Why do Firms Issue Guaranteed Bonds?

Transcription

1 Why do Firms Issue Guaranteed Bonds? Fang Chen, Jing-Zhi Huang, Zhenzhen Sun, Tong Yu October 4, 2017 Chen is at the College of Business, University of New Haven. Huang is at the Smeal College of Business, Pennsylvania State University. Sun is at the School of Business, University of Massachusetts at Dartmouth. Yu is at the Car H. Lindner School of Business, University of Cincinnati. s: and We appreciate the comments from Van Son Lai, Mike Eriksen, and the participants of the International Finance and Banking Society 2015 Conference, the 2015 Midwestern Finance Conference, and the 2017 Financial Management Association Meetings. All errors are our own.

2 Why do Firms Issue Guaranteed Bonds? Abstract Corporates typically use affiliated firms as guarantors to issue guaranteed bonds, combining external financing with internal credit enhancements. Our finding shows that issuers with fewer tangible assets or lower credit rating are more likely to issue guaranteed bonds. Moreover, firms with more pronounced debt overhang problems and greater managerial agency problems tend to issue guaranteed bonds. While on average the use of guarantees improves bond ratings, it has little impact on bond initial yield spreads. Keywords: Guarantee, Credit Enhancement, Corporate Bonds, Credit Rating, Corporate Investments

3 1 Introduction One of the latest developments in the corporate bond market is the emergence of guaranteed corporate bonds, also known as credit enhanced corporate bonds. Guaranteed corporate bonds accounted for about fourteen percent of newly issued corporate bonds in terms of the issuance amount in With the guarantee arrangement, guarantors make payments to bond investors in case of issuer defaults. Partly owing to the recent financial crisis, the use of credit enhancement has received substantial attention (e.g., Arora, Gandhi and Longstaff, 2012; Demyanyk and Van Hemert, 2011). The spotlight is nevertheless on the packaging of credit enhancement with commercial loans, mortgage- and asset-backed securities, municipal bonds, and sophisticated structured products. 1 Little is known about the use of guarantees in the corporate bond market. Guarantees used in the corporate bond market are an interesting hybrid. Guaranteed bonds are typically insured by third party guarantors, making them different from traditional internal credit enhancement devices which improve bond rating by pledging collaterals, imposing covenants to restrict issuers activities or offering senior securities. Even so, since guarantors, the third party insuring guaranteed bonds, are either parents or subsidiaries of issuers, guarantees are neither an external arrangement to bond issuance. Motivated by the fact that guaranteed bonds are typically issued by affiliated firms with a trivial explicit guarantee cost, we introduce hypotheses that associate guaranteed bond issuance with debt overhang and agency problem. It is well noted that financial constraints limit corporates ability to borrow debt from investors. 2 Guarantees improve the bond creditworthiness and corporate financing ability. However, among variety of ways to enhance the bond creditworthiness, it is not clear which factors were the driving force for firms to choose guarantee. Given the unique feature of internal-arrangement insurance and trivial 1 On the applications of credit enhancements in mortgage backed securities, see Ashcraft and Santos, 2009; Griffin and Tang (2012); Arora, Gandhi and Longstaff (2012). Related to the applications of credit enhancement in municipal bonds, refer to Braswell, Nosari and Browning (1982), Kidwell, Sorensen and Wachowicz (1987), Nanda and Singh (2004). 2 For the literature regarding the effect of financial constraints, see Kaplan and Zingales (1997), Lamont, Polk and Saaá-Requejo (2001), and Baker, Stein and Wurgler (2003). 1

4 cost, we propose that the motivation of using guarantees in overcoming financial constraints is in two folders. On one hand, under the assumption that the use of guarantees lowers corporate required debt payments, we show that guaranteed bond issuance alleviates financial constraints due to debt overhang by increasing firm equityholder value. On the other hand, in a parent-subsidiaries context, agency problem stemming from misaligned incentives between parent firms and the subsidiaries is regarded as one of the main reasons for the inefficiency (Bolton and Scrafstein, 1998; Scharfstein and Stein, 2000; Ozbas and Scharfstein, 2010; Holod, 2012). In an inefficient internal market, the capital allocation may not obey the rule on project NPVs. Rather, firms undertake projects benefiting issuing companies but harmful to affiliated guarantors. The simple story leads to clear empirical predictions. First, firms with greater financial constraints are more likely to issue guaranteed bonds. Second, holding the firm rating constant, firms with more pronounced debt overhang problem are more inclined to use guarantee. Third, firms with greater agency problems rising from the misaligned incentives between the parent firms and the subsidiaries have higher odds of issuing guaranteed bonds after considering the firm rating factor. Empirical findings offer supports to the above predictions. While we do not find evidence that the likelihood of using guarantee is significantly related to three conventionally used financial constraint proxies, we show firms with poorer credit rating or less collateral are more likely to use guarantee on bonds. This is in line with the financial constraint explanation. Moreover, we find that firms with greater debt overhang problem are more likely to issue guaranteed bonds. That is, firms debt overhang, measured by the current value of bondholders rights to recoveries in default (Hennessy, 2004; Hennessy et al., 2007), is found to be significantly positively associated with the odds of issuing guaranteed bonds. Further, we present the evidence that guaranteed bond issuance may be a consequence of empirebuilding of the parent firms or corporate socialism among the subsidiaries. Specifically, firms with more severe agency problems featured with higher free high cash flow and lower growth opportunities are more inclined to issue guarantee bonds. 2

