amount of each tranche (INTEREST), the natural log of the loan amount scaled by firm total assets (AMOUNT), the natural log of the maturity (MATURITY), the number of lenders in the lending syndicate (NUMLENDERS), the number of financial covenants (NUMCOV). I also include indicator variables for the presence of a revolving line of credit (REVL V), a borrowing base (BORROWBASE), collateral (SECURED), a pricing grid (PPP),and a prior lending relationship with lenders (RELINT). Loan renegotiations can also be triggered by changes in market conditions. Therefore, I also include two macroeconomic factors. I use both the change in one year treasury yield (A Treasury Yield) and the change in BB-AAA credit spread on publicly traded bonds (A Credit Spread) over the two quarters subsequent to loan origination to measure the change in credit market conditions (Roberts and Sufi 2009b).27 An increase in credit spread indicates a decrease in demand for corporate debt and higher lender bargaining power in the renegotiation One year treasury yield data is from Federal Reserve Bank 1-Year Treasury Constant Maturity Rate. Credit spread data is calculated by subtracting the Bank of America Merrill Lynch US Corporate AAA Effective Yield from the Bank of American Merrill Lynch US High Yield BB Effective Yield. 27 23 process. Finally, I incorporate the Fama-French 12-industry fixed effects to capture the institutional structure of syndicated lending and year fixed effects to capture any common trends in the data. The standard errors are also clustered by firm and year. All continuous variables are winsorized at the top and bottom 1% levels. 3.5 Test of Time lo Renegotiale H1(b) also hypothesizes that firms with more timely reporting of good news are more timely in renegotiating their contracts after a positive credit quality shock. Since the variable of interest is the time to renegotiate, I use survival analysis which is also called time to event analysis. As I only search for renegotiations in the six quarters after a positive shock, my data has right censoring at six quarters. Therefore, I estimate the widely used COX proportional hazard model (Cox 1972), which corrects for right-censoring problem and yields unbiased coefficient estimates of the covariates. 28 Recent treatments of survival analysis tend to focus on the hazard function. The hazard function allows us to approximate the probability of exiting the initial state within a short interval (i.e., hazard rate), conditional on having survived up to the starting time of the interval.2 9 Specifically, I test this hypothesis by estimating the following hazard model: (5) In hi(t) = P(t) + 81G-Scorei + X'4 + el. Wooldrige (2010) specifically explains the survival analysis with right censored data, and shows that hazard model corrects for it (Section 20.3). For programming, I use the stata command "setset TimetoReneg, failure (Reneg)" to correct for right-censoring problem. 29 If the data is with discrete time (time measured in large intervals such as months, quarters, years or even decades), we can get an intuitive idea of the hazard rate. For discrete time the hazard rate is the probability that an individual will experience an event at time t while that individual is at risk for having an event. Thus, the hazard rate is really just the unobserved rate at which events occur. If the hazard rate is constant over time and it was equal to 1.5 for example this would mean that one would expect 1.5 events to occur in a time interval that is one unit long. Furthermore, if a person had a hazard rate of 1.2 at time t and a second person had a hazard rate of 2.4 at time t then it would be correct to say that the second person's risk of an event would be two times greater at time t. It is important to realize that the hazard rate is an un-observed variable yet it controls both the occurrence and the timing of the events. It is the fundamental dependent variable in survival analysis. 28 24 The term hi(t) represents the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t; p(t) is the baseline hazard function. H 1(b) predicts that /3 > 0, i.e., the hazard rate of renegotiation following a positive shock to credit quality increases with the timeliness of good news reporting. Compared to the probit regression, the hazard model uses the information in the timing of the renegotiations rather than just the incidence of the renegotiations, providing more insight about the interaction between the timely reporting of good news and renegotiations following a positive shock to credit quality. The control variables are the same as in the probit regression in Section 3.4.2. 3.6 Cross-Sectional Tests on the Credibility ofAccounting Information My second hypothesis posits that the effects of the timely reporting of good news on renegotiations after positive shocks are more pronounced when the good news can be more credibly communicated. I test this hypothesis using equations (4) and (5), adding interaction terms.30 Renegi =,8o + PIG-Scorei + P2 Crediblei + /3 3G-Score * Credible + X'i + ci. (6) in hi(t) =,8(t) + ,81G-Scorei+ /2 Crediblei + /3G-Score * Credible, + X'1 ( + ci. (7) Crediblei are variables that proxy for the credibility of accounting information. First, I use Restatement to proxy for the financial misreporting. Restatement is an indicator variable that takes the value of one if the firm reports any restatement prior to renegotiation, zero otherwise. If the firm does not report a renegotiation within the search period, Restatement equals one if the Norton et al. (2004) points out and explains why computing the marginal effect of a change in two variables is more complicated in nonlinear models than in linear models. Given this paper, therefore, I use command inteff and margins that computes the correct marginal effect of a change in two interacted variables for a logit or probit model, as well as the correct standard errors. 30 25 firm reported any restatement before the maturity date or the end of the sample period, (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. H2(a) predicts /3 <0. To test H2(b), I define industry specialization as an increasing function of market share. I measure market share using the total sales examined by an audit firm within an industry (Palmrose, 1986). 3I define an industry as all companies within each two-digit primary Standard Industry Classification (SIC) code in the Compustat database. Following the prior literature, I create an indicator variable AudilorSpecialist;it equals one if the client's audit firm examines at least 20% of sales and the largest proportion of sales in the client's two-digit SIC code industry, as reported by Compustat, and zero otherwise (Mayhew and Wilkins 2003; Dunn and Mayhew 2004).32 This variable is measured at the fiscal year-end immediately prior to loan origination. H2(b) predicts #3 > 0. Finally, to test H2(c), I use Transitory to proxy for the degree to which accounting earnings are affected by transitory earnings. Following Collins, Maydew, and Weiss (1997), I calculate Transitory as the ratio of the value of one-time items to the value of core earnings over the two quarters subsequent to loan origination, where core earnings are reported earnings minus one-time items. One-time items refer to discontinued operations, extraordinary items, and special items. This constructs shows that the higher the ratio of Transitory, the greater proportion of transitory earnings might be excluded from reported earnings, leading to fewer renegotiations. Therefore, H2(c) predicts 133 < 0. Appendix I is the list of variable definitions. Prior research shows a positive correlation between audit fees and client size measured by sales or assets (Simunic, 1980; Palmrose, 1986). Therefore, my definition of specialization assumes that audit firns with large market shares, and hence industry specific audit fees, have strong incentives to deliver high quality audits and services. 32 For robustness, I also measure Auditor Specialist in another standard manner. That is, Auditor Specialist takes the value of one if the client's audit finn audits the largest proportion of the sales as well as the largest number of firms in the client's two-digit SIC code industry as reported by Compustat, zero otherwise. 31 26 4. Empirical Results 4.1 Descriptive Statistics Table 2 presents descriptive statistics of all the variables used in both the main and crosssectional tests. Panel A presents the descriptive statistics of full sample, Panel B for subsample of firms that renegotiate their contracts, and Panel C for subsample of firms that do not. The mean value of Reneg indicates that approximately 50% of the sample firms renegotiate their loan contracts with existing lenders after their credit quality improves. 33 Within the sample of firms that renegotiate their contracts, the mean value of TimetoReneg reveals that on average, a borrower initiates a renegotiation 1.58 quarters after the positive shock. The distribution of G-Score is generally consistent with the statistics in Khan and Watts (2009). The distribution of interaction terms shows that on average, 17% of firms have a history of restatement before either renegotiation or the end of sample period and 23% of my sample firms are audited by industry specialists. In addition, the mean value of Transitory reveals that the sample firms have an average 31% of one-time items as a percent of core earnings. Finally, Panels D and E present the distribution of deals across years and industries. Table 3 reports Pearson (above the diagonal) and Spearman (below the diagonal) correlations of the key variables. The panel reveals a positive correlation between G-Score and the likelihood of renegotiations after positive shocks, indicating that firms with more timely reporting of good news are more likely to renegotiate their contracts. The panel also reports a 33 This number is lower than the 90% found in Roberts and Sufi (2009b), mainly because I only explore renegotiation from the end of the positive shock rather than from loan origination. This helps alleviate the endogeneity problem, as it is otherwise difficult to distinguish between a credit quality shock causing a renegotiation and a renegotiation causing a credit quality shock. In addition, Roberts (2012) finds that renegotiations in the early stage of contracts are mostly related to the initial contract's restrictiveness, while later-stage renegotiations are largely triggered by the uncertainty of the borrower's credit quality. 