THE EVOLVING WORLD OF RULE 144A MARKET: A CROSS-COUNTRY ANALYSIS

Transcription

1 THE EVOLVING WORLD OF RULE 144A MARKET: A CROSS-COUNTRY ANALYSIS Usha R. Mittoo Zhou Zhang* I.H. Asper School of Business Faculty of Business Administration University of Manitoba University of Regina Winnipeg, MB, Canada, R3T 5V4 Regina, SK, Canada, S4S 0A2 Tel: (204) Tel: (306) Fax: (204) Fax: (306) umittoo@cc.umanitoba.ca zhou.zhang@uregina.ca Abstract We compare debt issuances by U.S. and international firms to examine differences in the borrowing costs, measured by yield spread, between the Rule 144A and public debt markets across countries and over time in period. We find that the yield spread is 48 basis points higher in the 144A market than in public bond market, after controlling for other determinants of yield spread. The non-u.s. developed country issuers pay a similar borrowing cost as their U.S. peers do, but emerging country issuers pay significantly higher costs to access the U.S. debt market. The results hold after controlling for additional country-specific legal and institutional variables. We also find that the borrowing costs increased after the enactment of the 2002 Sarbanes-Oxley Act, and surged to the highest level in 2008 during the financial crisis for both U.S. and international issuers. JEL Classification: G32; G15; G01 Keywords: Rule 144A; Public debt issuance; Credit quality; Cross-country; Sarbanes-Oxley Act *We thank Arnold R. Cowan and conference participants at the 2011 Midwest Finance Association, 2010 Northern Finance Association and the 2010 Administrative Sciences Association of Canada annual meetings, and the seminar at the Nanjing University for their comments and suggestions. Usha Mittoo acknowledges financial support from the Social Sciences and Humanities Research Council of Canada (SSHRC) and the Bank of Montreal Professorship. Zhou Zhang (corresponding author) acknowledges support from the University of Regina Dean s Research Grant and Viterra Faculty Fellowship. 1

2 1. Introduction On December 9, 1998, Intrawest Corporation, a Canadian publicly traded company, issued $77 million private debt placement in the U.S. under Rule 144A of the Securities Act of In the following six years, Intrawest tapped the 144A market five times and raised $984 million for the acquisition and renewal of existing debt. By comparison, it raised debt only once ($133 million) in the U.S. public debt market in August Why does Intrawest a firm with access to both Canadian and U.S. public equity and bond markets prefer to raise capital in the Rule 144A market? What are the costs differences for Intrawest to raise funds in public debt versus Rule 144A market? How do these costs compare with those of the U.S. and other international issuers? The main objective of this study is to provide some insights into these issues by examining the evolution of the Rule 144A market. The U.S. Securities and Exchange Commission (SEC) approved Rule 144A in April 1990 to ease restrictions on the resale of private placements. Prior to the adoption of Rule 144A, firms accessed the private placement market to raise capital, primarily to escape the SEC s registration and disclosure requirements. However, they faced generally higher costs than in the public debt market because of higher information costs and a small number of investors. Moreover, the Investors were also not allowed the immediate resale of restricted securities increasing the liquidity problem. 2 The Rule 144A removed this restriction on immediate resale, providing firms with an additional channel of financing. As a result, the size of Rule 144A debt market increased from $6 billion in 1991 to about $212 billion in Source: the Thomson Reuters SDC Global New issues database. 2 The exemption of disclosure requirements from the SEC causes securities in private placement market to have higher levels of asymmetric information compared to securities in the public market. Blackwell and Kidwell (1988) examine the yield spreads between public and private placement debt and show that firms select the market with the lowest borrowing cost in their issuance decision. Low liquidity in the private placement market is because buyers can only resell the securities if they hold the securities for at least two years or subsequently register them. 2

3 Despite the growing importance of Rule 144A market, few studies have examined the costs and benefits of raising capital in the Rule 144A (hereafter 144A ) and public debt (hereafter PD ) markets and most have focused on U.S. issuers. 3 Fenn (2000) examines the post-issuance registration with the SEC for the 144A high-yield bonds issued by the U.S. nonfinancial firms during the period of and finds that more than 97% of the issues were subsequently registered with the SEC. He concludes that the 144A market provides significant benefits to both issuers and investors because it provides speculative-grade rated firms a speedy way of raising capital, and subsequent registration with the SEC lowers the liquidity premium for the 144A investors. Livingston and Zhou (2002) examine both investment- and speculative-grade non-convertible bond issues in the 144A and PD markets by U.S. non-financial firms in the period. They find that 144A bond issues have a higher yield spread than PD issues consistent with their profiles of lower liquidity and high information uncertainty. They also find that the impact of registration rights on yield spreads is significantly higher for speculative-grade issuers. Chaplinsky and Ramchand (2004) is the only study that compares the 144A and PD issues by non-u.s. firms. They examine issuances in the period and show that there is a bifurcation of 144A and PD markets for investment- and speculative-grade issuers. They find that non-u.s. investment-grade issuers raise capital on both markets but pay a higher financing cost in the 144A market. In contrast, speculative-grade issuers largely rely on the 144A market, and face a similar yield spread between the 144A and PD markets. They argue that 144A benefits international firms because it allows non-u.s. firms to access the U.S. capital 3 These studies have exclusively focused on the 144A and public debt issuance rather than equity issuance. Chaplinsky and Ramchand (2004) argue that the popularity of the debt issuance is substantially higher than equity issuance (about 8 times larger in total dollar amount issued). More recent studies expand the analysis to more securities such as convertibles bonds (see e.g., Huang and Ramirez, 2009). 3