5 We perform two sets of additional analyses regarding guaranteed bond issuance for robustness. In the first, we differentiate between firms issuing guaranteed bonds only and firms alternatively issuing guaranteed and non-guaranteed bonds within a financial year. The result of multinominal logistic analysis shows that debt overhang is the main driver for issuing guaranteed bonds only as well as rotating between guarantee and non-guarantee bonds. Interestingly, the agency problem plays a significant role in the firms odds in firms issuing guaranteed bonds only, not in firms issuing both guaranteed and non-guaranteed bonds. In the second analysis, we address the concern that COMPUSTAT data reflects the consolidated financial information of corporates with a parent-subsidiary structure. We use the minority interest to estimate the financial variables at the parent firm level and then rerun the determinants analysis. Using the estimated income at parent firm level, we distinguish the operating firms from the financial holding firms. The result of the regressions with estimated variables at parent firm level and the dummy variable for the operating firms is consistent with that of the regression with the consolidated financial data. Given the fact that guaranteed bonds are issued with alternative drives, one may wonder whether the rating agency and the market have captured the incentives. We shed light on this question by performing further analysis to test the sequential rating and yield effects of guaranteed bond issuances. The result reveals that on average guarantee improves bond ratings but has no effect on initial yield spreads. More interestingly, when we respectively examine the individual effects of issuers financial constraints, debt overhang, and managerial agency problems on the relation between guarantee use and bond yield spreads, we find that guarantees used by bond issuers facing greater financial constraints or with more pronounced debt overhang indeed reduce yield spreads, while guarantees used by bond issuers with more severe managerial agency problems increase yield spreads. The offset of two effects provides a plausible explanation for no effect on yield spreads from guarantees overall. To summarize, guarantees enhance corporate debt capacity by allowing issuers to reach collateral resources beyond corporate borders. Such resources, however, are from affiliated firms. Our study focuses on the hybrid nature of guaranteed bonds. We show that, acting 3

6 like a double-edged sword, guaranteed bonds have the advantage to expand corporate debt capacity and alleviate debt overhang but its use may aggravate the agency conflict between parent companies and subsidiaries. The findings add to the stream of research on the efficiency of the internal capital market (Bolton and Scharfstein, 1998; Scharfstein and Stein, 2000; Ozbas and Scharftein, 2010). The remainder of the paper is organized as follows. Section 2 provides the institutional background of guarantees used in the corporate bond market and reviews the relevant literature. In Section 3, we introduce the hypotheses. Section 4 describes our sample and data used in the empirical analysis. Section 5 presents empirical findings. Section 6 concludes. 2 Background We begin this subsection with an example of a guaranteed corporate bond. On September 12, 2000, MGM Mirage issued a 10-year 8.5% senior bond using all its wholly owned domestic subsidiaries to provide a guarantee. The aggregate par value of the bond issuance is 850,000,000 USD. The guarantee is an unsecured senior obligation of the guarantor. At the time of bond issuance, MGM Mirage has a rating right below the investment grade: Ba1 by Moody s 3 while the newly issued bond is rated at an investment grade, BBB- by S&P and Baa3 by Moody s. The use of guarantees on newly issued bonds has increased substantially since its inception in the early 1990s. Reported in Table 1, based on par value, the market share of the guaranteed bonds in all US corporate bonds issued in a year rises from 1.4% in 1993 to 18% in Guaranteed bond issues peaked in 2009, with 37% of US corporate bonds being guaranteed. The aggregate par value of US corporate bonds during the period is roughly $17 trillion, of which 14% (i.e., $2.3 trillion) were issued with guarantees. Based on the Mergent Fixed Income Securities database (Mergent FISD), guarantees 3 Issuer ratings are assessed based on issuers ability to honor senior unsecured financial obligations and contracts. Issuer ratings share the same scheme as corporate bonds. Based on Moody s rating scheme, Baa3 is the lowest investment grade. 4

7 are one of three types of credit enhancements for corporate bonds, in addition to bond insurance and letters of credit. In terms of issuance amount, guaranteed bonds account for over 96% of corporate bonds with credit enhancement in the period from 1993 to Corporate bonds issued with bond insurance and letters of credit merely account for 3% and less than 1%, respectively, of the market. Documented in Nanda and Singh (2004), bond insurance is frequently used in the municipal bond market; roughly 50% of municipal bonds are packaged with bond insurance. Nanda and Singh (2004) suggest that the tax-exempt benefit of municipal bonds can be extended through third-party insurance, and is the main reason for the frequent use of bond insurance in municipal bonds. Corporate bonds are not eligible for tax exemption, thus tax benefit does not provide incentive to corporates to use guarantee on bonds. 4 The super majority of guarantors of guaranteed corporate bonds are either parent companies or subsidiaries of bond issuers. A clear advantage of using internal resources from affiliated firms to guarantee a corporate bond is its efficiency making use of available corporate resources. While a parent firm and its subsidiaries are separate legal entities, guarantee bounds their risk together. For example, without guarantees, if subsidiaries are in default, subsidiary bondholders do not have any recourse to the parent companies unless the parent companies are involved in some wrong-doing (Thomson, 1991). With a guarantee from the parent company, the subsidiary debtholders have recourse to its parent guarantors should the subsidiary default. In practice, most of the guarantees to the public issuers are provided jointly by most or all of their domestic subsidiaries. The guarantees are normally senior obligations of the guarantors, ranked equally with all other existing and future senior debt of the guarantors in right of payment. It is also a common practice to contain the covenants in the indenture to limit the payments of the guarantors, such as dividend payout, share repurchase or making principle payment prior to the schedule, among other arrangements. Consequently, this arrangement helps to reduce credit risk of guaranteed bonds. Alternative types of credit enhancements differ in issuance expenses. In early years the 4 Our sample only contains 4 bond issues backed by bank letter of credits, out of 11,226 corporate bonds issued with credit enhancements. As a result, we focus our comparison on bond guarantees and insurance. 5