27 negative correlation between Reneg and Restatement, indicating that firms with a misreporting history are less likely to renegotiate their contracts. In addition, the likelihood of renegotiations after positive shocks is positively correlated with the size of the credit quality shock (JBSM Prob), as well as with loan characteristics such as initial interest rates (Interest), prior lending relationship (Relint), the number of financial covenants (Num_Cov), and performance pricing provision (PPP). As TimetoReneg is inversely correlated with Reneg, the relations described above are generally the opposite for TimetoReneg. 4.2 Likelihood and Timeliness of Renegotiations after Positive Shocks Table 4 reports the results of my first set of hypotheses. 34 Consistent with both H 1(a) and Hi (b), the probit model and hazard model regressions show that G-Score is positively associated with both the likelihood and timeliness of renegotiations after positive shocks. These results are also economically significant: a one standard deviation increase in G-Score is associated with a 6.2% increase in the predicted probability of renegotiation and a 2.7% increase in the hazard rate. In terms of the control variables, I find the change in credit quality shock (JBSM Prob) and firm size (LNASSET) have significant positive effects on both the likelihood and timeliness of renegotiations. This is consistent with my prediction, as both a larger improvement in credit quality and firms' being a larger size provide borrowers with more bargaining power in the renegotiation process, leading to more renegotiations. Other firm characteristics do not load in my regressions, which is consistent with Roberts and Sufi (2009b). In terms of loan characteristics, I find that loans with higher interest rates and larger deal amounts, as well as The timely reporting of good news may affect the likelihood of renegotiations through two potential channels. First, the timely reporting of good news may influence the initial contract design choice. Second, the timely reporting of good news influences ex post renegotiations. However, my study focuses on the second channel, so I hold constant the initial design choice by controlling for ex ante loan characteristics in the regression analyses. 1 28 revolving loans, are more likely to be renegotiated. In addition, the coefficients on PPP are positive and significant in both of the table's regressions, implying that loans with performance pricing provisions are more likely to be renegotiated. These results are consistent with Roberts and Sufi (2009b, 181), that demonstrate that "contingencies are designed to shape the renegotiation game rather than to avoid the renegotiation game." However, they also caveat their result: the choice of covenant on which to contract is endogenous with respect to a particular renegotiation outcome. That is, it is possible that a performance pricing provision is included in contracts that are more likely to be renegotiated ex ante.3 5 In general, this table provides strong evidence for my main hypotheses. 4.3 Cross-Sectional Test Results Panel A of Table 5 presents the cross-sectional test results of the probit model regression. The dependent variable in all regressions is an indicator variable (Reneg) that is equal to one if the contract is renegotiated between the end of the positive shock and either maturity or the six quarters after the positive shock, whichever is shorter. Consistent with my second set of hypotheses, I find that the effects of the timely reporting of good news on the likelihood of renegotiations after positive shocks are more pronounced for firms that do not have prior restatements, that are audited by industry specialists, and whose performance improvement is less transitory. When adding all the interaction terms into the regression , the coefficients on each interaction terms remain significant. In terms of economic significance, a one standard deviation increase in G-Score is associated with a 7.6% increase in the predicted probability of renegotiation (Reneg) when firms have no prior restatements, versus a 3.5% decrease for firms that restated. In addition, a one standard deviation increase in G-Score is associated with a 7.7% 35 f performance pricing provisions (PPP) are used to avoid renegotiation, I predict renegotiation would have been even more likely had the PPPnot been incorporated into the contract. 29 increase in Reneg for firms audited by an industry specialist, as opposed to a 5.0% increase for those with non-specialist auditors. Finally, a one standard deviation increase in G-Score is associated with an 8.3% increase in Reneg when the proportion of one-time items is below the sample median, in contrast to a 3.7% increase when it is above. The results of control variables in the main regression (Table 4) generally still hold in these cross-sectional tests. Panel B of Table 5 presents the Cox proportional hazard model of the time to renegotiate (Time to Reneg) for each cross-sectional test. Similar to Panel A, the results about the interaction terms are consistent with my second set of hypotheses, implying that the effects of the timely reporting of good news on the timeliness of renegotiations after positive shocks are more pronounced when the firm does not have a history of financial misreporting, when it is audited by an industry specialist, and when its improvement in performance is less transitory. 4.4 Additional Tests 4.4.1 Economic Impact of Timely Reporting of Good News on Renegotiations after Positive Shocks Up till this point, my paper shows that firms with more timely reporting of good news are more likely to renegotiate their debt after positive shocks, and are likely to renegotiate their debt in a more timely manner. It indicates that the timely reporting of good news can benefit borrowers by increasing the chance of renegotiations. Naturally, it is also interesting to quantify how much firms can benefit from renegotiating their contracts by reporting good news more quickly. Specifically, how are contract terms amended during renegotiations after positive shocks? To investigate this economic impact of timely reporting of good news on renegotiations after positive shocks, I collect renegotiation data in two ways. If only a part of the original loan 30 contract is amended, then an amendment report will be issued and disclosed in the exhibit of a financial report (e.g., 10-K, 10-Q, 8-K). From the amendment report, I can manually collect information on how different contract terms have been changed. In my sample, there are 165 amendment reports issued among 234 renegotiations. In contrast, borrowers and lenders may choose to replace the whole original loan contract with a new one (i.e., restated loan contract), and also attach the restated loan contract in the exhibit of a financial report. 36 In this case, I can rely on the Dealscan database to collect information on how different contract terms have been changed. This is because restatements of original loan contracts generate independent observations in Dealscan, from which I can compare contract terms (e.g., interest rates, loan amount, and maturity) between the original loan contract and the restated loan contract. Appendix II shows an example of an amendment report and the corresponding part from the original contract. In total, four contract terms have been changed in this amendment (i.e., Clayton Williams Energy Inc. - Amendment report on July 1, 2000). First, the performance pricing grid is changed. While the structure of the performance pricing grid is not changed, the interest rate in each grid has been reduced by 50 bps. This is equivalent to the fact that the borrower can obtain a 50 bps lower interest rate after the renegotiation. Second, the maturity has been extended by one year till July 31, 2002. Third, the limitation on the sale of collateral has been relaxed from $500,000 to $1,000,000. Lastly, the restriction on investment has also been relaxed so that the borrower can make up to $1,000,000 other investments. From this example, we can see that the amendment of original contracts can be very diverse, involving changes in different kinds of contract terms. Whether issuing an amendment report or a restated loan depends on how material the changes to original loan contracts are. In addition, renegotiations must be approved by a certain percentage of lenders, depending on the proposed changes. Amendments usually require a simple majority (typically 51% of the votes). Restated loan contracts usually involve changes to the rate, term, or collateral which are considered material changes, and therefore often require a unanimous vote (Standard & Poor's 2006; Roberts and Sufi 2009b; Nikolaev 2013). 36 31 Panel A of Table 6 summarizes the occurrence of change in each contract term in 165 amendment reports. The two most frequent changes to original loan contracts are changes to loan amount, in 49% of contracts, and changes to financial covenants (e.g., minimum EBITDA, minimum interest coverage, maximum leverage), in 41% of contracts. In addition, 52% of amendments charge an amendment fee, which is consistent with the argument that renegotiation is not costless (Roberts and Sufi 2009b). Lenders need to spend time and effort in understanding the new condition of borrowers, so they usually charge an amendment fee during renegotiations. I notice that the percentages of changes to interest rates and changes to maturity are relatively small, only 17% and 15%, respectively. This is because changes to interest rates and maturity are typically considered material changes and often require a unanimous vote, so borrowers and lenders may choose to restate the loan contract if they decide to change them (Roberts and Sufi 2009b; Nikolaev 2013). My data on restated loans show that 65% of them involve changes to interest rate and 85% involve changes to maturity. While changes to contract terms during renegotiations can be very diverse, in this paper I focus on investigating changes to three contract terms during renegotiations - interest rate, loan amount, and loan maturity. This is mainly because changes to these three contract terms are generally considered material changes, and they are also easy to quantify. This method is also consistent with Roberts and Sufi (2009b), which focus on examining renegotiations of these three contract terms. Panel B of Table 6 reports the frequency