4 market without paying a higher premium complying with the more stringent U.S. accounting and disclosure requirements. In this study, we examine non-convertible debt issued in the U.S. 144A and PD markets by U.S. and international issuers in the period. We contribute to the literature in two ways. First, in contrast to the prior Rule 144A literature that focuses either on the U.S. or international issuers, we study both U.S. and non-u.s. issuers in 144A and PD markets. We conduct a cross-country analysis to examine whether the borrowing costs vary across issuers in different countries and to test whether international firms pay a premium to access the U.S. capital market compared to their U.S. peers, and whether the premium differs between non-u.s. developed and emerging country issuers. Second, we cover an expanded time period that includes two major events that are likely to affect the yield spreads across countries and firms: the passage of Sarbanes-Oxley Act (hereafter SOX ) on July 30, 2002, and the 2008 global financial crisis. A large number of studies document that the SOX increased the reporting costs for U.S. and non-u.s. firms, especially for small and high risk firms. Several studies also document that the number of foreign firms cross-listing their equity on U.S. exchanges declined and many firms even delisted in the post-sox period due to the increased costs associated with internal controls and disclosure mandates for SEC reporting (see e.g., Engel, Hayes and Wang, 2008). However, to the best of our knowledge, no prior study has examined the effect of the implementation of SOX on the issuances in the 144A and PD markets. Our study fills this gap in the literature by comparing the issuance activity for international and U.S. issuers. Using a large sample of 144A (N=5,890) and PD (N=11,104) issues, we first study the relative attractiveness of 144A and PD markets for issuers from different countries over time. In 4

5 the full sample period ( ), the U.S. and other developed country issuers placed 33% of the total proceeds in the 144A market and the remaining 67% in PD market. This pattern is opposite for the emerging country issuers who raised 72% of their total proceeds in the 144A market and only 28% in the PD market. The analysis in the pre-sox and post-sox periods, however, shows that international issuers reduced their reliance on the 144A market in post-sox period whereas there is no significant difference for the U.S. issuers in pre-sox and post-sox periods. The decrease is pervasive among both developed and emerging country issuers; the percentage of total proceeds raised from the 144A market in post-sox period decreased by 55% for non-u.s. developed countries and 53% for emerging countries. Based on the analysis of the 144A market from 1991 to 1997, Chaplinsky and Ramchand (2004, p1095) argue that the evidence suggests that 144A could soon eclipse the public debt market for international firms. Our evidence, however, shows an opposite trend in post-sox period and suggests that the evolution of the 144A market could be impacted by other factors. In our multivariate analysis, we examine whether the borrowing costs between 144A and PD markets, estimated by yield spread between the original debt yield and the corresponding treasury yield, differ across countries and in the pre- and post-sox periods. Our main findings are as follows. First, consistent with prior studies, we find that the yield spread is 48 basis points (and statistically significant) higher in the 144A market than in the PD market, after controlling for other determinants of yield spread. Using U.S. issuers as a benchmark, we also show that issuers from other developed countries pay a similar borrowing cost, but issuers from emerging countries have to pay significantly higher costs (about additional 62 basis points in yield spread) to access the U.S. debt market. Second, the yield spread increases significantly by about 30 basis points in post-sox period: about 25 basis points increase in PD market and 45 basis points 5

6 increase in 144A market. The 2008 financial crisis also moved the debt financing costs to a historical high level. Using year 1991 as a benchmark, the additional increase in yield spread is 161 basis points in 2008, but are -19 basis points in 2007 and -38 basis points in We also show that emerging country issuers pay 50 basis points higher yield spread than the U.S. issuers in the PD market whereas other developed countries have similar borrowing costs as their U.S. issuers do. However, the U.S and international firms share similar yield spread in 144A market. We include several commonly used country-specific variables (such as legal dummies, creditor rights and rule of law) in the regression. We conduct the analysis in a sub-sample that matches each international issue with a U.S. issue based on issue type, year, industry, and issue size. We find that our main results hold in the robustness checks that include additional country-specific variables and in the matched sub-sample. Why do international firms rely less on the U.S. debt market in post-sox period and why does the yield spread increase more significantly in 144A market than in PD market? To answer these questions, we explore two alternative but not mutually exclusive explanations. First, if the passage of SOX increases the borrowing costs in the U.S., international firms may decide to seek funds elsewhere, but for most U.S. firms accessing to the domestic market maybe the only option and they have to bear higher borrowing costs after the SOX (country effect). Second, it is possible that firms with lower credit quality tend to choose the 144A market over the PD market, and thus, the higher yield spread in the 144A market could be driven by the lower credit quality of issuers (credit quality effect). To disentangle these competing explanations, we first we run separate regressions for the U.S, other developed countries, and emerging countries subsamples. We find that the borrowings costs increase significantly after the SOX only for U.S. issuers, not for other developed and 6