8 parent/subsidiary guarantee was treated as an internal arrangement which requires a filing to the Securities and Exchange Commission (SEC) (no such filing requirement for insurers or banks when they offer credit enhancements). Nevertheless, the SEC filing expense for guaranteed bonds was as trivial as a few thousand dollars. The requirement of filing was abandoned by the SEC in 2003, thus no more filing expense is associated with guarantees. The other direct floatation cost is the guarantee fee. As the internally arranged guarantees are considered as an arm s-length transaction, parents/subsidiaries guarantee fee is seldom included in the contract. In contrast, the insurance fee is explicit and ranges from 0.5 to 2.0 percent of total debt (Cole and Officer, 1981). In summary, as an internal arrangement, guarantees to corporate bonds virtually have no direct cost. This is different from bond insurance and letters of credit, which typically involve explicit expenses. Nevertheless, as indicated in the later analysis, the use of guarantees is not costless indirect costs may arise. Like the example discussed in the beginning of the subsection, credit ratings of guaranteed bonds differ from those of bonds with insurance or letters of credit. At the time when bonds are issued, bond insurers and letters of credit providers typically have an AAA rating. Therefore bonds backed up by insurance or letters of credit have an AAA rating. By contrast, guarantors and guaranteed bonds have their ratings varying from AAA as the highest to D as the lowest. Moreover, the interconnection between issuers and guarantors has a potential effect on credit ratings of bonds. For example, if a parent rating is CCC+ or lower while a subsidiary has a stand-alone credit rating of B- or higher, the subsidiary s final credit rating could be lowered accordingly due to the concern of extraordinary negative intervention from the parent firms. Guarantee strengthens the interconnection. When the subsidiary, acting as a guarantor, is downgraded, the rating agency would subsequently reevaluate the bond and lower the rating of the guaranteed bond. The liability entails the default risk to the guarantor. For example, Vitro, the largest manufacturer of glass containers and flat glass in Mexico, suffered a dramatic decline in operating income as a result of the global economic and financial crisis that began in The decline in its operating income caused Vitro to 6

9 default on certain financial obligations, including $1.216 billion in outstanding notes. Those notes were unsecured, but guaranteed by certain subsidiaries of Vitro, including some located in the United States. The default caused the involuntary bankruptcy of its U.S. subsidiaries (Porzecanshi, 2011). 3 Hypothesis Development When firms have free access to capital, guarantees and credit enhancements could play a little role in the corporate bond market and they would not be introduced in the first place. The real market, nonetheless, is far from being perfect. As noted in the existing literature, financial constraints (Kaplan and Zingales, 1997; Lamont, Polk and Saaá-Requejo, 2001; and Baker, Stein and Wurgler, 2003) and debt overhang (e.g., Hennessy, 2004; Hennessy et al., 2007) are major detriments to corporate financing activities and their investments. By issuing guaranteed bonds, firms internalize external funding activities. Therefore, the choices between guaranteed and non-guaranteed bonds, therefore, reflect corporate status quo prior to bond issuance. First consider the effect of financial constraints on the issuance of guaranteed bonds. Consider a financially constrained firm has a positive NPV project. The project s NPV is q (> 0) before adjusting the funding cost. Owing to financial constraints, the firm relies on capital infusion to materialize the investment opportunity. There is a wedge between the cost of internal capital and that of external capital, c i and c e respectively. The value of the project, q, is between c i and c e. That is, c i < q < c e. Clearly, the firm would invest in the project when it has internal capital but it has to forgo the opportunities when it has to raise capital externally. Using subsidiaries as guarantors, corporates issue guaranteed bonds to lower the external capital cost thus undertake the positive NPV project. Consequently, we expect that the greater financial constraints, the stronger incentive of the firm to use guarantees. This gives rise to the first hypothesis. 7

