Unethical Behavior and Debt Contracting: Evidence from Backdated Option Grants

Transcription

1 Unethical Behavior and Debt Contracting: Evidence from Backdated Option Grants Veljko Fotak SUNY Buffalo Feng Jiang SUNY Buffalo Hae Kwon Lee SUNY Buffalo Current draft: July 13, 2016 We study the impact of unethical behavior on debt contracting. We find that, after the revelation of option backdating, borrowers pay higher spreads on loans, by about 19 bp. The increase is larger for loans originating from lenders with no prior relationship with the borrower, for geographically distant lenders, and for opaque borrowers. Our results are consistent with the notion that unethical behavior leads to an increase in perceived information risk. On the other side, we do not find any impact on the cost of public debt and find that, after the revelation, backdating borrowers rely more on public than on private debt. JEL Classification: G32 Keywords: Corporate culture; Cost of debt; Option backdating; Soft information; Unethical behavior Please address correspondence to: Feng Jiang 344 Jacobs Management Center, Buffalo, NY Tel: (716) * We thank Michael Dambra, Sahn-Wook Huh, William Kross, William Megginson, Inho Suk, Cristian Tiu, Brian Wolfe, Fei Xie, and seminar participants at the University at Buffalo for their valuable feedback and comments.

2 1. Introduction It has been long recognized that culture plays an important role in explaining individual behavior and economic growth (e.g., Weber, 1930; North, 1990; Greif, 1994; Lal, 1999; Lehrer, 2004; Stulz and Williamson, 2003). The impact of firm culture on corporate decisions and asset prices, however, is less understood (Hermalin, 2001). 1 As a consequence, there is a growing interest in understanding the effect of corporate culture on corporate behavior and performance (e.g., Guiso, Sapienza, and Zingales, 2015; Pan, Siegel, and Wang, 2014; Benmelech and Frydman, 2015; Biggerstaff, Cicero, and Puckett, 2015). In this paper, we extend this line of research to investigate the impact of corporate culture on debt contracting. In particular, we examine the impact of unethical corporate culture on the perceived level of information risk of the firm and, consequently, on the price and non-price terms of bank loans, on the price of public debt and, ultimately, on the choice of private versus public debt financing. Stein (2003) argues that failures in the collection, processing, and sharing of information are the most pervasive and important violations of the Modigliani and Miller (1958) perfect capital market assumption. Hence, in the absence of perfect information, lenders are concerned about the quality of disclosure of borrowers and about the perceived level of information risk of the firm. Extant literature on debt contracting highlights the importance of information transparency as a determinant of a firm s cost of capital (Diamond and Verrecchia, 1991), while a stream of studies on information transparency has established that unethical behavior affects the credibility of corporate disclosure (Karpoff, Lee, and Martin, 2008). Accordingly, we expect that unethical behavior will increase the perceived information risk of the firm and, in turn, affect debt contracting. One empirical challenge in this area is that ethics and corporate culture are hard to observe and quantify. We build on prior literature finding that executive option backdating may be a manifestation of lax ethical norms in the firm (e.g., Armstrong and Larcker, 2009) to identify firms with unethical 1 Hirshleifer (2014) argues that Most importantly, there is a need to move from behavioral finance to social finance (or social economics). Social finance includes the study of how social norms, moral attitudes, religions and ideologies affect financial behaviors [ ], and how norms that affect financial decisions form and spread (p. 44). 2

3 culture. Our approach is validated by Biggerstaff, Cicero, and Puckett (2015), who use option backdating as a proxy for unethical culture and report evidence that firms with CEOs who personally benefit from options backdating are more likely to engage in other corporate misbehaviors, suggestive of an unethical corporate culture. We accordingly exploit the revelation of executive option backdating in 2006 to study the impact of revealed unethical culture on debt contracting. Lie (2005) defines the backdating of stock options as the practice of retroactively choosing a favorable date (i.e., when the stock price was low) as the date on which stock options were supposedly granted. 2 Extant research shows that backdating was usually initiated by CEOs or other high-level executives and widespread around the turn of the century, with as many as 30% of public firms engaging in the practice (Heron and Lie, 2007; Heron and Lie, 2009; Bizjak, Lemmon, and Whitby, 2009; Collins, Gong, and Li, 2009; Bebchuk, Grinstein, and Peyer, 2010). The main advantage of using the revelation of prior backdating of option award grants as a proxy for unethical behavior is that, while indicative of unethical behavior, it does not reveal new information about the underlying performance of the firm and has limited cash flow implications (Bernile and Jarrell, 2009), which hence allows us to separate a wealth effect from an information effect (Graham, Li, and Qiu, 2008). We focus on syndicated loan contacting for three main reasons. First, bank loans are an inherently important market to investigate as they constitute, worldwide, the largest source of external funding for corporations (Lin et al., 2010; Chui et al., 2010). In 2013, global syndicated lending for the full year reached USD 4.7 trillion (USD 2.3 trillion in the USA), while global bond issues, the second 2 Because the value of an option is higher if the exercise prize is lower, executives should prefer being granted options when the exercise prize is at its lowest. Backdating would not be illegal if it were clearly communicated to shareholders, adequately accounted for in both earnings and taxes, and no document was forged. However, this is rarely true in practice, making most instances of backdating illegal. Academic interest in option grant dates started with the finding that returns are abnormally low leading up to the grants and positive afterwards (Yermack, 1997). Early research documented abnormal stock returns around stock option grants but it was only in Lie (2005) established that official grant dates had been timed retroactively by many firms. 3