of the occurrence of changes to these three major contract terms. Among 234 firms that report a renegotiation (either issuing an amendment report or issuing a restated loan) after positive shocks, 30% of firms receive a reduction in interest rates, 38% of firms receive an increase in loan amount, and 34% of firms extend their 32 maturity. A few firms receive seemingly unfavorable outcomes (an increase in interest rates, a decrease in loan amount, or a shortened maturity) during renegotiations. There may be several reasons. For instance, borrowers want to reduce commitment fees on the unused portion of any credit line. Therefore, they may be motivated to reduce the loan amount. Banks, however, prefer firms to maintain the original size and therefore may increase the interest rates to compensate for this reduction in interest rates. In addition, banks may also charge higher interest rates if borrowers want to shorten the loan maturity (Roberts and Sufi 2009b). However, these cases are uncommon, especially increases in interest rates (2%) and shortening of maturity (1%), indicating that firms usually obtain more favorable outcomes during renegotiations after positive shocks. Panel C of Table 6 presents descriptive statistics of changes to these three contract terms. On average, firms can receive a reduction of 86 bps in interest rates (relative to the average initial interest rate of 265.71 bps), an increase of 53% of original loan amount, or an increase of 92% in original loan maturity during renegotiations. Thus, my results indicate a large economic significance of reporting of good news, and suggest that by increasing the chance of a renegotiation, firms with more timely reporting of good news are likely to obtain considerably better contract terms. 4.4.2 Implicationsfor Accounting Conservatism While my paper examines the impact of timely reporting of good news in debt contracting, it has important implications for accounting conservatism (i.e., the asymmetric timeliness of bad news recognition). Conceptually, the timely reporting of good news is not necessarily inversely related to accounting conservatism. In my study, however, the two constructs are negatively correlated (Table 3). This suggests that more conservative firms tend to 33 delay reporting good news, and thus are less likely to renegotiate their loans after positive shocks. Consistent with my prediction, I find that the estimated coefficients on C-Score (i.e., Khan and Watts (2009) measure of the asymmetric timeliness of bad news recognition) are significantly negative, indicating that after positive shocks to their credit quality, more conservative firms are less likely to renegotiate their loan contracts and when they do, to do so later (Table 7). This finding provides a rationale for prior findings on the ex ante benefits of accounting conservatism (Zhang 2008; Wittenberg-Moerman 2008). Specifically, to the extent that these less likely and less timely renegotiations may benefit lenders ex post by protecting them from interest rates reductions, one would expect lenders to share these benefits by charging lower interest rates ex ante (i.e., at the time of loan origination) to more conservative firms. In addition, this finding provides a nice complement to the existing research on accounting conservatism. Previous literature (Zhang 2008) shows that conservative reporting benefits lenders by accelerating covenant violations after negative shocks, whereas my paper indicates that conservative reporting benefits lenders by delaying renegotiations after positive shocks. Therefore, my paper contributes to the literature on accounting conservatism by enhancing our understanding of the complete role of accounting conservatism in debt contracting. 4.4.3 Nature of Good News Received An alternative interpretation of the findings in this paper is that it is the variation in the nature of information received, rather than the variation in the timeliness with which firms incorporate good news into financial statements, that influences contract renegotiations following positive shocks. For example, for high growth firms, a greater proportion of their market value shock may be due to growth options which cannot be incorporated into financial statements, leading to lower probability of renegotiations. Another example could be firms in 34 R&D-intensive industries (e.g., software, biotechnology and telecommunications). This is because the granting of a patent to these firms could induce a positive shock to their credit quality, which also cannot be incorporated into accounting records in a timely manner, leading to fewer renegotiations. To address these concerns, I conduct two sensitivity tests (Table 8). First, I re-run probit and hazard model regressions only on firms that have no increase in market-tobook ratio (my proxy for expansion in investment opportunity). All results still hold. Second, I eliminate firms in four R&D-intensive industries with two-digit SIC codes of 28, 35, 36 and 38 from my sample (Lev and Sougiannis 1996; Shi 2003; Jones 2010).37 I re-run both regressions and also find robust results. 4.4.4 Alternative Measure of Timely Reporting of Good News I modify the approach in Ball et al. (2008) to measure how well accounting numbers predict future credit rating upgrades (Table 9). The debt-contracting value of accounting in the context of credit rating upgrades, DCV Upgrade, is calculated at the industry level. Specifically, for any given year, I estimate a probit model using quarterly data in the past five years for each Fama-French industry (48 categories): P(Upgrade,, = 1) = P(ao + aiAE,_.,i + a2AE -2. + a3 AEt-3 ,i + a 4AEt-4 ) (10) Where Upgrade is an indicator variable equal to one if firm i's credit rating is upgraded in the current quarter I and equal to zero otherwise. AE-k.i is the seasonally adjusted change in quarterly earnings before extraordinary items scaled by total assets in the kth quarter prior to the current quarter, t. Somers' D, also known as the accuracy ratio, is a popular statistic that is used to measure the quality of credit-rating systems. Somers' D measures the extent of concordance between the model-predicted upgrades and the actual upgrades. The higher the Somers' D, the Four R&D-intensive industries are chemicals and pharmaceuticals (Standard Industrial Classification [SIC] 28), machinery and computer hardware (SIC 35), electrical and electronics (SIC 36), and scientific instruments (SIC 38). 3 35 higher the prediction ability of earnings changes (Altman and Sabato 2007). This measure captures how well accounting numbers predict future credit rating, implicitly signaling how quickly accounting numbers reflect current good news. Replacing G-Score with DCV Upgrade does not change my results. 4.4.5 Timely Reporting of Good News Interactedwith the Size of Positive Shock One way to confirm the mechanism that financial statements provide verification value to lenders and thereby affect renegotiations is to have the timely reporting of good news interacted with the size of positive shock. It can be expected that the larger the size of positive shock, the more pronounced are the effects of timely reporting of good news on the likelihood of renegotiations. This is because the larger the size of positive shock, the more the gains that can be incorporated in financial statements if firms report good news in a timely manner, leading to higher probability of triggering renegotiations. I find results consistent with my prediction, implying that accounting properties do play an important role in contract renegotiations (Table 10). 4.4.6 Firm Loans without Earnings-BasedPerformancePricingProvision According to Asquith, Beatty, and Weber (2005), in a performance pricing contract the lender and the borrower agree at loan inception to a pricing grid that designates how interest rates will change in response to changes in financial performance. The pricing grid provides important mechanisms that can reduce the re-contracting costs associated with loans, namely, reducing the probability of renegotiations, especially those related to interest rate changes. Therefore, if the loan includes an earnings-based performance pricing grid (e.g., debt-toEBITDA ratio, interest coverage, fixed charge coverage), it is less likely that the timely reporting of good news in accounting earnings will affect renegotiations after positive shocks. 36 Therefore, I exclude loans that have earnings-based performance pricing (21 loans) and re-run the probit and hazard model regressions. All results still hold (Table 11). 4.4.7 Switch of Lenders Although I examine the renegotiation of existing loan contracts, one concern is that as firms have outside financing options, they may switch to outside banks to refinance their current loans as opposed to renegotiating with existing lenders after the credit quality improves. To alleviate this concern, I identify new loans that are "refinancing from outside lenders" as renegotiations and re-run the probit and hazard model regressions. All results still hold (Table 12). 4.4.8 Controlfor Changes in Firm Characteristics Roberts and Sufi (2009b) find that ex post changes in firm-specific variables are important determinants of renegotiations (e.g., LNASSET, LEV). As I already include the change in the probability of bankruptcy (JBSM-Prob), which is a comprehensive proxy for the change in credit quality, I do not expect the changes in firm-specific variables to affect my previous results. Nevertheless, I still measure the change in each of the firm-specific characteristic variables from loan origination to the two quarters after origination and add them to my regressions. Consistent with my prediction, all results still hold, and the coefficients on these change variables are not statistically significant (Table 13). 