10 H1. Firms with more financial constraints are more likely to issue bonds with guarantees, all else being equal. An alternative driver for a firm to forgo a positive NPV project is the so-called debt overhang problem. First suggested by Myers (1977), equityholders could forgo the positive NPV project if the return of such investment goes to bondholders. Note that debt overhang is harmful to both bondholders and equityholders because when rationally expecting equityholders to have the underinvestment incentive, potential bondholders would be reluctant to lend their money or charge a high cost of debt at issuance. Myers (1977) suggests various remedies, such as including protective covenants to restrict equityholders activities in order to alleviate such a debt overhang problem. In spite of variations, the essential purpose of the proposed remedies renders equityholders less likely to forgo positive investment projects. Aligned with this idea, Stulz and Johnson (1985) suggest that firms may issue secured bonds to reduce the debt overhang problem in the sense that because secured debt has collateral, secured creditors rely less on the new investments than unsecured creditors when firms are at default, thus equityholders are less likely to forgo the investment opportunity. Alternatively, Mayers and Smith (1987) suggest to include a bond covenant that requires the issuer to purchase protective coverage. The protection pays off the loss or makes up the cash flow shortfall, reducing the expected downside loss of the project. This results in greater incentive for equityholders to invest in the positive NPV project. Issuing guaranteed bonds potentially combines the benefits of secured bond issuance and protective coverage. When the parent defaults, the bond guarantors hold the responsibility to make promised payments This makes a guaranteed bond like a secured bond. Guarantees from affiliated firms also act as if a standby insurance to the bond issuer. Thus, we expect that firms issue guaranteed bonds to address the debt over hang problem. This gives rise to the second hypothesis. H2. Firms with more pronounced debt overhang problem are more likely to use guarantees on bonds, all else being equal. 8

11 Despite that there is no or little explicit fee for the the guarantee provided by affiliated firms, the use of guaranteed bonds to overcome financial constraints is not completely free of costs, such as that the internal arrangement of guaranteed bonds may restrict guarantors to issue new bonds. Especially in the parent-subsidiary context of guarantee, the incentive of a parent firm may be misaligned with that of subsidiaries, regardless of the cost of guarantee issuance to subsidiaries. As a result, the allocation of internal resources in the form of a guarantee may not follow the investment opportunities. Instead, the parent company could use the guarantee for an empire-building purpose 5. When there are financial constraints for conventional bond issuance and the implicit cost of guarantee is not the concern of the parent company due to the agency conflict, guarantee is a preferred way to raise capital even when the benefit from the new investment opportunity is negative. This reasoning results in a link between corporate agency conflicts and incentive to issue guaranteed bonds. This leads to our third hypothesis. H3. Firms with greater agency problem are more likely to issue guaranteed bonds, all else being equal. Finally we propose a hypothesis related to yield spreads of guaranteed bonds. For firms facing financial constraints from default risk, as the use of internal guarantees reduces the default risk and thus mitigates constraints, yield spreads of guaranteed bonds are naturally lower than those of non-guaranteed bonds, all else being equal. H4a. Guaranteed bonds have a lower yield spread than non-guaranteed bonds when issuers face financial constraints, all else being equal. The debt overhang is an internal constraints to the investment due to the conflict be- 5 For example, Bolton and Scharfstein (1998) point out that allocating capital to divisions with opportunities aligns with parents empire-building preference. Scharfstein and Stein (2000) argue that a two-tier agency problem, stemming from misaligned incentives at parents and at divisions, is necessary for corporate socialism in internal capital allocation. Ozbas and Scharfstein (2010) find that the ownership stakes of top management have a positive relation with the extent of Q-sensitivity differences, suggesting that agency problem leads to the inefficient capital market. 9

12 tween bondholders and equityholders. Guarantees alleviate the debt overhang problem by increasing the equityholders value and therefore increase the chance of investing on positive NPV projects. The higher cost from debt overhang, the higher benefit from guarantee and the lower yield spread is required by rational bond investors. H4b. Guaranteed bonds have a lower yield spread than non-guaranteed bonds when issuers have more debt overhang, all else being equal. However, under the agency conflict scenario, the purpose of issuing guarantee bonds to overcome the financial constraints is mainly for empire building. A rational bond investor may require higher yield due to the risk from the misalignment of the growth opportunities and guarantee bond issuance. H4c. Guaranteed bonds have a higher yield spread than non-guaranteed bonds when issuers have greater agency problem, all else being equal. Taken together, the combined effect of guarantees on debt value can be positive or negative which remains as an empirical question. 4 Data 4.1 Sample This study utilizes the Mergent FISD database which provides issues and issuers information for bonds issued by both public and private firms. We select U.S. corporate bonds including U.S. corporate debenture, corporate midterm notes (MTNs), asset-backed securities and other corporate bonds issued by U.S. firms. Mergent FISD contains information on whether corporate bonds were issued with any type of credit enhancements: (i) guarantees, (ii) letters of credit (LOC) and (iii) bond insurance. Our sample ranges from 1993 to Among a total 123,034 corporate bonds, 11,226 corporate bonds were issued with credit enhancements. Specifically, there were 11,551 guar- 10