4 largest source of financing for corporations, reached approximately USD 3.5 trillion. 3 About 80% of public companies headquartered in the USA have outstanding bank loans compared to only 20% with outstanding public debt. Thus bank loans provide an optimal testing ground offering evidence generalizable to a large cross-section of firms. Second, bank loan contracts allow for the identification of both borrowers and lenders, allowing us to test whether the observed impact is due to an increase in perceived information risk. Extant literature finds that lenders that have previous relationships with the borrower (Bhattacharya and Chiesa, 1995; Boot, 2000; Bharath et al, 2009), or lenders that are geographically closer to the borrower (Agarwal and Hauswald, 2010), have superior access to information about the borrower and thus enjoy lower information asymmetry. Accordingly, we posit that such lenders would be less sensitive to information risk and test whether the impact of unethical behavior on loan terms differs for lenders with prior relationships and who are geographically close to borrowers. Third, according to Melnik and Plaut (1986), loan contracts are effectively a bundle of contract terms including both price and nonprice terms such as maturity, collateral and covenant requirements, seniority levels, and others. In this sense, syndicated loans allow for the investigation of non-price mechanisms employed to mitigate perceived information risk. Our main sample includes 7,509 loans to US firms over the years spanning the interval between 2000 and 2012, of which approximately 34% are to firms identified as likely backdaters. 4 We find strong evidence of an increase in spreads (over LIBOR) for loans to borrowers identified as likely backdaters. Our results are both statistically significant and economically meaningful, as the estimated increase in the cost of loans is of approximately 19 basis points (bp), or about 8% of average loan spreads. We further find that the increase in the cost of loans is related to the number of option grants that are likely to have been backdated. 3 The totals are, respectively, from the Thomson Reuters Debt Capital Markets Review (Full Year 2014) and the Thomson Reuters Global Syndicated Loans Review (Full Year 2014), both available at 4 We follow Lie (2005) and Heron and Lie (2007, 2009) in identifying firms that are likely to have modified option grant dates ex-post. Here, and in the remainder of the paper, we refer to such firms as backdaters for brevity and we conversely refer to all other firms as non-backdaters. 4

5 Further, we find that the impact of backdating on the cost of loans is mitigated by lenders with lower information asymmetry. In particular, we find that the increase in the cost of loans for backdaters is much smaller if they have previous lending relationships with the borrowers. While the increase in the cost of the loan is of approximately 34 bp for loans originating from new lenders, the increase is of only 6 bp for loans originating from relationship lenders. Similarly, we find that the increase in the cost of debt is smaller for borrowers who are geographically close to the lender. After revelation of option backdating, we find that the cost of loans increases by 2 bp for every 100 miles of distance between lender and borrower who is a likely backdater. In additional tests, we find that the revelation of option backdating has a stronger impact on the cost of debt of firms that are ex-ante less transparent: smaller firms, firms with a smaller analyst following, and firms not included in the S&P 500 index. Overall, these findings suggest that the increase in the cost of debt is due to a revised perception of information risk following the revelation of unethical behavior, rather than to the revelation of failures of internal governance and control mechanisms. In contrast, we find no significant impact on the non-price terms of loans, suggesting that lenders do not try to mitigate the perceived increase in information risk by relying on risk-mitigating loan features such as financial covenants, collateral, or shorter maturities. While our main interest lies in studying the impact of managerial unethical behavior on the terms of loans, we extend our investigation to public debt markets. Given that bond investors have both lower incentives to monitor borrowers and generally do not share the superior access to firm level information that banks have, we would expect them to react even more forcefully to a deterioration in perceived disclosure quality. On the other side, extant literature finds that, while banks value soft information (Berger and Udell, 2002; Berger et al, 2005), public debt markets tend to price debt mostly on the basis of hard information. Accordingly, this might suggest that the revelation of unethical behavior would have a stronger impact on bank lenders than on public debt investors. Consistent with the latter explanation, our tests reveal no change in the cost of public debt following the revelation of the backdating practice. 5