5. Conclusion Renegotiations play a critical role in corporate financial contracting. Despite the extensive research on renegotiations in the context of corporate default, recent empirical studies show that numerous renegotiations are voluntarily initiated by borrowers, especially when the 37 borrower's credit quality improves. In my paper, I extend this literature by investigating the role of accounting in renegotiations after positive shocks. Using a hand-collected sample of private debt contracts between U.S. publicly traded firms and financial institutions, I examine whether and how the timely reporting of good news affects the renegotiation of debt contracts following a positive shock to the borrower's credit quality. I find that firms with more timely reporting of good news are more likely to renegotiate their loan contracts and they do so sooner than firms with less timely good news reporting. Further, these effects are more pronounced for firms whose positive shocks can be more credibly communicated through financial reporting. My paper contributes to the literature on the role of accounting in incomplete contracting. In particular, my paper fills a gap in the literature by examining whether and how accounting information, proxied by the timely reporting of good news, plays a critical role in renegotiations following a positive shock to a borrower's credit quality. In addition, an implication from my paper is that conservative firms tend to delay disclosing good news and thus are less likely to renegotiate their contracts after positive shocks. To the extent that this benefits lenders ex post, my paper provides a rationale for the evidence in previous literature on the ex ante benefits of accounting conservatism, i.e., lower initial interest rates charged for firms with delayed good news disclosure (Zhang 2008; Wittenberg-Moerman 2008). 38 References Aghion, P., and P. Bolton. 1992. An Incomplete Contracts Approach to Financial Contracting. The Review of Economic Studies 59(3): 473-94. Ahmed, A. S., B. K. Billings, R. M. Morton, and M. Stanford-Harris. 2002. The Role of Accounting Conservatism in Mitigating Bondholder-Shareholder Conflicts over Dividend Policy and in Reducing Debt Costs. The Accounting Review 77(4): 867-90. Altman, E. I. 1968. Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. The Journalof Finance23(4): 589-609. Armstrong, C. S., W. R. Guay, and J. P. Weber. 2010. The Role of Information and Financial Reporting in Corporate Governance and Debt Contracting. Journal of Accounting and Economics 50(2-3): 179-234. Asquith, P., A. Beatty, and J. P. Weber. 2005. Performance Pricing in Bank Debt Contracts. JournalofAccounting and Economics 40(1-3): 101-28. Baber, W. R., S. Kang, and K. R. Kumar. 1998. Accounting Earnings and Executive Compensation: The Role of Earnings Persistence. Journal of Accounting and Economics 25(2): 169-93. Ball, R. 2001. Infrastructure Requirements for an Economically Efficient System of Public Financial Reporting and Disclosure. Brookings-Wharton Papers on Financial Services 2001(1): 127-69. Ball, R., S. Jayaraman, and L. Shivakumar. 2012. Audited Financial Reporting and Voluntary Disclosure as Complements: A Test of the Confirmation Hypothesis. JournalofAccounting and Economics 53(1-2): 136-66. Ball, R., and L. Shivakumar. 2005. Earnings Quality in UK Private Firms: Comparative Loss Recognition Timeliness. JournalofAccounting and Economics 39(1): 83-128. Ball, R., R. M. Bushman, and F. P. Vasvari. 2008. The Debt-Contracting Value of Accounting Information and Loan Syndicate Structure. JournalofAccounting Research 46(2): 247-87. Balsam, S., J. Krishnan, and J. S. Yang. 2003. Auditor Industry Specialization and Earnings Quality. AUDITING: A Journal of Practice& Theory 22(2): 71-97. Basu, S. 1997. The Conservatism Principle and the Asymmetric Timeliness of Earnings. Journal ofAccounting and Economics 24(1): 3-37. 39 Berlin, M., and L. J. Mester. 1992. Debt Covenants and Renegotiation. Journal of Financial Intermediation 2(2): 95-133. Bharath, S.T., S. Dahiya, A. Saunders, A. Srinivasan. 2011. Lending Relationships and Loan Contract Terms. Review of FinancialStudies, 24(4), pp. 1141-1203. Bolton, P., and D. S. Scharfstein. 1996. Optimal Debt Structure and the Number of Creditors. Journal of PoliticalEconomy 104(1): 1-25. Burns, N., and S. Kedia. 2006. The Impact of Performance-based Compensation on Misreporting. Journal of FinancialEconomics 79(1): 3 5-67. Bushman, R. M., and J. D. Piotroski. 2006. Financial Reporting Incentives for Conservative Accounting: The Influence of Legal and Political Institutions. Journal of Accounting and Economics 42(1-2): 107-48. Cassar, G. 2011. Discussion of the Value of Financial Statement Verification in Debt Financing: Evidence from Private U.S. Firms. Journal ofAccounting Research 49(2): 507-28. Christensen, H. B., and V. Nikolaev. 2012. Capital versus Performance Covenants in Debt Contracts. Journal ofAccounting Research 50(1): 75-116. Collins. D. W., E. L. Maydew, and I. S. Weiss. 1997. Changes in the Value-relevance of Earnings and Book Values over the Past Forty Years. Journal of Accounting and Economics 24(1): 39-67. Costello, A. M., and R. Wittenberg-Moerman. 2011. The Impact of Financial Reporting Quality on Debt Contracting: Evidence from Internal Control Weakness Reports. Journal of Accounting Research 49(1): 97-136. Cox, D. R. 1972. Regression Models and Life-Tables. Journal of the Royal Statistical Society. Series B (Methodological)34(2): 187-220. Daniel, N. D., D. J. Denis, and L. Naveen. 2008. Do Firms Manage Earnings to Meet Dividend Thresholds? JournalofAccouniing and Economics 45(1): 2-26. DeAngelo, L. E. 1981. Auditor Size and Audit Quality. Journal of Accounting and Economics 3(3): 183-99. Dechow, P., W. Ge, and C. Schrand. 2010. Understanding Earnings Quality: A Review of the Proxies, Their Determinants and Their Consequences. Journal of Accounting and Economics 50(2-3): 344-401. 40 DeFond, M., and J. Jiambalvo. 1994. Debt Covenant Violation and Manipulation of Accruals. JournalofAccounting and Economics 17(1-2): 145-76. DeFond, M., and J. Zhang. 2013. A Review of Archival Auditing Research. Working Paper. Dichev, I. D., and D. J. Skinner. 2002. Large-Sample Evidence on the Debt Covenant Hypothesis. JournalofAccounting Research 40(4): 1091-1123. Dou, Y. 2013. The Debt-Contracting Value of Accounting Numbers, Renegotiation, and Investment Efficiency. Working Paper. Dunn, K. A., and B. W. Mayhew. 2004. Audit Firm Industry Specialization and Client Disclosure Quality. Review ofAccounting Studies 9(l): 35-58. Ertimur, Y. 2004. Accounting Numbers and Information Asymmetry: Evidence from Loss Firms. Working Paper. Financial Accounting Standards Board (FASB). 1980. Qualitative Characteristics of Accounting Information. Stamford, CT: FASB. Garleanu, N., and J. Zwiebel. 2009. Design and Renegotiation of Debt Covenants. Review of FinancialStudies 22(2): 749-81. Gorton, G., and J. Kahn. 2000. The Design of Bank Loan Contracts. Review of FinancialStudies 13(2): 331-64. Graham, J. R., S. Li, and J. Qiu. 2008. Corporate Misreporting and Bank Loan Contracting. JournalofFinancialEconomics 89(1): 44-61. Guay, W., and R. Verrecchia. 2006. Discussion of an Economic Framework for Conservative Accounting and Bushman and Piotroski (2006). Journal of Accounting and Economics 42(1-2): 149-65. Hart, 0., and J. Moore. 1988. Incomplete Contracts and Renegotiation. Econometrica 56(4): 755-85. Hart, 0., and J. Moore. 1998. Default and Renegotiation: A Dynamic Model of Debt. The QuarterlyJournalof Economics 113(1): 1-41. Healy, P. M., and K. G. Palepu. 1990. Effectiveness of Accounting-based Dividend Covenants. Journal ofAccounting and Economics 12(1-3): 97-123. Hennes, K. M., A. J. Leone, and B. P. Miller. 2008. The Importance of Distinguishing Errors from Irregularities in Restatement Research: The Case of Restatements and CEO/CFO Turnover. The Accounting Review 83(6): 1487-1519. 41 Hillegeist, S. A., E. K. Keating, D. P. Cram, and K. G. Lundstedt. 2004. Assessing the Probability of Bankruptcy. Review ofAccounting Studies 9(1): 5-34. Hribar, P., and N. T. Jenkins. 2004. The Effect of Accounting Restatements on Earnings Revisions and the Estimated Cost of Capital. Review ofAccounting Studies 9(2-3): 337-56. Khan, M., and R. L. Watts. 2009. Estimation and Empirical Properties of a Firm-year Measure of Accounting Conservatism. Journal ofAccounting and Economics 48(2-3): 132-50. Kim, J., D. A. Simunic, M. T. Stein, and C. H. Yi. 2011. Voluntary Audits and the Cost of Debt Capital for Privately Held Firms: Korean Evidence. Contemporary Accounting Research 28(2): 585-615. Kothari, S. P., K. Ramanna, and D. J. Skinner. 2010. Implications for GAAP from an Analysis of Positive Research in Accounting. Journal of Accounting and Economics 50(2-3): 24686. Kravet, T., and T. Shevlin. 2010. Accounting Restatements and Information Risk. Review of Accounting Studies 15(2): 264-94. Krishnan, G. V. 2003. Does Big 6 Auditor Industry Expertise Constrain Earnings Management? Accounting Horizons 17: 1-16. Leftwich, R.. 1983. Accounting Information in Private Markets: Evidence from Private Lending Agreements. The Accounting Review 58(1): 23-42. / Levitt, A. 2000. Renewing the Covenant with Investors. SEC Speech. (http: // www.sec.gov news / speech / spch370.htm). Li, N. 2010. Negotiated Measurement Rules in Debt Contracts. Journalof Accounting Research 48(5): 1103-44. Maines, L. A., and J. M. Wahlen. 2006. The Nature of Accounting Information Reliability: Inferences from Archival and Experimental Research. Accounting Horizons 20(4): 399425. Maskin, E., and J. Moore. 1999. Implementation and Renegotiation. The Review of Economic Studies 66(1): 39-56. McVay, S. E. 2006. Earnings Management Using Classification Shifting: An Examination of Core Earnings and Special Items. The Accounting Review 81(3): 501-31. Minnis, M. 2011. The Value of Financial Statement Verification in Debt Financing: Evidence from Private U.S. Firms. JournalofAccounting Research 49(2): 457-506. 42 Nikolaev, V. 2013. Scope for Renegotiation in Private Debt Contracts. Working Paper. Norton, E.C., Wang, H. & Ai, C., 2004. Computing Interaction Effects and Standard Errors in Logit and Probit Models. The StataJournal, 2004(2), pp. 1 54-167. Ohlson, J. A. 1999. On Transitory Earnings. Review ofAccounting Studies 4(3-4): 145-62. Palmrose, Z. 1986. The Effect of Nonaudit Services on the Pricing of Audit Services: Further Evidence. JournalofAccounting Research 24(2):405-11. Petersen, M. 2004. Information: Hard and Soft. Mimeo, Nothwestern University. Ramakrishnan, R. T. S., and J. K. Thomas. 1998. Valuation of Permanent, Transitory, and PriceIrrelevant Components of Reported Earnings. Journal of Accounting, Auditing & Finance 13(3): 301-36. Roberts, M. R. 2012. The Role of Dynamic Renegotiation and Asymmetric Information in Financial Contracting. Working Paper. Roberts, M. R., and A. Sufi. 2009a. Financial Contracting: A Survey of Empirical Research and Future Directions. Annual Review of FinancialEconomics 1(1): 207-26. . 