13 anteed bond issues, 441 issues with bond insurance, and 4 issues backed by bank LOC. We identify the guarantors of the issues with guarantees via two ways. First, Mergent FISD directly lists most guarantors as Subsidiaries. Second, the database lists the parent of the issuers. If a parent s identification number matches a guarantor s identification number, the bond was guaranteed by the parent firm of the issuer. If guaranteed bonds have no information on guarantors, we manually search issuers SEC filings. The guarantors information is disclosed in the 424B, S-4, 8-k, 10-Q or 10-K. This study investigates the public issuers only. We restrict our sample to the issuers that are listed at the time of bond issuance by matching the CUSIP of the issuers in FISD with that in the Center for Research in Security Prices (CRSP) database. Consistent with prior studies, e.g., Custdio, Ferreira, and Laureano (2013) and Becker and Josephson (2016), we exclude financial firms (SIC codes ) and regulated utility firms (SIC codes ) because they have significantly different capital structures and are subject to different regulations. We further exclude bonds issued with bond insurance or letters of credit (LOCs) to focus on the use of guarantee. There are 8,797 corporate bonds issued by public firms, of which 825 are guaranteed bonds. To facilitate the analysis, we remove bonds when issuers total assets, firm ratings and bond ratings are not available. We measure firm ratings using the most recent S&P long term issuer credit ratings right before bond issuance (Frating(AAA=26,D=1)) in Compustat, also in line with prior studies such as Norden and Kampen (2013), Alp (2013), and Baghai, Servaes and Tamayo (2014). We obtain bond ratings at issuance from Mergent FISD. As there are ratings from multiple rating agencies, we use the S&P rating if it is available for a specific bond. In case the S&P rating is missing in Mergent FISD, we alternatively use the Moody s rating. If both S&P and Moody s ratings are missing, we use the Fitch s rating. The highest rating number is 26 (equivalent to S&P s AAA) and the lowest rating number is 1. Our treatments result in a final sample consisting of 7,201 corporate bonds, of which 731 bonds with guarantees (see Figure 2). Since all guaranteed bonds are guaranteed by either the subsidiaries or the parent firms, 11

14 to run a fair comparison, we require non-guaranteed bonds in our sample to hold similar corporate structures. We use Capital IQ to search relevant information manually. Capital IQ lists the parent firms and the subsidiaries in its corporate tree section. In the final sample, 5,949 bonds were issued by firms which have parent firms or subsidiaries, of which 647 bonds were packed with guarantees. This is the subsample used in the regressions of the determinants of guarantee use. Figure 2 provides a detailed description of the number of bonds with guarantee issued by publicly listed companies in the sample. We obtain accounting data from Compustat, stock return data from CRSP, and bond issuance data from Mergent FISD. From FISD, we obtain the bond issuance information such as time to maturity, initial bond yield spread, bond ratings, the indicators for callable bonds, putable bonds, and secured bonds, and so on. Then we merge sample bonds with the Compustat data by the issuer ID. We obtain firm variables from the Compustat data including firm rating, total assets, operating income, tangible assets, sales, age, dividend, debt, cash flow and free cash flow. The accounting variables are at the fiscal year end before the debt issuance. As Compustat provides the financial data from the consolidated statements, the financial variables are measured at the group level except the issuer s firm rating from S&P. When S&P assesses the credit rating of a firm within a parent-subsidiary structure, S&P considers the firm s stand-alone credit profile and the support or intervention from other firms within the group (see Standard & Poor s, 2013). We consider having no financial variables available at the parent firms level as a limit of our data when we run the regressions on the parent firms. 6 In order to examine the validation of our test, we use the minority interest method to approximately estimate the financial variables in the parent firms only and run a robustness check. 6 We also checked FactSet, one database available to academics that includes information for both parent companies and subsidiaries. Unfortunately, FactSet has very little financial data on subsidiaries available. We find that data are available only for those subsidiaries that used to public and only for the period when they were public. 12

15 4.2 Summary Statistics In Table 1, we present the distribution of guaranteed bonds by both public and private firms in Mergent FISD. It shows that the fraction of guaranteed bonds in Mergent FISD is about 14% (9%) in terms of the aggregate par value of bonds (the number of bonds). Focusing on the public firms, the percentage of guaranteed bonds issued by public firms is 9% (8%) in terms of the aggregate par value (number) of bonds. The fraction of aggregate par value of bonds issued by public firms ranges between 1.19% in 1994 and 24% in Figure 3 presents the percentage of guaranteed bonds in terms of aggregate issuance amount over time. Shown in Panel A, the percentage of guaranteed bonds issued by all firms increases over time and peaks in Panel B shows that the percentage of guaranteed bonds issued by public firms varies and peaks in 2009 as well. This finding is consistent with that in Table 1. In Table 2, we provide the summary statistics of guaranteed bonds and non-guaranteed bonds as well as their issuers characteristics. In both panels A (for non-guaranteed bonds) and B (for guaranteed bonds) of the table, the first three columns report the distributions (mean, median and standard deviation) of bond issuer and issue attributes, under the heading All. The average firm s credit rating (16.13, corresponding to S&P s BB+, relative to 26 for an AAA rating) of guaranteed bonds is lower than that of non-guaranteed bonds (17.80, corresponding to S&P s BBB). Given the fact that the threshold point between investment grade bonds and speculative grade bonds is BBB- by S&P (the numeric rating of 17.00), the mean firm rating for guaranteed bonds is right below the threshold and the mean firm rating for non-guaranteed bonds is right above. Not tabulated, we find the correlation between rating numbers and a dummy variable of whether a guaranteed is used to be -0.15, suggesting that, without considering issuer and bond characteristics, the average rating of guaranteed bonds is unconditionally lower than that of non-guaranteed bonds. Besides issuer rating, it is also shown in the table that the guaranteed bond issuers have slightly lower Tobin s Q (1.97) than non-guaranteed bond issuers (2.07). Guaranteed bond issuers also make less profit, borrow more debt, and hold fewer assets than non-guaranteed 13