6 Given that the bond market appears not to penalize firms after the revelation of option backdating, we expect an increased reliance on public over private debt by firms likely to have backdated option grants. Our findings support this conjecture, as we document that, post-revelation, backdaters are more likely to seek public debt financing post-revelation and issue a greater share of public debt. This migration to public debt markets is particularly marked for large borrowers and borrowers with a large analyst following, suggesting that large, transparent borrowers who have access to bond-market financing rely less on bank financing post-revelation, while smaller, more opaque borrowers maintain their reliance on bank financing and suffer a greater increase in the cost of debt. Extant literature further finds that lenders respond to higher information risk by charging higher interest rates, by employing non-price risk mitigating loan terms, but also by rationing capital (Stiglitz and Weiss, 1981). Accordingly, we document that the revelation of option backdating leads to an increased sensitivity of investments to cash flows, which has been interpreted in prior studies as a metric of financial constraints (Almedia, Campello, and Weisbach, 2004; Fazzari, Hubbard, and Petersen, 1998). Our evidence suggests that unethical behavior not only increases the cost of debt of backdaters, but that it also leads to capital rationing and increased financial constraints. Our main contribution is to the literature on unethical behavior, as we show that the revelation of malfeasance increases perceived information risk, which leads to both a higher cost of debt and increased financial constraints. In this sense, we add to the findings by Graham, Li, and Qiu (2008), who show that earnings restatements have an impact on loan contracting terms. Yet, while Graham, Li, and Qiu (2008) recognize being empirically unable to separate new information about the underlying cash flows of the firm from the impact of increased information risk, our empirical setup more clearly identifies the impact on loan contracting as a consequence of a break in trust. 5 Our second contribution lies in offering evidence that unethical behavior by managers has long lasting consequences on the 5 Bernille and Jarrell (2009) extensive examine the channels through which option backdating impact firms and conclude that the practice has a strong impact on firm value due to an increase in information risk related to lower perceived quality of disclosure, while having little impact on firms fundamentals and cash flows. 6

7 perceived information quality of the firm s disclosure, which indicates a persistent impact on firm culture. We further contribute to the literature on option backdating, which has examined how the practice impacts firms from the vintage point of equity holders, but has so far failed to capture the impact on debt contracts. We also add to the literature on relationship lending, by providing evidence that banks that engage in multiple transactions with borrowers or that are located in geographical proximity can mitigate the negative impact of a shock to information risk. Finally, we provide insights into firms choice of private vs. public debt, by showing that private debt markets react more strongly to the revelation of unethical behavior, which leads to an increase in the cost of private debt of unethical firms and a shift in borrowing from private to public debt markets. This paper is organized as follows. Section 2 develops testable hypotheses. Section 3 describes the data sources and identification standards. Section 4 focuses on the empirical analysis. Section 5 presents our conclusions. 2. Hypotheses development Early studies have documented unusual patterns in stock prices surrounding option grant dates (Yarmack, 1997; Aboody and Kasznik, 2000; Chauvin and Shenoy, 2001). Yet, it was Lie (2005) that first postulated a backdating hypothesis to explain that positive abnormal stock price returns following option grant dates were due to the ex-post manipulation of the dates. This led to strong attention in the media, including a story published in The Wall Street Journal on March 18, 2006, which has been cited as having attracted public interest in the backdating scandal (Bernille and Jarrell, 2009). As in extant studies, we hypothesize that it is only after this initial revelation that investors become aware of the practice and of the identity of specific firms likely to have backdated option grants. The revelation of unethical behavior by managers can increase a lender s uncertainty about the firm s financial information (Graham, Li, and Qiu, 2008). Extant literature finds that investor s perception of disclosure quality and, in turn, uncertainty about the quality of financial information have an impact on a firm s cost of capital and firm value (Diamond and Verrecchia, 1991; Duffie and Lando, 7

8 2001; Easly and O Hara, 2004; Yu, 2005; Kumar et al., 2006; Epstein and Schneider, 2008). Bernille and Jarrell (2009) find that option backdating leads to a loss of investors confidence in a firm s management and a consequent increase in perceived information risk. We accordingly hypothesize that the revelation of option backdating leads to an increase in the cost of debt of firms. Hypothesis H1: The revelation of option backdating increases the cost of debt of firms. Our main hypothesis is that option backdating affects firm s cost of capital because of an increase in information risk. Yet, option backdating might also reveal new information about the quality of governance and internal controls of the firm. Bebchuk, Ginstein, and Peyer (2010) find that option backdating is associated with weak internal governance few independent directors, no outside blockholders, and entrenched CEOs. Accordingly, we recognize that option backdating might have an impact on bank loan contracting terms as investors revise their expectations regarding agency costs stemming from the separation of ownership and control of the firm (Jensen and Meckling, 1976). In order to verify that the main impact is due to an increase in perceived information risk, rather than on the perceived quality of internal governance, we test whether the impact of option backdating on loan contracts is conditional on the level of information asymmetry between borrower and lender. The underlying assumption is that, if option backdating affects loan terms because of an increase in information risk, this effect should be weaker for lenders who have superior access to information about the borrower. Prior studies find that lenders that have repeat interactions with the same borrowers suffer from lower information asymmetry (Berger and Udell, 1995; Bharath et al, 2009). Similarly, extant literature finds that lenders that are physically close to the borrower are capable of mitigating information asymmetry problems via closer monitoring. Hence, we hypothesize that the impact on the cost of debt due to the revelation of option backdating will be lower for lenders who have a prior relationship with the borrower or who are located in geographical proximity to the borrower. Hypothesis H2: The impact of option backdating on the cost of loans is weaker if the lender has 8