7 emerging country issuers. This evidence supports the prediction in the country effect explanation and suggests that U.S. issuers bear higher borrowing costs after the SOX. We further provide evidence by examining a subsample that only includes Canadian and U.S. issuers. Canadian firms highly rely on the U.S. debt market and two countries have closely linked economies, share similar institutional structures, and follow similar credit rating guidelines. Canadian firms also played a dominant role in post-sox period, accounting for more than 60% of 144A issues and 45% of PD issues by international firms. We show that Canadian issuers share similar borrowing costs as U.S. issuers, supporting the country-effect explanation. Next, we run separate regressions for the investment-grade and speculative-grade subsamples to test the credit quality explanation. We show that the yield spread between 144A and public debt issues is larger in investment-grade subsample than in the speculative-grade subsample. We also show that the increase in borrowing costs after the SOX is much larger for speculative-grade firms than for investment-grade firms (109 versus 15 basis points), and is primarily driven by the PD issues and not by the 144A issues. These results, however, do not support the credit quality explanation. The remainder of the paper is organized as follows. Section 2 describes the data and sample. Section 3 discusses the main empirical results and Section 4 provides evidence on two alternative explanations. Section 5 summarizes the conclusions. 2. Data and Sample Characteristics We use the Thomson Reuters SDC Global New issues database as our main data source for this study. We collect all non-convertible debt issued in the U.S. 144A and public markets by non-financial firms for the period of We exclude financial firms because they account for a very small proportion of issues in the 144A market. Also, their motivation and 7

8 associated costs to raise capital could be different from non-financial firms. Following Chaplinsky and Ramchand (2004) and Livingston and Zhou (2002), we further restrict the sample to fixed rate and zero coupon debt and exclude perpetuity debt. [Insert Tables 1 and 2 about here] Table 1 shows the country of origin of debt issuance in the U.S. 144A and PD market. 4 We divide the issuers into three categories: the U.S., other developed countries and emerging countries. 5 Panel A of Table 1 shows that the U.S. issuers dominate the 144A market accounting for 85% of the total amount issued in the period, followed by other developed countries (12%) and emerging countries (3%). To examine the effect of the passage of SOX, we further divide the full sample into two time periods: pre-sox and post-sox. The univariate result shows that passage of SOX may have little impact on the issuing behavior for U.S. issuers, but has significant impact for international firms. In the pre-sox period, international firms use more of the 144A market than PD market; the proportion of debt financing from the 144A market to the total proceeds of 144A and PD markets are 49% for other developed countries and 80% for emerging countries. In post-sox period, the ratio decreases to 18% and 26% for other developed countries and for emerging countries, respectively. The ratio is 33% for the U.S. issuers in the pre- and post-sox period. Firms could also time the 144A and public debt markets when the market is hot, reflected as a high cluster in issuers and issued amounts. In an untabulated analysis, we find that market timing patterns are more pronounced in the 144A market than in the public market. Years 1998, 1999, 2001, 2003 and 2004 account for about 50% 4 Crown, Erwin and McConell (1999) argue that in the past financial innovation largely focuses on the kinds of securities issued. However, more recent change focuses on the way securities are issued and where they are issued. 5 The classification of developed and emerging countries is based on World Banks country classification table ( The World Bank uses gross national income (GNI) per capita as the main criterion for classifying economies into high-, middle-, and low-income groups. The high-income countries are classified as developed countries, and the middle- and low-income countries are classified as emerging countries. 8

9 of the total 144A issuance issued in the full sample ( ). Year 2008 is a slow year for 144A issuance but is a fairly good year for public debt issuance. Barry, Mann, Mihov and Rodriguez (2009) examine a large sample of private placements, 144A, bank loans and PD issues, and show that managers attempt to time the debt market but they are not successful in lowering the interest costs. In Table 2, we further look at the 144A issues by individual countries. Canada and the U.K. are the top two most active issuing countries among non-u.s. developed countries, and Mexico and Argentina are ranked as the top two active ones among emerging countries. Except Canada, these three countries experienced a significant decline in issuing activity in the post- SOX period. The number of 144A issues for three countries in the pre-sox period is 106, 59 and 55, respectively and drops to six, one, zero, respectively, in the post-sox period. In sum, the evidence in Tables 1 and 2 suggests that the passage of SOX could change the relative attractiveness between 144A and public debt for U.S. versus international firms. The tighter regulation of SOX implies that international firms should use more of the 144A market because of its lower requirement on information disclosure. It is surprising, however, to see the significant decline in 144A issuance by international firms, suggesting that other factors might play a role in explaining this trend. [Insert Table 3 about here] Table 3 shows the distribution of 144A debt by quintiles of proceeds (panel A) and by industries (panel B). Panel A of Table 3 shows that the 144A sample exhibits significant heterogeneity in the amount of raised proceeds. About 20% of the 144A issuance in the highest quintile rank raised $559 billion, or 51% of the total proceeds. The 144A issuance by other 9

10 developed countries shows a similar pattern. In contrast, emerging country issuers exhibit lower percentage of total proceeds in quintile 5 but higher percentage in quintile 4. Panel B of Table 3 shows that the industry composition of 144A issuers differs across countries. We use Kenneth French s 12 industry classification obtained from his homepage. In the U.S. utility firms are the largest players (17% of total proceeds) whereas telecomm firms dominate in other developed countries (31%) and emerging countries (28%). In sum, the evidence suggests that we need to control for the industry effect in our regression analysis. [Insert Table 4 about here] Table 4 reports issuer and issuance characteristics. Similar to the prior studies (Livingston and Zhou, 2002; Chaplinsky and Ramchand, 2004), we document significant differences between 144A and PD issuers. A typical 144A issuer raises about $220 million of proceeds, slightly less than the amount raised by the PD issuers ($232 million). This difference could be partly explained by the economy of scale consideration because a large proportion of floatation costs that the issuer pays to investment bankers are fixed and do not vary with issue size (see e.g., Blackwell and Kidwell, 1988; Krishnaswami, Spindt and Subramaniam, 1999). The average firm size of 144A issuers (proxied by total assets before issuance), however, is about 33% of the size of PD issuers ($5.3 billion versus $16 billion). The relative issue size (dollar amount of proceeds adjusted by the total assets before issuance) is much higher for 144A issuers than for PD issuers (89.6% versus 16.3%), implying that the increase in leverage after debt issuance is more pronounced for 144A issuers. The gross underwriter spread (%), a measure for the direct financing cost, suggests that the 144A issuers have to pay about 82 basis points higher than public debt issuers. This difference could reflect the fact that underwriting institutions have to devote more resources in analyzing the quality of 144A issuers because of 10