16 bond issuers. Taken together, the comparisons of issuer characteristics reveal that guaranteed bond issuers unconditionally have a poorer growth prospect and worse business condition than non-guaranteed bond issuers. The comparisons of bond characteristics are also reported in Table 2. The average yield spread of guaranteed bonds ( basis points) is much larger than that of non-guaranteed bonds ( basis points). Interestingly, guaranteed bonds typically have larger par value and shorter time to maturity than non-guaranteed bonds. Next, we take a close look of the distribution of guaranteed and non-guaranteed bonds by dividing them into eight sub-groups based on the issuers ratings: AAA, AA, A, BBB, BB, B, CCC and others. We compute the percentage of the aggregate par value in each issuer s rating for guaranteed bonds and non-guaranteed bonds respectively. This is plotted in Figure 4 the majority of guaranteed bonds were issued by firms with the ratings of BBB or BB while most of the non-guaranteed bonds were issued by firms with the ratings of A or BBB. We also report, in Table 2, the result of univariate comparisons for the variables between guaranteed bonds and non-guaranteed bonds in the above four firm rating groups. The comparison of all variables between guaranteed bonds (issuers) and non-guaranteed bond (issuers) in each rating group is consistent with the overall comparison reported in the first three columns. Nevertheless, the magnitude of the difference between the guaranteed bonds and the non-guaranteed bonds is dynamic in each firm rating group. The mean Tobin s Q of guaranteed bond issuers is 2.09 and that of non-guaranteed bond issuers is 2.49 in the AAA-A group, while it is 1.78 for both guaranteed bond issuers and non-guaranteed bond issuers in the B-CC group. In the AAA-A group, the mean yield spread of guaranteed bond issuers is basis points (bps) and that of non-guaranteed bonds issuers is bps, while in the B-CC group, the mean yield spread is bps for guaranteed bond issuers and bps for non-guaranteed bond issuers. 14

17 4.3 Regression Variables Now, we discuss all the variables to be used in hypothesis tests conducted in Section 5. To be specific, in order to examine the effect of financial constraints on corporate guarantee use (the first hypothesis), we include the following variables in panel regressions. The main dependent variable is GT, an indicator variable equal to one if it is issued with a guarantee and zero otherwise. The main independent variables include financial constraint measures, firm characteristics and bond characteristics. Financial constraint measures are: the KZ index (KZ ), the WW index (WW ), the SA index (SA), Tangible (Tangible) and issuer credit rating (Frating). We follow Lamont, Polk, and Sa-Requejo (2001) to construct the KZ index (Kaplan and Zingales, 1997) as follows: KZ = Cash Flow/K Tobin s Q (1) Debt/Total Capital Dividends/K Cash/K where Cash Flow is computed as (Income Before Extraordinary Item + Total Depreciation and Amortization), K as PP&E, Tobins Q as (Market Capitalization + Total Shareholder s Equity - Book Value of Common Equity - Deferred Tax) / Total Asset, Debt as (Total Long Term Debt + Current Portion of Long Term Debt), Total Capital as (Total Long Term Debt + Current Portion of Long Term Debt + Total Shareholder s Equity), Dividends as (Total Cash Dividends Paid), Cash as (Cash + Short-Term Investment). Following Whited and Wu (2006), the WW index is computed as WW = CF DIVPOS TLTD (2) LNTA ISG SG where CF is the ratio of cash flow to total assets, DIVPOS is an indicator that equals one if a firm pays cash dividend and 0 otherwise, TLTD is the ratio of long-term debt to total assets, LNTA is the natural logarithm of total assets, ISG is a firm s industry sales growth, and SG is a firm s sales growth. 15

18 The SA index is computed using the following formula as in Hadlock and Pierce (2010): SA = Size Size Age (3) where Size is the natural logarithm of inflation-adjusted total assets, Age is the number of years the firm is listed in Compustat. Another financial constraint proxy, Tangible is measured by the ratio of a firm s property, plant and equipment to its total assets. It is a proxy for collateral to increase the creditworthiness of bonds. Finally, Frating is the most recent S&P long term issuer rating before bond issuance. We set a numerical score 26 for AAA rated bonds and 1 for bonds receicing a rating D. Firms with higher KZ, higher WW, higher SA, lower Tangible and lower Frating are more likely to have financial constraints than firms with lower KZ, lower WW, lower SA, higher Tangible and higher Frating under tightened financial conditions. In unreported results, we find that five constraint measures are highly correlated with each other. For example, the rank correlation between size and the WW index, SA Index are and respectively. Although these proxies may be picking up similar information, it nonetheless appears that each measure picks up some unique information as well. KZ, WW and SA place a heavy weight on the firms ability of internal funding capacity while Tangible and Frating primarily focus on the creditworthiness. We thus use each of the alternative measures as financial constraints. Firms characteristics variables include Size, Profit, FCF and Dividend and bond characteristics variables include MktYld, Term, Par, Call, and Put. MktYld is the Moody s yield spread of corporate bonds (the average monthly yield of Moody s AAA and Baa bonds) over corresponding treasuries before issuance. It is used to control for the timing of bond issuance by a firm. The lower of the average yield of in the bond market (i.e., lower debt financing cost), the more likely firms are to use guarantees to increase their debt capacity. Term is the logarithm of the bond s time to maturity in years. Par is measured as the logarithm of the total offering amount in millions of dollars. Call (Put) is the indicator for a bond equal to one if the bond has a call (put) provision. A larger par (Par), a longer time to maturity 16