9 a prior relationship with the borrower or if the lender is geographically close to the borrower. 6 In a similar spirit, we recognize that some borrowers are more transparent than others. If option backdating affects firms cost of debt mainly by increasing information risk, we would expect this effect to be stronger for less transparent borrowers. Extant literature finds that large borrowers, borrowers with a large analyst following, and borrowers included in the S&P 500 index are generally more transparent. 7 Accordingly, we hypothesize that the impact on the cost of debt due to the revelation of option backdating will be lower for borrowers that are larger, have a greater analyst following, and are included in the S&P 500 index. Hypothesis H3: The impact of option backdating on the cost of loans is weaker if the borrower is large, has a large analyst following, or is included in the S&P 500 index Extant literature further finds that lenders not only respond to an increase in information asymmetry by requiring higher compensation in the form of higher spreads. Rather, lenders employ other non-price mechanisms to mitigate the risk level of a lending contract. Qian and Strahan (2007) and Bae and Goyal (2009) find that borrowers use loan maturity as a risk-mitigating mechanism. Chava and Roberts (2008) discuss in detail the use of financial covenants to mitigate loan risk. Accordingly, we posit that: Hypothesis H4: After the revelation of option backdating, lending contracts will include nonprice risk-mitigating features, including: shorter maturity, more frequent use of collateral, and a higher number of covenants. Esty and Megginson (2003) and Qian and Strahan (2007) find that risk affects the size and concentration of a lending syndicate. Their evidence is consistent with concentrated lending leading to 6 Finding an interaction between borrower sensitivity to information asymmetry and the revelation of backdating in determining the firm s cost of debt would provide strong evidence in support of option backdating increasing cost of debt via information risk. We nonetheless recognize that we cannot, ultimately, rule out the possibility that the impact is at least partially due to revise expectation about future agency costs. 7 Hong, Lim, and Stein (2000) find that information transfer to public is slower for firms with low analyst coverage. Seasholes and Zhu (2010) find that there is a great overlap between alternative measures of information asymmetry and S&P 500 inclusion. 9

10 better monitoring incentives and greater re-contracting flexibility. On the other side, riskier loans could also lead to larger lending syndicates, as lenders attempt to spread risk by retaining a smaller share of the loan. Accordingly, the net impact of information risk on the size of the lending syndicate cannot be easily predicted and the link between the revelation of option backdating and the size of the lending syndicate is worthy of empirical investigation. Hypothesis H5: The revelation of option backdating affects the size and concentration of the lending syndicate. As a further test of the information asymmetry channel, we recognize that banks are generally less sensitive to information asymmetry problems than public debt holders. First, banks have access to better information about borrowers (Fama, 1985; James and Smith, 2000). Second, banks have a stronger incentive to engage in costly monitoring (Berlin and Loeys, 1988; Diamond, 1991; Houston and James, 1996). Finally, banks have the ability to discipline firms more effectively than public debt holders (Gertner and Scharfstein, 1991; Denis and Mihov, 2003; Park, 2000). As a consequence, banks are less sensitive to changes in information quality than public debt holders (Leland and Pyle, 1977; Diamond, 1984; Boyd and Prescott, 1986; Houston and James, 1996; Denis and Mihov, 2003). Accordingly, we expect to see a stronger impact (a larger increase in spreads) for bonds issued by backdaters, in comparison to loans received by backdaters. We do recognize, however, that extant literature finds that banks tends to rely more on soft information, such as borrower character (Berger and Udell, 2002; Petersen, 2004; Berger et al, 2005), than public sector lenders, who generally tend to give greater emphasis to hard information such as operating performance metrics. As the revelation of option backdating confers information about unethical firm culture soft information we could observe a stronger reaction from banks than from bond investors. Ultimately, these two streams of the literature offer conflicting predictions about the relative reaction of private versus public debt markets to the revelation of unethical behavior. Accordingly, we recognize that the matter is open for empirical investigation. Hypothesis H6: The revelation of option backdating impacts the cost of public debt of firms to 10

11 a different degree than the cost of private debt. Not only we expect a different impact on public debt markets, but we note that a difference in market reaction to the revelation of unethical behavior should affect the firm s optimal choice between public and private debt (e.g. Leland and Pyle, 1977; Campbell and Kracaw, 1980; Diamond 1984, 1991; Fama, 1985; Berlin and Loeys, 1988; Rajan, 1992; Park, 2000). Accordingly, we expect the revelation of option backdating to impact the firm s choice between bank financing and public debt markets, with reliance increasing on the type of debt whose cost is less affected by the revelation of unethical behavior. Hypothesis H7: The revelation of option backdating affects the proportion of private loans to public debt issues. Finally, extant literature further finds that lenders respond to higher information risk by charging higher interest rates, by employing non-price risk mitigating loan terms, but also by rationing capital (Stiglitz and Weiss, 1981). We accordingly hypothesize that backdaters could suffer from higher capital constraints. One of the proxies used in the literature for capital constraints is the sensitivity of investments to cash flows (Almedia, Campello, and Weisbach, 2004; Fazzari, Hubbard, and Petersen, 1998). We accordingly expect an increased sensitivity of investments to cash flows for backdaters following the revelation of the practice. Hypothesis H8: The revelation of option backdating increases the sensitivity of firm s investments to cash flows. 3. Data 3.1. Backdating of executive option grants We follow the methodology in Lie (2005) and Heron and Lie (2007, 2009) to construct the sample to identify firms that are likely to have modified option grant dates ex-post. Here, and in the remainder of the paper, we refer to such firms as backdaters for brevity and we conversely refer to all other firms as non-backdaters. 11