2009b. Renegotiation of Financial Contracts: Evidence from Private Credit Agreements. Journalof FinancialEconomics 93(2): 159-84. Simunic, D. A. 1980. The Pricing of Audit Services: Theory and Evidence. Journal of Accounting Research 18(1): 161-90. Sloan, R. G. 1996. Do Stock Prices Fully Reflect Information in Accruals and Cash Flows About Future Earnings? The Accounting Review 71(3): 289-315. Sweeney, A. P. 1994. Debt-covenant Violations and Managers' Accounting Responses. Journal ofAccounting and Economics 17(3): 281-308. Tan, L. 2013. Creditor Control Rights, State of Nature Verification, and Financial Reporting Conservatism. JournalofAccounting and Economics 55(1): 1-22. Watts, R. L., and J. L. Zimmerman. 1986. Positive Accounting Theory. Prentice-Hall, Englewood Cliffs, NJ. Wittenberg-Moerman, R. 2008. The Role of Information Asymmetry and Financial Reporting Quality in Debt Trading: Evidence from the Secondary Loan Market. Journal of Accounting and Economics 46(2-3): 240-60. Wooldrige, J.M. 2010. Econometric Analysis of Cross Section and Panel Data. Massachusetts Institute of Technology. C 2010, 2002 43 Wright, S., and A. M. Wright. 1997. The Effect of Industry Experience on Hypothesis Generation... BehavioralResearch in Accounting 9: 273. Zhang, J. 2008. The Contracting Benefits of Accounting Conservatism to Lenders and Borrowers. JournalofAccounting and Economics 45(1): 27-54. 44 Appendix I: Variable Definitions Variable Dependent variables Reneg Time toReneg Treatment variables G-Score and C-Score Definition Indicator variable that equals one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. The number of quarters between the end of the positive shock and renegotiation. If the firm does not report a renegotiation within the search period, TimetoReneg is the number of quarters between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter. Khan and Watts (2009) Measure: Xi=Pi+P2Di43R+ DRiei; G-Score=p3=i+ji 2 Sizei+p 3 M/Bi+p4 Levi; C-Score=P4=3%+X2Sizei+k3M/Bi+4Levi. Where X is earnings scaled by the lagged market value of equity, R is annual returns compounded from monthly returns beginning the fourth month after fiscal year-end. D is a dummy variable equal to one if returns (R) are negative, and zero if returns are positive. M/B is the ratio of the market value of equity to the book value of equity at the end of the year. Size is the natural log of the market value of equity. Lev is leverage, defined as long-term and short-term debt deflated by the market value of equity. DCV Upgrade + U3 AEt 3,i + a 4 AEt4,i) Where Upgrade is an indicator variable equal to one if firm i's credit rating is upgraded in the current quarter t and equal to zero otherwise. AE,ki is the seasonally adjusted change in quarterly earnings before extraordinary items scaled by total assets in the kth quarter prior to the current quarter, t. Somers' D, also known as the accuracy ratio, is a popular statistic that is used to measure the quality of credit-rating systems. Restatement Indicator variable that takes the value of one if the firm reported any restatement prior to renegotiation, and zero otherwise. If the firm does not report a renegotiation within the search period, Restatement equals one if the firm reported any restatement before the maturity date, or the end of the sample period, (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. P(Upgradeti= 1)= cD(ao + ai AEt-, i+ U2 AEt-2, 45 A udilor Specialist Indicator variable that equals one if the client's audit firm audits at least 20% of the sales and also the largest proportion of the sales in the client's two-digit SIC code industry as reported by Compustat, zero otherwise Transitory The ratio of the value of one-time items to the value of core earnings, where core earnings are reported earnings minus one-time items. One-time items refer to discontinued operations, extraordinary items, and special items. Hillegeist et al.'s (2004) probability of bankruptcy (BSM-Prob) is based on Black-Scholes-Merton option-pricing model at firm-year level. I compute this measure at firm-quarter level by replacing their yearly variables with quarterly variables. ABSMProb is the absolute value of changes in the probability of bankruptcy in the two quarters after the origination of loan contracts. BSM-Prob = N(- UA VT ) Firm-specific variables ABSM Prob Where VA is the market value of assets which is defined as total liabilities plus the market value of equity; uA is the standard deviation of asset returns; It is the expected return on assets; 6 is the dividend rate which is calculated as the sum of the prior year's common and preferred dividends divided by the market value of assets VA; X is the face value of debt maturing at time T. The average of natural log of book assets (atq) over quarter q - 3 to q, where q is the quarter before loan origination. DTE The average of the debt-to-EBITDA ratio ((dlcq + dlttq)/oibdpq) over quarter q - 3 to q, where q is the quarter before loan origination. LEV The average of the debt-to-book-assets ratio ((dlcq + dlttq)/atq) over quarter q - 3 to q, where q is the quarter before loan origination. MTB The average of the market-to-book ratio (((ltq + pstkl - txditcq) (prccq*cshoq))/atq) over quarter q - 3 to q, where q is the quarter before loan origination. ROA The average of the EBITDA-to-book assets ratio (oibdpq/atq) over quarter q - 3 to q, where q is the quarter before loan origination. + LNASSET 46 The standard deviation of the EBITDA-to-book assets ratio (oibdpq/atq) over the past eight quarters. ZSCORE The average Z-score (1.2*((actq - lctq)/atq) + 1.4*(req/atq) 3.3*(piq/atq) + 0.6*((prccq*cshoq)/ltq) + 0.999*(saleq/atq)) over quarter q - 3 to q, where q is the quarter before loan origination. A LNASSET Change in the value of LNASSET in the two quarters after the origination of loan contracts. J DTE Change in the value of DTE in the two quarters after the origination of loan contracts. JLEV Change in the value of LEV in the two quarters after the origination of loan contracts. JMTB Change in the value of MTB in the two quarters after the origination of loan contracts. JROA Change in the value of ROA in the two quarters after the origination of loan contracts. ASTDROA Change in the value of STD ROA in the two quarters after the origination of loan contracts. AZSCORE Change in the value of ZSCORE in the two quarters after the origination of loan contracts. + STD_ROA Macroeconomic variable 4 Credit Spread The value of change in the BB-AAA credit spread on publicly traded bonds over the two quarters subsequent to loan origination. Credit spread data is calculated by subtracting the Bank of America Merrill Lynch US Corporate AAA Effective Yield from the Bank of American Merrill Lynch US High Yield BB Effective Yield. A Treasury Yield Loan-specific variables Interest Spread Change in one year treasury yield, which data is from Federal Reserve Bank 1-Year Treasury Constant Maturity Rate. The average all-in-drawn spread over LIBOR for all tranches in the deal, weighted by the amount of each tranche. AInterest Change in interest rate spread during renegotiations. Amount The sum of the amounts of all tranches in each deal scaled by total assets. 47 zAimount Percentage change in loan amount from original loan amount during renegotiations. Log(Amount) The natural log of the sum of the amounts of all tranches in each deal scaled by total assets. Maturity The length of the loan in months of all tranches in the deal, weighted by the amount of each tranche. AMaturity Percentage change in loan maturity from original loan maturity during renegotiations. Log(Maturity) The natural log of the length of the loan in months of all tranches in the deal, weighted by the amount of each tranche. NumLenders The number of lenders in lending deal. Revolver Indicator variable that equals one if the deal involves a revolver loan. Secured Indicator variable that equals one if the deal is secured and zero otherwise. PPP Indicator variable that equals one if the deal involves a performance pricing provision and zero otherwise. Relint Indicator variable that equals one if there is a prior lending relationship with the lender for the past five years, zero otherwise. Num Cov The number of financial covenants. 48 Appendix II: Example of Amendment Report Clayton Williams Energy Inc. - Excerpts from Amendment Report on July 1, 2000 1. Section 1(v) is hereby deleted in its entirety and the following inserted in lieu thereof: "(v) EURODOLLAR MARGIN - the fluctuating Eurodollar Margin in effect from day to day, shall be: (i) two percent (2%) per annum whenever the Total Outstandings are greater than 75% of the Elected Borrowing Limit in effect at the time in question; (ii) one and three-quarters percent (1.75%) per annum whenever the Total Outstandings are greater than 51%, but less than or equal to 75%, of the Elected Borrowing Limit in effect at the time in question; (iii) one and one-half percent (1.50%) per annum whenever the Total Outstandings are greater than 25%, but less than or equal to 50%, of the Elected Borrowing Limit in effect at the time in question; (iv) one and one-fourth percent (1.25%), whenever the Total Outstandings are 25% or less of the Elected Borrowing Limit in effect at the time in question." The corresponding part from the original contract: (v) Eurodollar Margin - The fluctuating Eurodollar Margin in effect from day to day shall be: (i) two and one-half percent (2.50%) per annum whenever the Total Outstandings are greater than 75% of the Elected Borrowing Limit in effect at the time in question; (ii) two and one-quarter percent (2.25%) per annum whenever the Total Outstandings are greater than 50%, but less than or equal to 75%, of the Elected Borrowing Limit in effect at the time in question; (iii) two percent (2%) per annum whenever the Total Outstandings are greater than 25%, but less than or equal to 50%, of the Elected Borrowing Limit in effect at the time in question; (iv) one and three-quarters percent (1.75%), whenever the Total Outstandings are 25% or less of the Elected Borrowing Limit in effect at the time in question. 2. Section 1 (jj) is hereby deleted in its entirety and the following inserted in lieu thereof: "(jj) MATURITY DATE - July 31, 2002." The corresponding part from the original contract: (jj) Maturity Date - July 31, 2001. 