11 the lack of public information on the firm. The yield spread, which is the key interest in our paper, is the difference between the original debt yield and the corresponding treasury yield and is the measure for the borrowing costs. The yield spread is significantly higher for 144A issuers than for public debt issuers (397 basis points versus 152 basis points). This result could be partially driven by the credit quality because the average credit rating (calculated as the average of Moody s and S&P s ratings) is substantially lower for the 144A issues. We also report proportions of sub-samples between the 144A and PD issues. First, the 144A issues are less likely to be rated by rating agencies, and among the rated 144A issues, speculative-grade firms comprise a higher proportion compared to PD issues. It is also clear that the 144A issues consist of a lower proportion of senior debt and a higher proportion of first time issuers compared to PD issues. The lower credit quality and less information availability for 144A issues suggest that their financing costs could be higher than their PD peers. We control for the issuance and issuers characteristics in our multivariate analysis. 3. Empirical results In this section, we examine whether financing costs differ between 144A and public debt issuance, and whether the costs vary after the SOX for developed and emerging country issuers. Our main regression model is as follows: YLDSPD i,t = α + β 1 144A i,t + β 2 DVP i,t + β 3 EMG i,t + β 4 SOX i,t + β 5 (144A*DVP) i,t + β 6 (144A*EMG) i,t + β 7 (144A*SOX) i,t + β 8 (Credit Ratings) i,t + β 9 log(years to Maturity) i,t + β 10 log(proceeds) i,t + β 11 Senior i,t + β 12 Public i,t + β 13 (First Time Issuer) i,t + θ t (Year dummies) t + σ t (industry dummies) t + ε i,t (1) where the dependent variable YLDSPD is treasury spread in percentage, and i and t represent i th issuance and t th year, respectively. Our main focus is on the difference between 144A and PD issues and on the cross-country and time-varying differences. We include three dummy variables for this analysis. The dummy variable 144A is equal to one for a 144A issuance and zero 11

12 otherwise. The dummy variable, DVP (or EMG) equals one if the issuer is from a developed (or emerging) country, and zero otherwise. SOX is an indicator variable for issues after July 30, 2002, and zero otherwise. Similar to the prior studies, we use several commonly used control variables in the regression. We use the S&P and Moody s ratings to measure issuer s credit risk. Following Fenn (2000) and other studies, we translate the ratings into conventional numerical scores. Specifically, Credit rating is a numerical variable that takes a value of 22 if the S&P rating is AAA or Moody s rating is Aaa. The value declines by one to 21 if the S&P rating is AA+ or Moody s rating is Aa1, and so on. The lowest value is 2 corresponding to the S&P rating of C or Moody s rating of C. If a firm is rated by both rating agencies, we take the average of the numerical value. The yield spread and credit ratings is expected to be inversely related since higher credit quality firms will have lower financial risk and consequently, lower borrowing costs relative to their low quality peers. Log (Years to Maturity) measures issue characteristics and will be expected to be positively correlated with yield spread because the farther the year to maturity, the greater the uncertainty about the firm s future prospects. Log (proceeds) is a proxy for issuer size; large issuers are likely to have lower borrowing costs because of their lower bankruptcy risk. Three additional indicator variables measure issuer or issue characteristics. The dummy variable Senior equals one for a senior debt issuance and zero for a subordinate debt issuance. Public is a dummy variable equal to one if the firm and/or its ultimate parent is a publicly traded firm, and zero otherwise. This variable proxies for the degree of publicly available information for the firm, and a public firm is expected to have more information disclosure, and therefore, a lower yield spread compared to a private firm. First Time Issuer equals one if at the time of 144A or public debt issuance the firm has never issued fixed income securities in the SDC 12

13 database since 1970, and zero otherwise. The lack of information about the firm s past history for the firm that has never issued a fixed income security in the debt market implies is likely to result in a higher yield spread for such firm compared to its peers who have accessed the debt market in the past. To control for time varying financing costs, we use year dummies instead of a time index equal to 0 in year 1991 which increases by 1 in subsequent years. For brevity we only report the coefficients in Years 2006, 2007 and 2008 in the tables. We also control for fixed industry effect by introducing dummies based on French s 12-industry classification. We use Ordinary Least Squares (OLS) (controlling for the cluster by issuers) with robust standard errors (White, 1980) as our main regression model. [Insert Table 5 about here] We report our main regression results in Table 5. In column 1, the coefficient on 144A is positive (0.48) and significant at the 1% level (t=9.90). This result suggests that 144A issuers pay 48 basis points higher yield spread than PD issuers, after controlling for other determinants of borrowing costs. As expected, the coefficient on Credit Ratings is negative and significant (- 0.34, t=40.32), indicating that as a firm s credit rating increases by one micro level (e.g. from BBB to BBB+) the yield spread decreases by 34 basis points. The coefficient on Log (Years to Maturity) and First Time Issuers are positive and significant at the 1% level, consistent with the prior studies. The positive coefficient on Senior (0.15, t=2.28) is opposite to the theoretical prediction but is consistent with Fenn (2000). The coefficients on year dummies show varying yields in different economic conditions. For example, compared to the base year of 1991, the yield is 38 (19) basis point lower in the pre-crisis years 2006 (2007) but jumps up by 161 basis points in 2008, the year of global financial crisis. The yield increase in 2008 is the largest and is 13