12 We first obtain the sample of stock option grants to CEOs from the Thomson Financial Insider Filing database. This database captures insider transactions reported on SEC forms 3, 4, 5, and 144. We restrict the sample to transactions that occurred from January 1996 to December We further require that stock returns be available from 20 trading days before to 20 trading days after the grant date. Finally, following Heron and Lie (2009), we only include grants to the CEO, President, or Chairman of the Board. We include all three categories because in many instances, CEOs identify themselves by an alternate title (such as the President) in their SEC filings (Heron and Lie, 2009). We eliminate any duplicate grants that occur on a given grant date, so for each firm we have only one grant event for a given date. Like other studies, we focus on unscheduled awards because these grants are much more likely to be manipulated (Heron and Lie, 2007 and 2009). A grant is identified as scheduled if a grant is issued on the same date, plus or minus one day, during the preceding year; otherwise, it is classified as unscheduled. Our final CEO option grants sample consists of 29,421 grants across 4,326 companies over the period We follow the methodology in Bizjak, Lemmon, and Whitby (2009) to identify backdated option grants. The key assumption used by Bizjak, Lemmon, and Whitby (2009) is that if option grant dates are chosen randomly instead of manipulated, there will not be any unusual performance pattern in the stock price surrounding the grant date. Alternatively, if firms use hindsight to identify past dates with particularly low stock prices when setting option grant dates, the stock prices will exhibit a reversal around the reported grant date. Consistent with backdating, Lie (2005); Heron and Lie (2007); Narayanan and Seyhun (2008); Bizjak, Lemmon, and Whitby (2009) find that, on average, stock option grants are preceded by a fall in the stock price, with a subsequent increase in the stock price following the reported grant date. 8 Similar to Bizjak, Lemmon, and Whitby (2009), we begin with 1996 because it is the first year Thomson began collecting data on option grants, and we end our sample period in 2002 because Heron and Lie (2009) report that the incidence of backdating drops dramatically after the implementation of new insider reporting guidelines associated with the passage of SOX in August of

13 As discussed in Section 2, we recognize that investors become aware of the practice of option backdating only after the backdating scandal originating from a series of academic and media articles. Given that multiple articles appeared in the press from the early spring of 2006 and the strong media coverage continuing through the year, we identify 2006 as a watershed period and label the pre-2006 period (ending on December 31, 2005) as pre-revelation and the post-2006 period (starting January 1, 2007) as the post-revelation period Syndicated loans The main source of data for syndicated loans used in this study is the Thomson Reuters Loan Pricing Corporation Deal Scan database ( DealScan ). DealScan includes loans, high-yield bonds, and private placement transactions spanning the globe. The version of the database used in this study covers loans initiated between January 1980 and December The database includes data on loan pricing, contract details, terms and conditions, plus information on loan participants (borrower and lender identities and sparse accounting data). The loans are organized by package and by facility. Each package represents a loosely-defined deal and may contain one or multiple facilities on an average, there are approximately 1.5 loans in each package. All loans within the same package share the same borrower, but the identity of the lender, or composition of the lending syndicate, type of loan, loan initiation date and other contract characteristics can all vary between loans from the same package. 9 For each loan, we obtain, as an estimate of cost to the borrower, the all-in-drawn spread. We further record the loan maturity (recording at initiation, in months), the facility amount (in USD), the number of lenders, headquarter addresses of lenders, indicator variables identifying collateralized loans and senior loans, and information on the number of financial and general covenants. We also create indicator variables based on the database fields identifying loan type and loan purpose. We limit our sample to loans identified as 364-Day Facility, Bridge Loan, Term Loan of 9 Chava and Roberts (2008) describe the database extensively. Some recent empirical studies using data from this database include Guner (2006), Qian and Strahan (2007), Sufi (2009), Bae and Goyal (2009) and Haselmann and Wachtel (2010). 13

14 all types, Revolver line of all maturities and Other Loan, thus excluding not only bonds and private placements, but also credit letters and guarantees. We further exclude loans whose status is Cancelled or Rumor. Further, we exclude from our sample all loans for which data on the composition of the lending syndicate is missing and loans with conflicting information (for example, loans marked as single-lender loans for which multiple lenders are listed). Finally, we include in the sample only loans to US firms that issue options to executives as part of their compensation packages related data is discussed in more detail in the following sections. The rationale for this restriction is to make sure we have a truly comparable benchmark of firms: given that our backadaters, by definition, issue options, we want to make sure that factors correlated with the decision to issue options are not inducing sample selection biases in our analysis. Our final sample includes 7,509 loans to 1,847 firms Bonds In additional tests, we compare and contrast the impact of the revelation of unethical behavior between private and public debt markets. The bond sample is the newly issued bonds by U.S. companies provided by Thomson Reuters s SDC Global New Issues database. For each bond issue, SDC provides detailed information, including the cost of the bond in terms of spread, the issue date, yield-to-maturity (YTM), maturity, proceeds, and ratings. Bond spread is the difference between the yield-to-maturity of the corporate bond of interest and that of a U.S. Treasury bond with comparable maturity, measured in bp. Bond issues without information on spread and maturity are excluded. We further exclude convertible bonds and bonds for which we are unable to obtain necessary financial statement information. Our final sample includes 2,820 bonds issued by 641 firms Additional data We obtain accounting information from the COMPUSTAT database. Size (log of total assets), coverage (coverage ratio), leverage (debt-to-asset ratio), profitability (ROA), valuation (Tobin s Q), firm age, cash volatility and bankruptcy risk (Altman Z-score) are estimated and used as a firm level control variables. The Dealscan-Compustat link file provided by Professor Roberts is used to merge the 14