3. Section 13(e) is hereby deleted in its entirety and the following inserted in lieu thereof: 49 "(e) LIMITATION ON SALE OF COLLATERAL. Neither Borrower nor Guarantor will sell, assign or discount any of the Collateral or Negative Pledge Property other than (i) sales of oil and gas production in the ordinary course of business, (ii) sales or other disposition of obsolete equipment which are no longer needed for the ordinary business of Borrower or Guarantor or which are being replaced by equipment of at least comparable value and utility, and (iii) sales or other dispositions not exceeding $1,000,000 in the aggregate between Borrowing Base redeterminations. If and as any of such Collateral or Negative Pledge Properties and interests are sold, conveyed or assigned during the term of the Revolving Commitment, Borrower or Guarantor will prepay against the Notes or Guarantor's obligation under its guaranty agreement, as the case may be, 100% of the Release Price, provided, however, that no such payments shall be required from Borrower or Guarantor until the aggregate proceeds received between any Borrowing Base redetermination exceeds $1,000,000. Provided, however, notwithstanding the foregoing, if an Event of Default has occurred and is continuing all such amounts received by Borrower and/or Guarantor from such sale during the continuance of an Event of Default shall be paid to the Agent for the ratable benefit of the Banks. The term "Release Price" as used herein shall mean the loan value of the Collateral or the Negative Pledge Property being sold as determined by the Agent. Any such prepayment of principal on the Notes required by this Section 13(e) shall not be in lieu of, but shall be in addition to, any Monthly Commitment Reduction or any mandatory prepayment of principal required to be made pursuant to Section 9(b) hereof. Any such prepayment shall be applied pro rata to the principal due on the Revolving Notes until such Revolving Notes are paid in full, principal, interest and other amounts." The corresponding part from the original contract: (e) Limitation on Sale of Collateral. Neither Borrower nor Guarantor will sell, assign or discount any of the Collateral or Negative Pledge Property other than (i) sales of oil and gas production in the ordinary course of business, and (ii) sales or other disposition of obsolete equipment which are no longer needed for the ordinary business of Borrower or Guarantor or which are being replaced by equipment of at least comparable value and utility. If and as any of such Collateral or Negative Pledge Properties and interests are sold, conveyed or assigned during the term of the Revolving Commitment, Borrower or Guarantor will prepay against the Notes or Guarantor's obligation under its guaranty agreement, as the case may be, 100% of the Release Price. The term "Release Price" as used herein shall mean the loan value of the Collateral or the Negative Pledge Property being sold as determined by the Agent. Any such prepayment of principal on the Notes required by this Section 13(e) shall not be in lieu of, but shall be in addition to, any Monthly Commitment Reduction or any mandatory prepayment of principal required to be made pursuant to Section 9(b) hereof. Any such prepayment shall be applied pro rata to the principal due on the Revolving Notes until such Revolving Notes are paid in full, principal, interest and other amounts. Provided, however, that the Borrower and Guarantor may, without consent of Banks and Agent and without prepaying the Notes, sell Negative Pledge Properties where the sales proceeds from any such sale do not exceed $500,000 on an annual basis. 50 4. Section 13(k) is hereby amended by deleting the period (".") at the end of Subsection 13(k)(vii) and substituting a semi-colon and the word "and" ("; and") in lieu thereof and by the addition of a new Subsection (viii) thereto as follows: "(viii) other investments not exceeding $1,000,000 in the aggregate made from and after the First Amendment Effective Date." The corresponding part from the original contract: 13(k) Investments. Neither Borrower nor Guarantor shall make any investments in any person or entity, except that the foregoing restriction shall not apply to: (i) investments and direct obligations of the United States of America or any agency thereof; (ii) investments in certificates of deposit issued by the Agent or certificates of deposit with maturities of less than one year issued by other commercial banks in the United States having capital and surplus in excess of $500,000,000 and have a rating of (A) 50 or above by Sheshunoff and (B) "B" or above by Keef-Bruett; (iii) investments such as insured money market funds, Eurodollar investment accounts and other similar accounts with the Agent or such investments with maturities of less than ninety (90) days at other commercial banks in the United States having capital and surplus in excess of $500,000,000 and having a rating of (A) 50 or above by Sheshunoff and (B) "B" or above by Keef-Bruett; (iv) investments in the Subsidiaries (other than Guarantor) existing on the Effective Date; (v) Borrower's investments in Guarantor; (vi) the repurchase of Borrower's stock as permitted by Section 13(j) hereof; and (vii) Guarantor's initial investment of $425,000 in Clayton Williams Acquisition Partnership, Ltd. ("Acquisition") and additional investments in Acquisition equal to one percent (1 %) of all capital contributions made to Acquisition by its general and limited partners, but only to the extent such contributions are required by the partnership agreement of Acquisition, provided that immediately before and after giving effect to such additional investments no default or Event of Default shall exist. 51 Figure 1 Timeline of the Measurement Period/lbrthe Variables JBSM-Prob A Treasury Yield ACredit Spread Transitory Time _toReneg 2 quarters Loan origination End of positive shock Renegotiation End of sample period 4', G-Score, C-Score, DCV Upgrade are and A uditor Specialist measured at the fiscal year-end to loan prior immediately origination. 2. variables are Firm-specific measured at the quarter-end prior immediately to loan origination. 3. Loan-specific variables are taken original loan from the agreements. 4. If firms report a renegotiation, then Restatement is measured at the year-end before renegotiation. Otherwise, Restatement is measured at the year-end before maturity or the end of sample period, whichever is shorter. 52 Table 1 Sample Selection No. of Obs. Remaining Selection Criteria Firm loans that experienced a decrease in the probability of bankruptcy for the years 1996-2005 3,127 Firm loans that experienced at least a 1% decrease in BSM-Prob 1,195 Firm loans with available data to calculate loan characteristics 786 Firm loans with data available to calculate firm characteristics and required accounting attributes 466 466 Final Sample * Among the 466 firm loans in the final sample, 230 disclose a renegotiation in their 10-K, 10-Q, or 8-K filings within sample period. * As the starting year for the restatement data is 1998, the sample for testing H2(a) is reduced to 409 firms. 53 Table 2 Descriptive Statistics Variable Dependent variables Reneg TimetoReneg (Quarters) Treatment variables G-Score C-Score DCV Upgrade Restatement Auditor Specialist Transitory Firm-specific control variables JBSM Prob LNASSET DTE LEV MTB ROA STDROA ZSCORE Macroeconomic control variable ACredit Spread A TreasuryYield Loan-specific control variables Interest Spread Amount Maturiy (Months) NumLenders Revolver Secured PPP Relint Num Cov All variables are defined in Appendix N Panel A: Full Sample STD Mean Median N Mean Median STD N Panel B: Reneg=1 Panel C: Reneg=O Mean Median STD 466 466 0.50 3.60 1.00 4.44 0.50 2.52 234 234 1.00 1.58 1.00 1.02 0.00 1.33 232 232 0.00 5.95 0.00 6.00 0.00 0.23 466 466 466 409 466 466 0.04 0.22 0.22 0.17 0.23 0.31 0.04 0.20 0.17 0.00 0.00 0.00 0.05 0.21 0.19 0.34 0.42 1.61 234 234 234 206 234 234 0.05 0.20 0.24 0.12 0.26 0.05 0.05 0.20 0.19 0.00 0.00 0.00 0.05 0.13 0.20 0.29 0.44 0.90 232 232 232 203 232 232 0.03 0.23 0.20 0.20 0.20 0.56 0.03 0.20 0.17 0.00 0.00 0.00 0.05 0.27 0.18 0.38 0.40 2.23 466 466 466 466 466 466 466 466 0.08 5.44 11.89 0.31 1.42 0.02 0.02 0.04 5.32 8.73 0.30 1.12 0.02 0.01 1.56 1.16 0.09 1.86 43.21 0.18 0.95 0.03 0.02 2.94 234 234 234 234 234 234 234 234 0.09 5.62 9.99 0.31 1.41 0.02 0.02 1.66 0.05 5.48 8.37 0.30 1.10 0.02 0.01 1.14 0.10 1.77 42.11 0.19 1.11 0.03 0.02 3.44 232 232 232 232 232 232 232 232 0.07 5.25 13.67 0.31 1.42 0.01 0.02 1.46 0.04 5.08 8.93 0.30 1.15 0.02 0.01 1.18 0.08 1.93 45.30 0.18 0.75 0.04 0.02 2.34 466 466 -0.16 -0.47 -0.17 -0.38 1.12 1.14 234 234 -0.21 -0.52 -0.26 -0.49 1.15 1.17 232 232 -0.10 -0.43 -0.17 -0.29 1.08 1.11 466 466 466 466 466 466 466 466 466 265.71 0.34 41.43 6.60 0.88 0.77 0.45 0.48 3.77 255.00 0.24 36.00 2.00 1.00 1.00 0.00 0.00 4.00 122.46 0.42 20.06 13.35 0.33 0.42 0.50 0.50 2.90 234 234 234 234 234 234 234 234 234 274.19 0.32 39.86 7.63 0.93 0.79 0.56 0.53 4.24 275.00 0.25 36.00 3.00 1.00 1.00 1.00 1.00 4.00 108.74 0.30 16.02 15.60 0.25 0.41 0.50 0.50 2.85 232 232 232 232 232 232 232 232 232 257.13 0.35 43.03 5.55 0.83 0.75 0.35 0.42 3.28 250.00 0.23 36.00 2.00 1.00 1.00 0.00 0.00 3.00 134.64 0.51 23.39 10.52 0.38 0.43 0.48 0.49 2.87 I. 54 Table 2 (Continued) Panel D: Deal Distribution by Initiation Years Year Freg. Percent 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Total 17 40 18 77 64 86 57 83 10 14 466 3.65 8.58 3.86 16.52 13.73 18.45 12.23 17.81 2.15 3.00 100.00 Freg. 16 17 59 36 14 96 18 Percent 3.43 3.65 12.66 7.73 3.00 20.60 3.86 Utilities 13 2.79 Wholesale, Retail, and Some Services Healthcare, Medical Equipment, and Drugs 72 48 15.45 10.30 Other 77 466 16.52 100.00 Panel E: Deal Distribution by Industry composition Industry Consumer NonDurables Consumer Durables Manufacturing Oil, Gas, and Coal Extraction and Products Chemicals and Allied Products Business Equipment Telephone and Television Transmission Total 55 Table 3 CorrelationMatrix This table presents Pearson (above the significant at the 5% level or higher). All (1) Variable 1.00 Reneg (1) -0.92 TimetoReneg (2) 0.12 G-Score (3) -0.07 C-Score (4) -0.12 Restatement (5) 0.07 AuditorSpecialist (6) 0.07 Transitory (7) 0.11 JBSMProb (8) 0.10 LNASSET (9) -0.04 A Treasury Yield (10) 0.07 InterestSpread (11) 0.05 (12) Log(Amount) -0.04 Log(Maturity) (13) 0.21 (14) PPP 0.12 Relint (15) 0.17 Num Cov (16) diagonal) and Spearman (below the diagonal) variables are defined in Appendix I. (7) (6) (5) (4) (3) (2) 0.12 0.10 -0.01 -0.12 0.07 -0.91 -0.07 -0.12 0.11 0.01 1.00 -0.09 0.38 1.00 -0.46 