19 (Term) or a call option (Call) increase the risk for investors and the need for guarantee as credit enhancement. A put option (Put) decreases the risk for investors. The under-investment hypothesis (the second hypothesis) implies that the debt overhang problem is an obstruction to new debt issuance and leads to under-investment. Guarantee can be used to mitigate the under-investment problem. The main independent variable used in the test of the effect of under-investments on corporate guarantee uses is DOH. We construct DOH by following the approach of Hennessy (2004) and Hennessy, Levy, and Whited (2007). DOH = Leverage t Recovery [ 20 ] ρ t+s [1 0.05(s 1)](1 + r) s s=1 where Leverage is the ratio of total debt to total assets. Recovery by SIC is based on Altman and Kishore (1996). ρ t+s denotes the probability of default in period t+s based on debt rating on date t. We use the average default rates by ratings from Moody s. Similar to Hennessy, Levy, and Whited (2007), we ignore short-term debt and assume that long-term debt mature in a straight-line fashion, with 5% of the original debt maturing each year. r is the yield to maturity for zero coupon bonds with maturity of 1 to 20 years. The yield data are obtained from Gurkaynak, Sack, and Wright (2007). Firms have higher tendency for over-investment when agency problem is more serious. It has been well noted in the literature that the lower growth opportunities and the higher free cash are associated with higher agency problem (Jenson, 1986; Opler and Titman, 1993). Following Kolasinski (2009) and Custdio, Ferreira and Laureano (2013), we use a firm s Tobin s Q to measure growth opportunities. We sort firms into three equal numbered groups based on the issuers Q and their free cash flow (FCF ) respectively. Q is the ratio of market value to book value computed as the total assets minus total book value of equity plus market capitalization and then divided by total assets. FCF is computed as the EBITDA minus the sum of XINT, TXT, DVC, and DVP and then divided by total assets. The main independent variable used in the test of the effect of over-investments on corporate guarantee uses (the third hypothesis) is LQ*HFCF. LQ*HFCF equals one if the issuer is simultaneously in the bottom Q tercile group and top FCF tercile group and zero otherwise. (4) 17

20 5 Empirical Results In this section, we empirically test the impact of financial constraints, debt overhand and agency problem on the odds of issuing guaranteed bonds. Further, we examine the effect of guarantees on bond ratings and yield spreads at issuance. 5.1 Determinants of Guarantee Use In this subsection, we explore the determinants for firms to issue guaranteed bonds. Specifically, we examine a set of factors in a logistic function that models the probability of a bond issuer to use the guarantee. In the logistic regression, the dependent variable is the guarantee dummy equal to one if a newly issued bond is a guaranteed bond and zero otherwise, and the independent variables are a set of variables explaining bond issuers propensity to use a guarantee Effect of Financial Constraints We test the first hypothesis that the high level of firms financial constraints drives guarantee uses. The main independent variables are five financial constraint measures (WW, SA, KZ, Tangible and Frating). The control variables include firm characteristic variables (Size, Profit, FCF, and Dividend) and bond characteristic variables (Term, Par, Call, and Put). To control for time-varying macroeconomic factors and industry specific factors, we also include the yield spreads of Moody s corporate bonds over corresponding treasuries before issuance (MktYld), the fixed year effect and fixed industry effect. All variables are defined in Section 4 and Appendix B. The results are reported in Columns 1-5 of Table 3. In all the columns, the standardized beta of the independent variables is reported. We first test the effect of three financial constraint indexes on the guarantee. Owing to the high correlations between three indexes (WW, SA, KZ ) and firm characteristic variables (Size, Profit, FCF, and Dividend), we include three financial constraints indexes and the firm characteristic variables in the separate regression 18

21 specifications. In Columns 1 to 3, the coefficients of WW, SA and KZ are insignificant. In Columns 4 and 5, we use two other proxies of the financial constraint: Tangible and Frating. In Column 4, we first include Tangible in the regression with firm and bond characteristic variables as control variables. The standardized beta of Tangible is and significant at the 1% level. In Column 5, we include Frating(AAA=26,D=1) and Tangible in the regression with the same set of control variables in Column 4. The standardized betas of Frating and Tangible are and -0.19, respectively. Both are significant at the 1% level. The result shows that firms with less tangible assets and lower credit rating are more likely to use guarantee on bonds. In terms of economics significance, one standard deviation increase of Frating results in is associated with standard deviations decrease in the log odds of using guarantees. In other words, one standard deviation increase in Frating is associated with % probability decrease of using guarantees. In sum, Table 3 shows the effect of five financial constraints measures on the guarantee use. Although the three financial constraints indexes that lean more on the internal funding capacity of firms are insignificant in the regressions, Frating(AAA=26,D=1) and Tangible that weight more on the creditworthiness of firms are statistically and economically significant. The result supports the first hypothesis that financial constraints is one of the main driver of the guarantee use by firms Effects of Debt Overhang and Managerial Agency Conflict In this subsection, we present the analysis to evaluate the effects of debt overhang and agency problem on corporate use of guarantees. The results are reported in Columns 6 and 7 of Table 3. In Column 6, we include DOH, Tangible and Frating and firm and bond characteristic variables. The standardized beta coefficients on Tangible and Frating(AAA=26,D=1) are significantly negative. This is consistent with the result of the effect of the financial constraints on guarantee use. The standardized beta of DOH is and significant at the 5% level. One standard 19