15 syndicated loan data and the option grant data. 10 We exclude bonds and loans issued by financial institutions and utilities. The number of analysts covering the firm is measured by the number of analysts issuing earnings forecasts for the relevant firm during the previous year. The information on analysts are obtained from Thomson Reuters I/B/E/S database. A full list of variables, with definition and sources, is included in Table 1. *** Insert Table 1 about here *** 4. Empirical analysis 4.1. Descriptive statistics As a first step in analyzing the impact of option backdating on loan contracting, we first present summary statistics regarding price and non-price loan terms in Table 2. Our total sample includes 7,509 loans with complete data. Of those, 34% are to borrowers we identify as backdaters. Approximately half (49%) are classified as relationship loans that is, loans from lenders that have previously lent to the same borrower. The average all-in-drawn spread is 229 bp over LIBOR. The average maturity is of months, or approximately 4 years. The average loan size is of USD million. The average syndicate is composed of 8.69 members. 72% of the loans are collateralized. Loans in our sample have an average of 2.15 financial covenants and 4.14 general covenants. Table 2 also contains descriptive statistics of borrowers. We note that borrowers in our sample are fairly large firms, with average assets equal to USD 3.92 billion and are often mature firms, with an average firm age of 24 years. *** Insert Table 2 about here *** 10 Michael Roberts extended the link data used in Chava and Roberts (2008) that includes Dealscan and COMPUSTAT links for the period between 1983 and August The file is available at 15

16 4.2. Price and non-price loan terms for backdating borrowers and matched borrowers To investigate the impact of the revelation of managerial unethical behavior on loan terms, we compare the terms of loans to borrowers likely having backdated options ( backdaters ) issued after the revelation of the practice of option backdating (that is, after 2006) to loans to backdaters issued prior to the revelation of option backdating (that is, prior to 2006). As a metric of the cost of the loan, we rely on the all-in-drawn spread. Our analysis of the non-price loan terms includes metrics of loan maturity, the number of lenders participated in the syndicated loan deals, size of the loan, a binary variable identifying collateralized loans (secured vs not-secured), and the number of financial and general covenants. To ensure that we are comparing loan terms to truly comparable loans, we match loans initiated after the revelation of backdating to loans to backdating borrowers from the same industry (first-digit SIC code) and classified as having the same Loan purpose and Loan type. Our analysis relies on a comparison of loans issued post-revelation to loans issued prior to the revelation of the practice of option backdating. As such, it suffers from the empirical problems of all analysis relying on a single event in time that is, the need to separate temporal trends from the impact of the event that we are investigating. Accordingly, we rely on a difference-in-difference comparison by comparing changes in loans to backdating firms to changes in loans to a sample of non-backdaters. We present means of variables related to price and non-price loan terms by sub-periods and t- tests for differences in means in Table 3, Panel A. For backdaters, we find that spreads increase postrevelation, from a mean of bp to bp. The approximately 63 bp increase is highly statistically significant and clearly economically meaningful. While we observe an increase in spreads in the benchmark sample of non-backdaters, this increase is smaller in magnitude (37 bp). The difference-in-difference indicates that, post revelation of option backdating, the spread on loans to backdaters increases by bp more than the spread on loans to non-backdaters, and the result is statistically significant at 1%. We further observe an increase in loan maturity for both samples, but it is of similar magnitude (4.8 months for backdaters, 4.6 months for non-backdaters) and the differencein-difference analysis is not statistically significant. Average loan size increased by USD million 16