14 almost three times the increase in the 2001 crisis (58 basis points), indicating the severity of the crisis. In column 2, we add two additional variables, DVP and EMG, to control for the differences across developed and emerging countries. The coefficient on DVP is not significant at any level (-0.02, t=0.35) but the coefficient on EMG is positive and significant at the 1% level (0.62, t=4.85). This result suggests that issuers from emerging countries pay substantially higher borrowing costs compared to the U.S. and other developed countries. This difference could be due to the lower credit quality of firms or countries or both. In an untabulated analysis, we also include several commonly used country-specific variables to the regression in column 3. They are legal dummies (French law, German law, and Scandinavian law), creditor rights (CR) and rule of law (ROL). The legal dummies and the CR data is from Djankov, McLiesh, and Shleifer (2007). The ROL data takes the value of rule of law in year 2000 and is from Kaufmann, Kraay and Mastruzzi (2008). We find that issuers from French law country have 29 basis points lower yield spread than issuers from English law country. We also find that the coefficient on ROL is (t=1.71), suggesting that firms from higher rule of law countries enjoy lower borrowing costs. However, we do not find that the CR has any significant impact on yield spread. 6 We examine the effect of SOX in the next regression. In column 3, the coefficient on SOX is positive and significant at the 1% level (0.30, t=4.04), indicating that the SOX increased borrowing costs by 30 basis points for both 144A and PD issues. We also use three interactive variables terms, 144A*DVP, 144A*EMG, and 144A*SOX, to test whether the 144A yield spread is significantly different from the U.S. issuers for other developed countries and emerging countries, and whether it is different before and after the passage of Sarbanes-Oxley Act. In 6 The results are available upon request. The explanatory power of the regression (adjusted R-square=0.66) does not improve with additional country-specific factors compared that in column 3 and the sample size also declines slightly in this sub-sample (N=13,416). 14

15 column 4, the coefficients on 144A*DVP and 144A*EMG are not significant at any level, indicating that the country risk premium may play a less important role in the 144A market than in the PD market. The significant and positive coefficient on EMG (0.50, t=2.69) supports this argument and suggests that for PD issues, firms from emerging countries need to pay 50 basis points higher costs than U.S. issuers. We do not find a significant yield spread difference between the U.S. and other developed countries either in 144A or PD market. 7 To check the robustness of our result, we minimize the sample difference between U.S. and international issuers and ask whether the difference in yield spread still exist across countries. For this analysis, we first construct a matched sample similar to Mittoo and Zhang (2008) and King and Segal (2009). We match each international issue with a U.S. issue (without replacement) based on issue type (144A or PD), issue year, French s 12-industry, and as close as possible on issued proceeds (the difference of proceeds between the U.S. and international issue must be within 50% of boundary). The matched sample (N=1,844) is substantially smaller than the full sample (N=13,492). We conduct the same regressions as in columns 3-4 of Table 5 and report the new results in columns 5-6. We find that our main results are robust to this analysis and hold in the matched sample. For example, in column 5 the economical and statistical significance of coefficients on SOX, DVP and EMG are very similar between the full and matched samples. In column 6, the coefficient on 144A*SOX is still positive and significant, and the coefficients on 144A*DVP and 144A*EMG are not significant at any level, similar to that in full sample. In sum, we show that our main regression results hold in the matched sample. 7 The significant impact of SOX on borrowing costs may be driven by the effect of financial crisis. We also run regressions excluding year 2008, and find that our main results hold. 15

16 4. Higher costs in post-sox period for 144A issues? Country or credit quality effect In the above analysis, we document that yield spreads increase significantly in post-sox period and this increase is more pronounced in the 144A market than in PD market. There are two alternative but not mutually exclusive explanations for this finding. The first possibility is associated with whether firms will choose other markets for financing instead of the U.S. debt market after the SOX (country effect). If the passage of SOX significantly increases borrowing costs, U.S. and international firms face different options: for most U.S. firms, accessing the home market seems the most viable choice because the U.S. still has the largest and most liquid debt market in the world. For international firms, they will tradeoff between the benefits and costs of using U.S. debt market, and may decide to use other markets (e.g., Eurobond) if the costs increase substantially for them after the SOX. If the country effect explanation stands, we expect to observe an increased yield spread after the SOX for U.S. firms than for international firms. The second possibility is the credit quality effect. Prior literature suggests that firms with lower credit quality tend to choose 144A market over the PD market, and thus, the higher borrowing costs in 144A market could be driven by the lower credit quality of issuers. In this section, we examine these explanations with different subsample analysis Country effect To examine the country effect, we run separate regressions for the U.S, other developed countries and emerging countries subsamples. This analysis allows us to examine the evolving financing costs of 144A against PD issues after controlling for the macro category of country effects. The results are reported in columns 1-3 of Table 6. In column 1, we find that the coefficients on 144A, SOX and 144A*SOX are all positive and significant (144A=0.41, t=6.90; SOX=0.25, t=3.32; 144A*SOX=0.21, t=2.53). This result suggests that for the U.S. issuers the 16