22 deviation decrease in DOH increases the odds of using guarantee by standard deviations. The marginal R 2 of adding DOH is by comparing specifications with and without DOH. The result suggests that firms with more debt overhang are more likely to use the guarantee on bonds. The implication is that with guarantee on bonds, firms mitigate the debt overhang problem and thus reduce the under-investment. The result supports the second hypothesis that firms with more pronounced debt overhang problem are more likely to issue guaranteed bonds. In Column 7, we use LQ*HFCF to proxy the agency problem. Other independent variables include DOH, Tangible, Frating, firm and bond characteristic variables. The standardized betas on DOH, Tangible and Frating is 0.088, and , respectively and all are statistically significant. The standardized beta of LQ*HFCF is and significant at the 5% level. One standard deviation decrease in LQ*HFCF increases the odds of using guarantee by standard deviations.the marginal R 2 of adding LQ*HFCF is by comparing specifications with and without LQ*HFCF. The result indicates that firms with greater agency problem are more likely to issue the guaranteed bonds. In other words, firms with fewer growth opportunities have more tendency to use guarantee to issue bonds and thus may induce the over-investment. The result supports the third hypothesis that firms with greater agency problem are more inclined to issue guaranteed bonds. Column 7 reveals some interesting findings with all independent variables in the regression. First, larger firms and less profitable firms are more likely to use guaranteed bonds. Second, putable bonds and shorter maturity bonds are less likely to be guaranteed bonds. Third, when the Moody s corporate bond yield spread is higher, firms are more likely to use a guarantee. Among all the significant independent variables, Frating has the largest marginal R 2 which is Tangible has the second largest marginal R 2 which is The marginal R 2 for DOH is and for LQ*HFCF is Frating also has the largest standardized beta. 20

23 5.1.3 Time Variation of Guarantee Use The above evidence suggests that guarantee use is a consequence of pronounced debt overhang or greater agency problem. However, guarantee use in our sample is not constant over time. In this subsection, we explore the time variation of guarantee use. In the logistic regression, we include the year dummy. Other independent variables for the test are the same as those in Column 7 of Table 3. The results are reported in Table 4. The standardized betas of 20 Year Dummy are presented. From 1993 to 2005, five out of thirteen Year Dummy are positive and statistically significant. The significant standardized beta of Year Dummy ranges from to From 2006 to 2012, all the Year Dummy are positive and statistically significant in a row. Moreover, overall the standardized betas have a large increase since year 2006 comparing to the years before. The impact of Year Dummy on the guarantee use starts to have a big jump in year 2006 and peaks at year Specifically, the Year Dummy for year 2006, 2007, 2008, 2009, 2010, 2011 and 2012 is 0.225, 0.197, 0.193, 0.295, 0.318, and 0.262, respectively. It shows that the use of guarantee is dynamic along the time. 5.2 Robustness Check of the Determinants of Guarantee Use Multinomial Logistic Regression on Guarantee Use It is not unusual for a firm to issue multiple bonds in a year. Some firms purely issued either non-guaranteed bonds or guaranteed bonds while other firms alternatively switched between them. In this subsection, we investigate the determinants of the firms choice of among three issuances strategies: non-guaranteed bonds only strategy, guaranteed bonds only strategy and switching strategy between guaranteed bonds and non-guaranteed bonds. If a specific set of determinants drives corporate to use guarantee on bonds, it is interesting to examine whether the determinants play a role solely in issuing guaranteed bonds or issuing mixed bonds or both. The multinomial logistic model is used and it is specified as follows: 21

24 Log ( ) P (Ym = 1, 2) = α m + P (Y i = 0) K β mk X ik + ε (5) k=1 The three categories of the dependent variable (Y ) are: 0: if the firm issues non-guaranteed bonds only in a year. 1: if the firm issues both non-guaranteed bonds and guaranteed bonds in a year. 2: if the firm issues guaranteed bonds only in a year. In the analysis, the baseline group is Y i = 0. X denotes a vector of independent variables: Tangible, Frating, DOH, LQ*HFCF, Size, Profit, FCF, Dividend, MktYld, Term, Par, Call, Put. The result is reported in Table 5. The regression result in Column 1 is for Y i = 1 in which the dependent variable is the log odds of issuing mixed bonds to purely issuing nonguaranteed bonds. The regression result in Column 2 is Y i = 2 in which the dependent variable is the log odds of purely issuing guaranteed bonds to purely issuing non-guaranteed bonds. The coefficient of Frating is in the mixed strategy group while it is in the pure guarantee strategy group. That is, holding other variables constant, with a one-unit decrease in the firm rating, the odds of using the mixed strategy (i = 1) to issuing nonguaranteed bonds only (i = 0) increases 0.496, and the odds of purely using the guarantee strategy (i = 2) over purely using non-guaranteed bonds (i = 0) increases For Tangible, the coefficient is in the mixed strategy regression and in the purely guaranteed bonds strategy regression. The result shows that firms with lower rating and less tangible assets prefer to issue either mixed bonds or guaranteed bonds to non-guaranteed bonds. The coefficient of DOH is significant at the 1% level in both regressions. Specifically, the coefficient is in the mixed bond group regression and in the purely issuing guaranteed bonds regression. Holding other variables constant, with a one-unit increase in DOH, the odds of using the mixed strategy (i = 1) to issuing non-guaranteed bonds only (i = 0) increases 0.251, and the odds of purely using the guarantee strategy (i = 2) over purely using non-guaranteed bonds (i = 0) increases The result indicates that firms with 22