17 for backdaters and by USD million for non-backdaters; the difference-in-difference is not statistically significant. We further find that lending syndicate shrink post-revelation for both backdaters (a loss of 3 lenders, on average) and non-backdaters (a loss of 2 lenders). The difference-in-difference is statistically significant at 5% and suggests that, after revelation of unethical behavior, lending syndicates tend to shrink in size. This is consistent with the prior literature finding that higher information risk borrowers optimally contract with smaller syndicates that, in virtue of their size, have superior monitoring incentives and advantages in renegotiating eventual covenant violations. We observe no change in the frequency of use of collateral in either sample. Finally, in both sample we see fewer financial and general covenants, but no significant difference between backdaters and nonbackdaters. One potential weakness of the matching test above is that it does not match loans to backdaters and non-backdaters by initiation year. Given the importance of controlling for temporal trends in analyzing lending patterns, we construct a second set of matched tests by matching to each loan to a backdaters a loan to a non-backdater from the same industry (first-digit of the primary SIC code) initiated during the same year and sharing the same Loan purpose and Loan type. We do so both for prerevelation and post-revelation loans and then compute difference-in-differences estimators. In this second set of tests, we find no statistically significant difference in the spreads on loans to backdaters and non-backdaters in the pre-revelation period, but a bp difference in the spreads post-revelation. The difference in difference is statistically significant at the 5% level. Our difference-in-difference analysis further reveals that backdaters receive larger loans both pre- and post-revelation of unethical behavior but the difference-in-difference is not statistically significant. Overall, the results in Table 3 indicate a statistically and economically significant increase in the cost of loans to backdaters after the revelation of the option backdating practice, which is consistent with the hypothesized increase in perceived information risk. We also find strong evidence of larger loan sizes for loans to backdaters even after the revelation. While extant theory does not offer any clear predictions regarding loan size, the larger loan size is somehow at odds with the idea that unethical 17

18 borrowers, once identified as such, are likely to suffer from borrowing constraints. The analysis of the non-price terms of the loan indicates, for the most part, that lenders do not rely on non-price loan terms to mitigate the impact of unethical behavior and the consequent increase in information risk. In this sense, the only statistically significant result is an indication that lending syndicates are more concentrated post-revelation, which is consistent with our priors but the result is not robust to the two different matching methodologies we employ. *** Insert Table 3 about here *** 4.3. The impact of backdating on loan spreads in regression analysis In this section, we conduct multivariate analysis to gauge the impact of the revelation of option grant backdating on the cost of bank loans for backdaters. In the regression, the dependent variable is all-in spread, which is defined by DealScan as the total annual cost, including a set of fees and fixed spread, paid over LIBOR for each dollar used under the loan commitment. The variable of interest is the interaction term between a binary variable identifying a backdating borrower and a binary variable identifying the post-revelation period (Backdater Post revelation). We also control for firm characteristic that have been found to be associated with the cost of debt in prior literature: Total assets, Coverage, Leverage, Profitability, Tobin s Q, Firm age, Speculative rating, Analysts, Cash volatility, and Altman Z, as defined in Table 1. We further control for loan characteristics, including loan size, loan maturity and use of collateral as in Graham, Li, and Qiu (2008) and Engelberg, Gao, and Parsons (2012). 11 We further control for macroeconomic factors by including the default spread (the yield spread between BAA and AAA corporate bond indices) and term spread (the yield spread between a ten-year Treasury bond and three-month Treasury bond). We include fixed effects for years, industries (2-digit 11 We recognize that price and non-price terms of the loans are co-determined during negotiations between members of the lending syndicate and the borrower. However, it is difficult to disentangle each component without any theory-based simultaneous model. Our previous analysis based on matched loans indicates that loans to backdaters do not differ substantially in terms of non-price loan characteristics, which ensures that this simultaneous determination should not affect our findings in regards to loan spreads. 18

19 SIC codes), loan type, loan purpose, and debt seniority. Standard errors are adjusted for heteroskedasticity and loan package level clustering. Estimated coefficients and levels of significance are reported in column 1, panel A of Table 4. We find that the coefficient associated with Backdater is not statistically significant, indicating that, prior to the revelation of unethical behavior, backdaters are paying a similar cost of loans as nonbackdaters. The variable identifying the post-revelation period is associated with a large and statistically significant coefficient of approximately 57 bp, likely due to the fact that the post-revelation period coincides with the onset of the financial crisis while we do not explore the matter further, academic and anecdotal evidence indicates that a drop in funding liquidity in the banking sector during the crisis lead to an increase in loan spreads for all borrowers. Finally, the coefficient of the interaction (Backdater Post revelation) is significantly positive, suggesting that the revelation of option backdating significantly increases the cost of bank loans for backdaters. The impact is also economically significant. While the cost of loans increases for all borrowers by 57 bp after 2006, the cost of bank loans for backdaters increase by an additional 19 bp after the revelation period. Given that the average firm in the study has a loan spread of 229 bp, this constitutes roughly an 8 percent increase in loan spreads for backdaters. In column 2, panel A of Table 4, we replace backdater by backdated grants (the number of backdated grants) to test if the frequency of backdating matters. We find that, post revelation, there is a significantly positive relation between the cost of loans and the number of option grants that are likely to have been backdated, with each additional backdated option grant leading to a 17 bp increase in the cost of debt. We note that firm size might affect the relation between option backdating (and the consequent perceived increase in information risk post revelation) and cost of loans. On one side, small firms are considered to be less informational transparent and subject to greater information asymmetry between insiders and investors. On the other side, large firms are likely to have access to multiple sources of capital, benefiting both from multiple banking relationships and the access to the bond market. Large 19