17 financing costs significantly increase after the SOX, and this increase is more pronounced for 144A than for PD issues. For both the developed and emerging countries subsamples, we find that SOX has no significant impact on yield spread; the coefficients on SOX and 144A*SOX are not significant at any level (columns 2 and 3). This evidence provides direct support to the country effect explanation. For U.S. issuers, even though the SOX substantially increased their yield spreads, financing from foreign markets seems not a viable solution and they will pay additional costs to finance through the domestic market. For international issuers, they have a choice to finance in the U.S. or elsewhere: the international issuers who expect a large increase in yield spread after the SOX choose to finance elsewhere whereas those who expect no significant increase in borrowing costs choose to stay in the U.S. market. The insignificant coefficients on SOX and 144A*SOX in columns 2-3 support this argument. Furthermore, our univariate evidence in Table 1 also suggests that the reduction of using the U.S. debt market after the SOX is pervasive among international firms: the number of 144A issuance was 495 for non- U.S. developed countries and 241 for emerging countries in pre-sox period, and decreased to 109 and 13 in post-sox period, respectively. Similar pattern is also documented for the public debt issues. [Insert Table 6 about here] Another angle to examine the country effect explanation is to look at the Canadian issuers. In post-sox period, we find that there are 109 Rule 144A issues and 193 PD issues by non-u.s. developed countries (Table 1). Among them, 64% (N=70) of 144A issues and 48% (N=92) of PD issues are by Canadian firms. In other words, Canadian firms played a dominant role among international issuers in accessing the U.S. debt market in post-sox period. Why do Canadian firms still finance through the U.S. market in post-sox period? Compared to other foreign 17

18 countries, Canadian and U.S. capital markets are more closely linked, share similar institutional structures, and follow similar credit rating guidelines. The 1991 Canada-U.S. Multijurisdictional Disclosure System (MJDS) allows eligible Canadian issuers to make cross-border security offerings in the U.S. using mostly Canadian disclosure documents. Freedman and Engert (2003) find that high-yield Canadian borrowers meet almost all of their financing requirements in the deep and liquid U.S. high-yield bond market. In sum, it is possible that the U.S. market is their primary consideration for Canadian firms, similar to their U.S. peers. We run the regression for the subsample that includes only the U.S. and Canadian issuers. In column 4 of Table 6, we find that the coefficient on CAN (a dummy variable equals to one for Canadian issuers, and zero for U.S. issuers) is not significant at any level, indicating that Canadian issuers pay similar borrowing costs as their U.S. peers. We also show that the coefficient on SOX is 0.32 (t=4.22), similar to the results in the main regression (column 3 of Table 5, SOX=0.30, t=4.04). In column 5, we also find that the coefficients on CAN and 144A*CAN are not significant, suggesting that the borrowing costs are similar between the two countries either in 144A or PD market. The coefficient on 144A*SOX is positive and significant (0.20, t=2.44) and is similar to the finding for the U.S. issuers subsample (column 1). In sum, we find that Canadian issuers share similar borrowing costs as U.S. issuers, supporting the country effect explanation. 4.2 Credit quality effect Literature suggests that the issuers credit quality determines the source of debt (see e.g., Denis and Mihov, 2003; Hale and Santoles, 2008). Fenn (2000) argues that the U.S. domestic speculative-grade issuers favor the 144A market because it allows firms to raise funds more quickly than in the public debt market. Chaplinsky and Ramchand (2004) suggest that 18

19 international issuers from emerging markets are more likely to issue in the 144A market because they can avoid complying with the more stringent U.S. accounting and disclosure requirements. Mittoo and Zhang (2009) also suggest that high and low credit quality firms face different tradeoffs between financial flexibility and concerns about maintaining credit rating and therefore might react differently to the bond market access. They show that Canadian high quality firms have the option to issue debt in the Canadian or U.S. market, and they tend to go to the U.S. primarily for large issues. By contrast, the LQ firms have no choice but to access the U.S. highyield market to enhance their financial flexibility and for economy of scale consideration. [Insert Table 7 about here] In sum, if low credit quality firms prefer the 144A market over the public debt market, naturally firms with a high risk profile and large asymmetric information will consist of a higher proportion in the 144A market, which drives up the average yield spread in this market (credit quality effect). To examine this effect, we run separate regressions for issues with investmentgrade and speculative-grade ratings and report the results in Table 7. We draw two main conclusions from this analysis. First, the coefficient on 144A is positive and significant in investment-grade subsample (0.24, t=5.94), about 5 basis points higher than that in speculativegrade subsample (0.19, t=2.51). This evidence does not support the explanation of credit quality effect because the credit spread is expected to be wider between 144A and PD in the speculativegrade subsample than the investment-grade subsample. Second, the coefficient on SOX in the speculative-grade sample is substantially higher than in the investment-grade sample (1.09 versus 0.15), implying that the increased yield spread in post-sox period is mainly driven by the speculative-grade firms. In columns 2 and 4, however, we find that the economic and statistical significance on SOX remain the same as columns 1 and 3 but the coefficients on 144A*SOX are 19