0.08 -0.03 -0.11 1.00 -0.07 0.02 -0.16 0.07 -0.47 0.14 0.06 1.00 0.10 -0.04 0.11 0.03 1.00 0.06 0.04 -0.07 -0.06 1.00 0.10 -0.01 -0.06 0.19 -0.02 0.01 0.04 0.02 -0.08 -0.06 0.10 0.21 0.13 -0.10 0.49 -0.35 0.19 -0.24 0.06 -0.03 -0.03 -0.15 0.03 0.04 0.03 0.11 0.05 -0.08 0.05 -0.04 -0.36 0.18 -0.18 -0.17 -0.21 -0.04 0.01 0.02 0.08 -0.02 -0.02 0.04 -0.04 -0.05 -0.17 0.08 -0.04 0.05 -0.01 0.07 -0.12 0.19 -0.10 0.04 0.05 -0.09 -0.05 -0.15 0.04 56 correlations of the key variables (correlation coefficients in bold are (8) 0.09 -0.08 -0.13 0.23 0.05 0.03 -0.05 1.00 -0.10 0.01 0.14 -0.03 -0.05 -0.06 -0.05 0.01 (9) 0.11 -0.09 0.37 -0.52 0.18 0.09 0.41 -0.18 1.00 -0.08 -0.11 -0.49 0.18 0.18 0.48 0.02 (10) -0.04 0.04 -0.31 -0.03 -0.04 -0.04 -0.28 0.02 -0.09 1.00 0.03 0.11 0.03 -0.01 0.00 -0.01 (11) 0.11 -0.13 -0.01 0.12 0.06 0.04 0.06 0.21 -0.18 0.02 1.00 -0.05 -0.09 -0.16 -0.07 0.11 (12) 0.04 -0.02 -0.30 0.22 -0.20 -0.14 -0.37 0.00 -0.41 0.16 -0.06 1.00 0.08 0.19 -0.07 0.22 (13) -0.03 0.09 -0.01 -0.10 0.04 -0.02 -0.01 -0.09 0.23 0.03 -0.09 0.13 1.00 0.14 0.02 0.06 (14) 0.21 -0.15 0.08 -0.06 -0.05 -0.04 0.10 -0.08 0.23 -0.02 -0.16 0.19 0.15 1.00 0.18 0.53 (15) 0.12 -0.09 0.17 -0.21 0.07 -0.01 0.23 -0.07 0.50 -0.01 -0.07 -0.06 0.05 0.18 1.00 0.16 (16) 0.17 -0.15 0.03 0.02 -0.09 -0.06 0.06 0.00 0.04 -0.01 0.12 0.21 0.04 0.53 0.17 1.00 Table 4 The Likelihood and Timeliness of Renegotiationsafter Positive Shocks Reneg Variables Inhi() Coef. t-stat Coef. Intercept G-Score -1.27 4.00** (-1.38) (2.11) 4.23** (2.12) ABSMProb LNASSET DTE LEV MTB ROA STDROA 2.09*** 0.13* 0.00 -0.42 -0.02 -0.53 5.46 (2.68) (1.75) (-0.64) (-0.91) (-0.1) (-0.19) (1.09) 2.26*** 0.12 0.00 -0.13 -0.09 0.23 6.33 (3.36) (1.48) (-1.01) (-0.25) (-0.37) (0.07) (1.18) ZSCORE 0.04 (1.38) 0.06* (1.86) zJCreditSpread ATreasury Yield 0.02 0.05 0.04 0.01 InterestSpread 0.00** (0.2) (0.39) (2.2) 0.00** (0.28) (0.06) (2.06) Log(A mount) Log(Maturity) NumLenders Revolver Secured PPP Relint NuinCov 0.18* -0.13 0.00 0.5 1** 0.00 0.36** 0.05 0.03 (1.93) (-0.80) (-0.45) (2.12) (0.00) (2.2) (0.32) (0.78) 0.20* -0.28 0.00 0.41 -0.05 0.40** 0.05 0.02 (1.75) (-1.46) (-0.66) (1.28) (-0.23) (2.07) (0.28) (0.61) Year FE Yes Yes Industry FE Yes Yes N 466 466 16% 3% Pseudo R 2 t-stat The first column of this table presents the estimated coefficients and z-statistics (in parentheses) from a probit regression of whether or not a renegotiation occurs after a positive shock, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. The second column of this table presents the Cox proportional hazard model of the time to renegotiate (Time toReneg), which measures the number of quarters between the credit quality shock and renegotiation. The dependent variable in this regression is h,(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. All variables are defined in Appendix 1. Continuous variables are winsorized at the 1" and 9 9 th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, * and ***, respectively. All hypothesis tests are conducted with standard errors robust to within-firm dependence and heteroskedasticity. 57 Table 5 Cross-SectionalTests Panel A: Cross-Sectional Tests on the Likelihood of Renegotiations Variables Intercept G-Score Restatement G-Score *Restatement Auditor Specialist G-Score*Auditor Specialist Transitory Coef. -1.12 6.22*** -0.63*** -6.74** t-stat (-1.15) (3.07) (-2.77) (-1.97) Dependent Variable = Reneg Coef. t-stat (-1.49) -1.38 3.59** (2.31) 0.36 1.32* 2.00** 0.13* 0.00 -0.45 -0.24 -0.96 6.21 0.04 -0.01 -0.01 0.00*** 0.15 -0.17 -0.01 0.51* 0.03 0.42** 0.04 0.02 Yes Yes 409 19% (2.33) (1.72) (-0.83) (-0.95) (-1.02) (-0.33) (1.19) (1.26) (-0.06) (-0.11) (2.73) (1.58) (-0.94) (-0.74) (1.91) (0.15) (2.34) (0.28) (0.52) 58 2.14*** 0.13* 0.00 -0.45 0.01 0.03 6.09 0.03 0.03 0.04 0.00** 0.18* -0.14 0.00 0.52** 0.01 0.36** 0.07 0.02 Yes Yes 466 17% t-stat (-1.17) (3.02) 0.07 -3.08*** 2.10*** 0.13* 0.00 -0.36 -0.02 -0.49 5.16 0.04 0.05 0.03 0.00** 0.19** -0.17 0.00 0.44* 0.01 0.34** 0.04 0.02 Yes Yes 466 17% (1.26) (-2.72) (2.63) (1.77) (-0.9) (-0.78) (-0.12) (-0.18) (1.03) (1.47) (0.37) (0.22) (2.34) (2.03) (-1.05) (-0.43) (1.83) (0.03) (2.09) (0.27) (0.75) (1.38) (1.81) G-Score *Transitory ABSM Prob LNASSET DTE LEV MTB ROA STDROA ZSCORE ACreditSpread J Treasury Yield InterestSpread Log(Amount) Log(Maturity) NumLenders Revolver Secured PPP Relint Nuni Cov Year FE Industry FE N Pseudo R 2 Coef. -1.08 5.85*** (2.75) (1.81) (-0.64) (-0.97) (0.03) (0.01) (1.2) (1.24) (0.21) (0.31) (2.24) (1.93) (-0.87) (-0.43) (2.14) (0.07) (2.24) (0.46) (0.77) Panel B: Cross-Sectional Tests on the Timeliness of Renegotiations Variables G-Score Restatement G-Score*Restatement Auditor Specialist G-Score*Auditor Specialist Transitory G-Score *Transitory Coef. 5.42*** -0.73*** -6.00* - Table 5 (Continued) t-stat (2.68) (-2.68) (-1.76) Dependent Variable = Inh;(t) Coef. t-stat 3.65** (2.25) 0.25 1.26* A BSMProb LNASSET DTE LEV MTB ROA STDROA ZSCORE z CreditSpread A Treasury_Yield Interest Spread Log(A mount) Log(Maturity) NumLenders Revolver Secured PPP Relint NumCov 1.72** 0.09 0.00 -0.04 -0.36 -1.32 9.06* 0.05* -0.03 0.00 0.00** 0.12 -0.21 0.00 0.51 0.07 0.40* 0.09 0.01 Year FE Industry FE Yes Yes N Pseudo R2 (2.41) (1.10) (-1.18) (-0.09) (-1.52) (-0.4) (1.68) (1.76) (-0.21) (0.02) (2.3) (1.25) (-1.06) (-0.6) (1.43) (0.31) (1.94) (0.55) (0.15) 2.29*** 0.12 0.00 -0.09 -0.08 0.69 6.74 0.06* 0.03 0.00 0.00** 0.19* -0.28 0.00 0.43 -0.05 0.41** 0.05 0.02 Coef. 6.04*** t-stat (2.97) 0.07 -2.90*** (1.36) (-2.87) 1.89*** 0.13 0.00 0.00 -0.06 0.37 5.69 0.06* 0.00 0.02 0.00** 0.21* -0.33* -0.01 0.38 -0.04 0.38** 0.02 0.02 (2.72) (1.51) (-1.08) (0) (-0.26) (0.11) (1.06) (1.74) (-0.01) (0.17) (2.18) (1.84) (-1.65) (-0.71) (1.16) (-0.19) (1.96) (0.09) (0.61) (1.49) (1.70) (3.46) (1.45) (-1.02) (-0.16) (-0.32) (0.21) (1.23) (1.79) (0.25) (0.03) (2.16) (1.71) (-1.45) (-0.61) (1.34) (-0.22) (2.12) (0.34) (0.54) Yes Yes Yes Yes 409 466 466 3% 3% 3% 59 Table 5 (Continued) Panel A of Table 5 presents the probit model of the likelihood of renegotiations on three interaction terms. The dependent variable in all regressions is an indicator variable (Reneg) that is equal to one if the original contract is amended after the positive shock and before the maturity date or the end of the sample period, (i.e., six quarters after the shock), whichever is shorter and zero otherwise. Panel B of Table 5 presents the Cox proportional hazard model of the time to renegotiate (Time to Reneg) on three interaction terms. The dependent variable in all regressions is hi(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. The three regressions in each panel correspond to the tests of Hypotheses H2(a), H2(b), and H2(c) respectively. As the sample period for testing restatements begins in 1998, the sample size is reduced to 409. All variables are defined in Appendix 1. Continuous variables are winsorized at the l and 9 9 th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. All hypothesis tests are conducted with standard errors robust to within-firm dependence and heteroskedasticity. 60 Table 6 Economic Impact Panel A: The Occurrence of Changes to Contract Terms in Amendment Report Change Items Change in interest rates (includingperformancepricing) Change in loan amount Change in maturity Change in collateral/security Change in borrowing base Change in investment restrictions Change in asset sale Change in financialcovenants Change in other covenants Charge amendmentfees Total N umber of Amendment Reports Freq. Percent 28 82 24 15 17% 49% 15% 9% 1 5% 25 43 26% 21 68 13% 41% 10 6% 85 52% 165 100% Panel B: The Occurrence of Changes to Three Major Contract Terms in Renegotiation Sample Change Items AInterest < 0 AInterest = 0 AInterest > 0 Total AAmount > 0 AAmount = 0 AAmount < 0 Total AMaturity > 0 AMaturity = 0 AMaturity < 0 Total Freq. Percent 69 161 30% 69% 4 2% 234 89 108 37 234 80 151 3 234 100% 38% 46% 16% 100% 34% 65% 1% 100% Panel C: Descriptive Statistics of Changes to Three Major Contract Terms N Mean P25 Median P75 STD AInterest (Reneg = 1 andzinterest 0) 73 -86.31 -105.88 -72.50 -43.00 72.53 AAmount (Reneg = I andAAmount# 0) 126 52.80% -16.67% 25.00% 57.14% 1.61 Variable 99.50% 135.90% 0.50 91.78% 55.77% 83 AMaturity (Reneg = I and AMaturity # 0) AInterest is the bps change in interest rate spread during renegotiations. AAmount is the percentage change in loan amount during renegotiations. AMaturity is the percentage change in maturity during renegotiations. 61 Table 7 Likelihood and Timeliness of Renegolialions on Conservatism Reneg Variables Intercept Coef. Inhi(t) t-stat Coef. t-stat -0.98 (-1.09) C-Score -1.11** (-2.49) -1.13** (-2.33) A BSMProb 1.95** (2.48) 2.06*** (3.02) LNASSET 0.12 (1.61) 0.13 (1.63) DTE 0.00 (-0.66) 0.00 (-1.01) LEV -0.24 (-0.51) 0.00 (0.00) MTB -0.19 (-0.9) -0.28 (-1.12) ROA 0.00 (0.00) 0.38 (0.12) STDROA 7.36 (1.51) 8.77* (1.66) ZSCORE 0.04 (1.46) 0.06* (1.94) ACredit_Spread 0.04 (0.36) 0.02 (0.13) Interest_Spread 0.00** (2.25) 0.00** (2.11) Log(Amount) 0.15 (1.64) 0.17 (1.57) Log(Maturity) -0.15 (-0.93) -0.29 (-1.56) NumLenders -0.01 (-0.66) -0.01 (-0.82) Revolver 0.54** (2.29) 0.54* (1.72) Secured 0.00 (-0.02) -0.02 (-0.09) PPP 0.34** (2.07) 0.35* (1.80) Relint 0.10 (0.69) 0.08 (0.47) NumCov 0.03 (0.89) 0.02 (0.74) Year FE Yes Yes Industry FE Yes Yes N 466 466 2 Pseudo R 20% 12% The first column of this table presents the estimated coefficients and z-statistics (in parentheses) from a probit regression of whether or not a renegotiation occurs after a positive shock, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended after the positive shock and before either the maturity date or the end of the sample period, i.e., six quarters after the shock, whichever is shorter, and zero otherwise. The second column of this table presents the Cox proportional hazard model of the time to renegotiate (Time to Reneg), which measures the number of quarters between the credit quality shock and renegotiation. The dependent variable in this regression is hi(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. All variables are defined in Appendix 1. Continuous variables are winsorized at the Is' and 9 9 th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. 