20 firms might be able to mitigate the impact on the cost of debt thanks to greater bargaining power, or by increasing their reliance on sources of capital that are less sensitive to soft information (for example, by relying on public debt markets rather than private loans). Hence, we expect the impact of the revelation of option backdating on the cost of loans to be greater in small, rather than in large firms. We accordingly subset our sample by total assets and report the results in column 1-4, panel B of Table 4. We find that, for small borrowers (total assets below the median) the impact is much greater; loan spreads increase by about 27 bp. On the other side, we find no statistically significant impact for large borrowers. To further test whether the increase in cost of debt is linked to information risk we report results using alternative measures of borrower transparency. In column 5-8, panel B of Table 4, we divide our sample into firms with more analysts following (analysts following above the median) and the ones with less analysts following (analysts following below the median). We find that loan spreads increase by 26 bp for firms with a smaller analysts following while the impact is not statistically significant for the borrowers with more analysts. Results in column 9-12, panel B of Table 4 also show consistent findings, using membership in the S&P 500 index as a proxy for transparency. Firms that are not members of S&P 500 face higher cost of loans (21 bp) while there is no statistically significant impact for the members of S&P 500. For all subsets, we document robust results when Backdater is replaced with Backdate grants (the number of backdated grants). *** Insert Table 4 about here *** 4.4. The impact of backdating on loan spreads for lenders with low information asymmetry in regression analysis The results thus far show that the cost of bank loans are higher for backdaters after the revelation of the backdating practice in 2006, which is consistent with the notion that banks are concerned about the quality of disclosure and the perceived level of information risk. Yet, the increase in the cost of debt is also consistent with the revelation of option backdating bringing to light weaknesses in internal governance and oversight. Accordingly, to test whether the impact of option backdating on loan spreads 20

21 is really due to an increase in perceived information asymmetry, we rely on an additional set of regression-based tests. In particular, we identify borrowers who are less affected by the decrease in perceived firm transparency, due to having gained superior knowledge about the firm via prior lending interactions. To explore this issue, we follow the prior literature to construct a dummy variable (Relationship loan) identifying lenders with prior lending relationship with the borrowers. We include Relationship loan by itself and an interaction term Backdater Post revelation Relationship loan. For completeness, we include also all the relevant two-way interaction terms. If prior lending relationship mitigates the adverse effect of increased information risk, we expect to see a negative coefficient associated with the three-way interaction term. The regression results are reported in Table 5. In this model, we again find that backdaters pay a similar cost of loans as non-backdaters prior to the revelation of the practice. We further find, in accordance with extant literature, that relationship loans are associated with lower spreads, by about 7 bp, for both backdaters and non-backdaters. Most importantly, we find that backdaters who receive loans from a new (non-relationship) lenders incur, post revelation, an increase in the cost of loan of approximately 34 bp. This result is highly statistically significant and the magnitude of the finding indicates that borrowers with no prior relationships with lenders are penalized even more than the average borrower. On the other side, the coefficient associated with the three-way interaction term is a negative 28 bp, indicating that the increase in the cost of loan for backdaters receiving relationship loans post revelation is tiny, approximately, 6 bp. We find similar result in column 2 of Table 5 by replacing backdater by backdated grants (the number of backdated grants). Consistent with our expectations, our evidence indicates that the increase in the cost of loans is mitigated by the existence of prior lending relationships with the borrower. *** Insert Table 5 about here *** Extant literature finds that lenders who are physically closer to borrowers are capable of mitigating information risk via more efficient monitoring (Agarwal and Hauswald, 2010). We hence test 21

22 whether the impact of the revelation of unethical behavior is mitigated by geographical proximity. In particular, we add a three-way interaction term between Backdater, Post revelation, and Average distance, measured as the average distance between the states in which the headquarters of the lender and borrower are located, in miles. The regression results are reported in Table 6. While the coefficient associated with the Backdater and Post revelation interaction variable is not statistically significant, the coefficient associated with the three-way interaction variable is positive and statistically significant. The cost of the loan to backdating borrowers increases, post revelation, by 2 bp for every 100 miles of distance between borrower and lender. Consistent with our expectations, we find that the increase in the cost of loans is mitigated by geographical closeness of borrowers and lenders. *** Insert Table 6 about here *** 4.5. The impact of backdating on bond spreads in regression analysis The results presented so far strongly indicate that lenders react to the revelation of unethical behavior by charging a higher cost of debt in response to a perceived increase in information risk. Consistently, our findings indicate that lenders with better access to firm-level information react less forcefully. In further analysis, we rely on the findings of extant literature, which identifies public-debt lenders (bond investors) as having both weaker incentives to monitor borrowers and inferior access to information. Based on this stream of extant literature, our expectation is that, given their greater exposure to information risk, public-debt lenders will react even more forcefully to the revelation of unethical behavior. On the other side, a second stream of extant research indicates weaker reliance by public debt markets on soft information, which could translate into a weaker impact at the revelation of unethical behavior. We replicate the analysis for the cost of debt presented in Table 4 by using a sample of public debt issues. Firm characteristics are controlled as was done in loan analysis. Bond characteristics are included in our model: Bond size, Bond maturity, Callable bond (binary variable), Puttable bond (binary variable), Subordinated bond (binary variable), and Bond rating. Prior bond issuance is also controlled which identifies whether the firm had issued bond before or not. 22