20 not significant in either investment-grade or speculative-grade subsamples, indicating that credit quality cannot explain the higher yield for low quality firms in post-sox period. Finally, we find that the sensitivities of issue characteristics to yield spread are significantly different between investment-grade and speculative-grade firms. 5. Conclusions We use a large sample of 144A and public debt issuance by U.S. and international firms to examine whether the borrowing costs, proxied by yield spread, vary across countries and time for the period of We show that the yield spread is 48 basis points higher in the 144A market than in the PD market, after controlling for other determinants of yield spread. We also show that in the PD market, other developed countries share similar borrowing costs as the U.S. issuers whereas emerging countries issuers pay 50 basis points higher yield spread than the U.S. issuers. In the 144A market, we find that firms from the U.S., other developed countries and emerging countries share similar yield spread. We further show the impact of the passage of Sarbanes-Oxley Act and shows that the yield spread increases significantly by about 30 basis points in the post-sox period: about 25 basis points increase in PD market and 45 basis points increase in 144A market. The 2008 financial crisis also moves the debt financing costs to a historical high level. We also examine two competing explanations on why higher yield spread after the SOX for 144A issues, and show that the country effect explanation is more dominant than the credit quality effect. Our study complements the findings in several recent studies. Qi, Roth, and Wald (2010) examine the impact of country-level political rights on borrowing costs for corporate bonds from 39 countries issued in the Eurobond and Yankee bond markets. They show that after controlling for legal institutions, higher political rights are positively associated with credit ratings and 20

21 negatively associated with yield spread. Our results that issuers from emerging countries pay significantly higher borrowing costs than issuers from developed countries support their findings. Miller and Puthenpurackal (2002) show that firms will benefit from tapping the global capital markets by accessing an expanded base of international investors. They examine 230 global bond issues (bonds sold simultaneously in several countries at the same price) in the U.S. and show that borrowing and issuing costs of global bonds are significantly lower than similar domestic bonds and Eurobonds. 21

22 Table 1. Issuance distribution in the U.S. Rule 144A and public debt markets Table 1 shows the distribution of debt issuance in the U.S. Rule 144A and the public debt market by the U.S and international issuers. We collect all non-convertible debt (including fixed rate and zero coupon debt but excluding perpetuity debt) by non-financial firms from the Thomson Routers SDC Global New issues database for the period of We divide issuers into three categories: the U.S., other developed countries and emerging countries. The categorization of developed and emerging countries is based on the World Banks country classification table. All 144A and public debt issues are further classified into two sub-samples (pre- and post-sox period) if the issue date is before (after) July 30, 2002 when the U.S. Congress passed the Sarbanes-Oxley Act (SOX). Panel A: Rule 144A debt issuance Full sample: Pre-Sox Post-Sox Countries of Issuers N Mean Median All All Proceeds All Proceeds Proceeds Proceeds N N Proceeds 1) The U.S. 5,032 1,102, , ,427 1, ,163 2) Other developed countries , , ,468 3) Emerging countries , , ,389 Panel B: Public debt issuance Full sample: Pre-Sox Post-Sox Countries of Issuers N Mean Median All All Proceeds All Proceeds Proceeds Proceeds N N Proceeds 1) The U.S. 10,324 2,252, ,147 1,365,938 2, ,304 2) Other developed countries , , ,893 3) Emerging countries 71 16, , ,235 22

23 Table 2. Debt issuance in the U.S. Rule 144A and public markets by top five most active international countries Table 2 shows the distribution of debt issuance in the U.S. 144A and the public debt market by the U.S. and international issuers. We collect all non-convertible debt (including fixed rate and zero coupon debt but excluding perpetuity debt) by non-financial firms from the Thomson Routers SDC Global New issues database for the period of The categorization of developed and emerging countries is based on the World Banks country classification table. All 144A and public debt issues are further classified into two sub-samples (pre- and post-sox period) if the issue date is before (after) July 30, 2002 when the U.S. Congress passed the Sarbanes-Oxley Act (SOX). Among other developed countries and emerging countries, we report the top five most active debt issuance countries measured by all proceeds during the period. Panel A: Rule 144A debt issuance Countries of Issuers Full sample: Pre-Sox Post-Sox N All Proceeds N All Proceeds N All Proceeds Top five issuing countries among other developed countries ranked by all proceeds 1) Canada , , ,445 2) United Kingdom , , ,148 3) Cayman Islands 25 15, , ) France 17 15, , ) Netherlands 31 9, , Top five issuing countries among emerging countries ranked by all proceeds 1) Mexico 60 11, , ) Argentina 55 6, ,805 3) Brazil 64 6, , ) Malaysia 9 4, ,781 5) Chile 15 3, , Panel B: Public debt issuance Countries of Issuers Full sample: Pre-Sox Post-Sox N All Proceeds N All Proceeds N All Proceeds Top five issuing countries among other developed countries ranked by all proceeds 1) Canada , , ,873 2) United Kingdom , , ,583 3) Netherlands 51 44, , ,694 4) Spain 9 6, ,710 5) France 12 6, , ,783 Top five issuing countries among emerging countries ranked by all proceeds 1) Mexico 18 5, , ,841 2) Chile 18 3, ,659 3) Argentina 15 3, ,013 4) Philippines 11 2, ,091 5) Brazil 3 1, ,394 23