62 Table 8 Nature of Good News Received Variables Intercept G-Score Panel A: Firms with AMTB <=0 Reneg Inhi(t) Coef. t-stat Coef. t-stat -1.63 (-1.56) 4.76** (2.15) 6.85** (2.57) BSMProb LNASSET DTE LEV MTB ROA STDROA ZSCORE J Treasury_ Yield ziCreditSpread InterestSpread Log(Amount) Log(Maturity) NumLenders Revolver Secured PPP Relint NumCov 3.16*** 0.19** 0.00 -0.57 -0.59 -2.08 -0.93 0.03 -0.10 -0.24 0.00 0.26** 0.51 0.00 0.34 0.13 0.73*** -0.06 -0.03 Year FE Industry FE Yes Yes (3.23) (2.00) (-0.76) (-0.96) (-1.28) (-0.58) (-0.16) (0.80) (-0.62) (-1.63) (1.06) (2.11) (1.39) (0.06) (1.13) (0.49) (3.38) (-0.30) (-0.62) 3.63*** 0.19 0.00 -0.62 -0.49 -1.85 -1.35 0.06 -0.08 -0.26 0.00 0.33* 0.67 0.00 0.15 0.04 0.84*** 0.01 0.00 (3.64) (1.50) (-0.77) (-0.74) (-1.45) (-0.39) (-0.18) (1.10) (-0.36) (-1.42) (0.89) (1.93) (1.16) (-0.14) (0.35) (0.13) (2.84) (0.05) (-0.07) Panel B: R&D Non-Intensive Firms Reneg Inhi(t) Coef. t-stat Coef. t-stat -1.27 (-1.20) 4.38** (2.14) 5.15** (2.41) 1.94** 0.12 0.00 -0.44 -0.07 -2.55 1.97 0.06** 0.07 0.08 0.00** 0.15 -0.19 0.00 0.56** -0.02 0.26 0.13 0.03 (2.11) (1.37) (-0.04) (-0.86) (-0.28) (-0.73) (0.35) (2.17) (0.5) (0.59) (2.01) (1.43) (-0.98) (0.22) (1.98) (-0.08) (1.43) (0.75) (0.81) 1.87** 0.11 0.00 -0.09 -0.16 -2.64 1.95 0.09*** 0.11 0.10 0.00** 0.16 -0.33 0.00 0.50 -0.10 0.21 0.09 0.05 (2.22) (1.24) (-0.4) (-0.15) (-0.54) (-0.57) (0.33) (3.29) (0.75) (0.77) (2.38) (1.28) (-1.51) (-0.03) (1.36) (-0.41) (0.94) (0.48) (1.15) Yes Yes Yes Yes Yes Yes e. N 339 339 355 355 Pseudo R2 19% 6% 14% 3% Panel A of Table 8 presents the probit model of the likelihood of renegotiations (Reneg) and the hazard model of the time to renegotiate (Time toReneg) for firms with AMTB <= 0. Panel B of Table 8 presents the probit model of the likelihood of renegotiations (Reneg) and the hazard model of the time to renegotiate (Time to Reneg) for R&D Non-Intensive firms. All variables are defined in Appendix 1. Continuous variables are winsorized at the 1s' and 9 9 h percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. 63 Table 9 Alternative Measure of Timely Reporting of Good News Inhi(t) Reneg Variables Intercept DCV Upgrade Coef. -1.59** t-stat (-2.03) Coef. t-stat 0.79** (2.00) 0.67* (1.75) JBSMProb LNASSET DTE LEV MTB ROA STDROA ZSCORE 1.91** 0. 18*** 0.00 -0.57 (2.41) (2.66) (-0.20) (-1.30) (-0.04) (0.01) (1.19) (1.07) (-0.59) (-0.71) (2.77) (1.61) (-1.01) (-0.44) (1.96) (-0.19) (2.28) (0.51) (0.88) 2.15*** 0.19*** 0.00 -0.38 0.01 1.06 6.84 0.04 -0.04 -0.03 0.00** 0.22* -0.33* -0.01 0.49 -0.08 0.39** 0.01 0.03 (3.06) (2.61) (-0.74) (-0.75) (0.02) (0.32) (1.39) (1.45) (-0.32) (-0.29) (2.23) (1.90) (-1.69) (-0.94) (1.51) (-0.35) (2.05) (0.03) (0.85) ACredit_Spread J Treasury_Yield InterestSpread Log(A mount) Log(Maturity) NumLenders Revolver Secured PPP Relint Num_Cov -0.01 0.02 5.38 0.03 -0.04 -0.05 0.00*** 0.15 -0.16 0.00 0.47** -0.04 0.36** 0.07 0.03 Year FE Industry FE Yes Yes Yes Yes N 466 466 2 Pseudo R 11% 3% The first column of this table presents the estimated coefficients and z-statistics (in parentheses) from a probit regression of whether or not a renegotiation occurs after a positive shock, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e.. six quarters after the shock), whichever is shorter, and zero otherwise. The second column of this table presents the Cox proportional hazard model of the time to renegotiate (Time toReneg), which measures the number of quarters between the credit quality shock and renegotiation. The dependent variable in this regression is h,(1), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. The independent variable in both regressions is DCVUpgrade, measuring how well accounting numbers predict future credit rating upgrades. All variables are defined in Appendix 1. Continuous variables are winsorized at the I" and 9 9 th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. 64 Table 10 Interact Timely Reporting of Good News with the Size of Positive Shock Inhi(t) Reneg Variables Coef. -1](- t-stat 42 Coef. t-stat Intercept G-Score 2.85* (1.77) 3.68** (2.11) ABSMProb 2.11** (2.28) 2.18** (2.35) G-Score*ABSM Prob 12.38* (1.71) 4.55* (1.73) LNASSET DTE LEV MTB ROA STDROA ZSCORE 0.13* 0.00 -0.41 -0.02 -0.47 5.64 0.04 (1.79) (-0.55) (-0.90) (-0.08) 0.12 0.00 -0.12 -0.09 0.24 6.43 0.06* ACreditSpread 0.05 0.02 (1.49) (-0.98) (-0.23) (-0.35) (0.07) (1.20) (1.81) (0.25) (0.05) (2.05) ATreasury Yield Interest Spread Log(Aniount) Log(Maturity) NumLenders Revolver Secured PPP Relint NumCov 0.00** 0.18* -0.12 0.00 0.5* -0.01 0.35** 0.05 0.02 (-0.17) (1.13) (1.34) (0.35) (0.19) (2.19) (1.89) (-0.75) (-0.45) (2.11) (-0.03) (2.18) (0.33) (0.75) 0.03 0.01 0.00** 0.19* -0.27 0.00 0.41 -0.06 0.40** 0.04 0.02 Year FE Industry FE Yes Yes Yes Yes N 466 466 2 (1.73) (-1.41) (-0.67) (1.29) (-0.25) (2.06) (0.28) (0.62) 14% 3% The first column of Table 10 presents the probit regression that has G-Score to be interacted with JBSMProb, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. The second column of Table 10 presents the Cox proportional hazard model of the time to renegotiate (Time to Reneg), which has GScore to be interacted with JBSMProb. The dependent variable in this regression is h,(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. All variables are defined in Appendix 1. Continuous variables are winsorized at the V4 and 9 9 th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. Pseudo R 65 Table 11 Firm Loans Without Earnings-BasedPertbrmancePricingProvision Reneg Variables Inhi(t) Coef. t-stat Intercept -1.32 (-1.41) G-Score 4.20** Coef. t-stat (2.12) 4.43** (2.17) 1.85** 0.12 0.00 (2.47) (1.45) -1.64 (2.00) (1.75) (-0.11) (-0.76) (-0.24) (-0.60) 4.80 0.04 (0.94) (1.56) 5.34 0.07** 0.07 0.03 0.00** (0.53) (0.27) (2.36) 0.06 0.02 0.00** (1.82) (-0.92) 0.19 -0.29 NumLenders 0.17* -0.15 0.00 Revolver 0.54** (-0.61) (2.24) -0.01 0.46 Secured PPP Relint Num Cov 0.03 0.40** 0.07 0.02 (0.17) (2.33) -0.03 0.41** 0.06 0.02 JBSMProb 1.66** LNASSET 0.13* DTE 0.00 LEV -0.35 -0.05 MTB ROA STDROA ZSCORE ACredit_Spread ATreasuryYield Interest_Spread Log(Amount) Log(Maturity) (0.44) (0.60) 0.01 -0.13 -1.18 Year FE Yes Yes Industry FE Yes Yes N 466 466 2 (-0.57) (0.02) (-0.5) (-0.36) (1.01) (2.25) (0.42) (0.17) (2.19) (1.61) (-1.51) (-0.78) (1.42) (-0.11) (2.02) (0.35) (0.51) Pseudo R 13% 3% Table 11 shows the estimated coefficients and z-statistics (in parentheses) from both probit regression and hazard regression on the sample of firm loans without earnings-based performance pricing provision. The first column of Table I1 presents the probit regression, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. The second column of Table I1 presents the Cox proportional hazard model of the time to renegotiate (TimetoReneg). The dependent variable in this regression is h,(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. All variables are defined in Appendix 1. Continuous variables are winsorized at the I" and 9 9th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. 66 Table 12 Refinancingfrom Outside Lenders Inhi(t) Reneg Coef. t-stat Intercept Variables -1.28 (-1.40) G-Score 3.94** ABSM Prob Coef. t-stat (2.10) 4.30** (2.15) 1.71** (2.19) 2.04*** (3.01) LNASSET 0.12* (1.68) 0.12 (1.43) DTE 0.00 -0.40 0.00 -0.12 (-1.02) LEV (-0.65) (-0.89) MTB -0.03 (-0.14) -0.09 (-0.37) ROA STDROA -0.21 4.83 (-0.08) (0.98) 0.53 5.99 (0.16) (1.13) ZSCORE 0.04 (1.38) 0.06* (1.87) ACredit Spread 0.07 -0.02 (0.54) (-0.20) 0.05 -0.01 (0.4) (-0.12) 0.00** (2.30) 0.00** (2.13) (1.90) (-0.80) 0.19* -0.30 (1.74) Log(Maturity) 0.17* -0.13 NumLenders 0.00 (-0.22) 0.00 (-0.49) (1.70) (0.03) (1.04) Secured 0.41* 0.01 0.33 -0.04 (-0.19) PPP 0.39** (2.40) 0.42** (2.19) Relint 0.05 (0.34) 0.04 (0.28) NumCov 0.03 (0.88) 0.02 (0.65) Year FE Yes Yes Industry FE Yes Yes N 466 466 J Treasury Yield Interest Spread Log(A mount) Revolver 2 (-0.24) (-1.58) 3% 14% The first column of Table 12 presents the probit regression, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended or a new loan issued by outside lenders replaces the original loan between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. The second column of Table 12 presents the Cox proportional hazard model of the time to renegotiate (Timeto Reneg). The dependent variable in this regression is h,(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. All variables are defined in Appendix 1. Continuous variables are winsorized at the s and 9 9 th percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, *, Pseudo R and ***, respectively. 67 Table 13 Controlfor Changes in Firm Characteristics Inhi(t) Reneg Variables t-stat (-1.56) (1.95) Coef. t-stat Intercept G-Score Coef. -1.49 3.66* 4.10** (2.10) JBSMProb LNASSET DTE LE V MTB ROA STDROA ZSCORE ALNASSET JDTE JLEV AMTB zROA ZSTDROA JZSCORE ACreditSpread ATreasury Yield Interest_Spread Log(Amnount) Log(Maturity) NumnLenders Revolver Secured PPP Relint NumCov 2.23*** 0.17** 0.00 -0.74 0.13 2.58 4.30 -0.04 0.91* 0.00 -1.38 0.60*** 19.83*** 5.97 -0.27*** -0.01 -0.02 0.00* 0.20** -0.13 0.00 0.60** -0.02 0.44*** -0.01 0.02 (2.71) (2.25) (0.38) (-1.55) (0.60) (0.81) (0.80) (-1.05) (1.90) (0.42) (-0.95) (3.29) (3.53) (0.65) (-3.72) (-0.09) (-0.16) (1.95) (2.08) (-0.78) (-0.60) (2.36) (-0.09) (2.70) (-0.04) (0.47) 2.22*** 0.14* 0.00 -0.51 0.06 3.57 6.62 -0.01 0.44 0.00 -0.76 0.45*** 13.63*** 2.52 -0.20*** 0.01 -0.02 0.00* 0.17 -0.26 0.00 0.48 -0.09 0.49** -0.07 0.02 (3.16) (1.78) (0.10) (-0.92) (0.26) (0.91) (1.24) (-0.35) (1.33) (0.95) (-0.62) (3.37) (2.67) (0.21) (-3.97) (0.04) (-0.15) (1.78) (1.58) (-1.32) (-0.65) (1.45) (-0.41) (2.54) (-0.39) (0.67) Year FE Industry FE Yes Yes Yes Yes N 466 466 Pseudo R 2 18% 4% The first column of this table presents the estimated coefficients and z-statistics (in parentheses) from a probit regression of whether or not a renegotiation occurs after a positive shock, where the dependent variable is an indicator variable (Reneg) that is equal to one if the original contract is amended between the end of the positive shock and the maturity date or between the end of the positive shock and the end of the sample period (i.e., six quarters after the shock), whichever is shorter, and zero otherwise. The second column of this table presents the Cox proportional hazard model of the time to renegotiate (Time_to_Reneg), which measures the number of quarters between the credit quality shock and renegotiation. The dependent variable in this regression is h,(t), which measures the instantaneous risk of a renegotiation at time t for borrower i conditional on i surviving to time t. All variables are defined in Appendix 1. Continuous variables are winsorized at the I" and 9 9 percentiles. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively. 68