24 Table 3. The distribution of Rule 144A and public debt issuance by proceeds quintile and industry Table 3 shows the distribution of debt issuance in the U.S. 144A and the public debt market by proceeds quintile and industry. We collect all non-convertible debt (including fixed rate and zero coupon debt but excluding perpetuity debt) by non-financial firms from the Thomson Routers SDC Global New issues database for the period of We classify issuers into 12 industries based on Kenneth French s criterion. Panel A: The distribution of 144A debt issuance by proceeds quintile U.S. issuers Other developed countries Emerging countries Quintile rank of 144A proceeds Mean proceeds quintile total proceeds % of total proceeds Mean proceeds quintile total proceeds % of total proceeds Mean proceeds quintile total proceeds % of total proceeds ,654 2% 27 3,194 2% 29 2,137 5% ,338 8% 96 12,208 8% 93 5,371 13% ,346 15% ,225 13% 153 7,210 17% ,409 24% ,731 18% ,603 26% ,843 51% ,447 58% ,919 39% Total 1,102, % 150, % 41, % Panel B: The distribution of 144A debt issuance across industries Industry No. of 144As U.S. issuers Other developed countries Emerging countries Industry total proceeds industry % of total proceeds No. of 144As Industry total proceeds industry % of total proceeds No. of 144As Industry total proceeds industry % of total proceeds Utilities ,523 17% ,107 11% 46 8,773 15% Others ,851 16% ,741 13% 49 10,184 18% Manufacturing ,024 9% ,934 10% 53 7,785 14% Telecomm ,253 15% ,350 31% 69 13,690 24% Wholesale and Retail ,871 9% 63 19,389 4% 16 1,976 3% Consumer NonDurables ,975 8% 90 17,615 4% 24 2,301 4% Energy ,945 6% ,498 16% 49 10,353 18% Business Equipment ,137 6% 37 9,567 2% % Consumer Durables ,918 4% 31 10,698 2% % Chemicals ,210 4% 36 8,439 2% % Healthcare ,129 5% 30 21,866 5% 24

25 Table 4. Issuer and issuance characteristics of 144A and public debt issues Table 4 compares the issuer and issuance characteristics of 144A and public issues. Issue size (%) is the proceeds over total assets before issuance. Gross underwriter spread (%) is the ratio of total management fees to the proceeds. Yield spread is the difference between the issues yield-to-maturity and the yield of the closest maturity treasury. Credit ratings are the average of S&P and Moody s ratings in the numerical value. 144A issuance Public issuance Difference test Characteristics Mean Median Mean Median t-test Wilcoxon rank test Proceeds ($mil.) ** *** Total assets before issuance ($mil.) *** *** Issue size (%) 89.6% 13.1% 16.3% 2.5% *** *** Gross underwriter spread (%) *** *** Original yield (%) *** *** Yield spread *** *** Average credit ratings *** *** Years to maturity *** *** Proportions Proportions test Proportions with rated debt 77.8% 99.3% *** Proportions with junk rated debt 53.4% 14.1% *** Proportions with senior debt 78.7% 95.9% *** Proportions with public firms 69.1% 95.5% *** Proportions with first time issuers 52.9% 8.7% *** *significant at 10%; **significant at 5%; ***significant at 1% 25

26 Table 5. Main regression analysis and evidence on the matched sample Table 5 reports the regression results of yield spread on issuance characteristics. Columns 1-4 report the full sample results. Columns 5-6 report the results based on a matched sample that matches each international issuance with a U.S. issuance on issue type, year, industry and as close as possible in proceeds. The dependent variable is yield spread. 144A equals one for a 144A issuance and zero otherwise. DVP (EMG) equals one for the issuers from non-u.s. developed country (emerging country), and zero otherwise. SOX takes the value of one if the issuance is offered after July 30, 2002, and zero otherwise. 144A*DVP, 144A*EMG, and 144A*SOX are interactive variables. Credit ratings is a numerical variable that takes the average of the S&P and Moody s ratings. Log (Years to Maturity) and Log (Proceeds) are the natural log of the number of years to final maturity and proceeds amount (in U.S. $million). Senior equals one for a senior debt issuance and zero for a subordinate debt issuance. Public equals one if the firm and/or its ultimate parent is a publicly traded firm, and zero otherwise. First Time Issuer equals one if the firm has never issued public fixed income securities in the SDC database at the time of 144A or public debt issuance. We use year dummies (Y1992 to Y2008) and Kenneth French s 12 industry dummies to control the fixed year and industry effects. For brevity we only report the coefficients on Years 2006, 2007 and 2008 in the tables. We use Ordinary Least Squares (OLS) with robust standard errors (White, 1980) and cluster in firms as our main regression model. Full Sample Matched subsample (1) (2) (3) (4) (5) (6) Constant 6.89*** 6.90*** 6.89*** 6.87*** 7.95*** 7.87*** (34.27) (34.07) (34.10) (34.11) (23.23) (22.48) 144A 0.48*** 0.48*** 0.48*** 0.40*** 0.48*** 0.38** (9.90) (9.87) (9.88) (7.01) (4.41) (2.36) DVP (0.35) (0.34) (0.05) (0.31) (0.42) EMG 0.62*** 0.61*** 0.50*** 0.59*** 0.46** (4.85) (4.83) (2.69) (4.00) (2.24) SOX 0.30*** 0.25*** (4.04) (3.34) (0.73) (0.22) 144A*DVP (0.36) (0.16) 144A*EMG (1.02) (1.01) 144A*SOX 0.20** 0.38* (2.51) (1.80) Credit Ratings -0.34*** -0.34*** -0.34*** -0.34*** -0.35*** -0.35*** (40.32) (40.33) (40.40) (40.53) (26.00) (25.65) Log(Years to Maturity) 0.17*** 0.17*** 0.17*** 0.17*** (7.84) (7.96) (7.85) (7.98) (1.53) (1.51) Log (Proceeds) -0.05** -0.05** -0.05** -0.05** -0.21*** -0.20*** (2.14) (2.15) (2.09) (1.99) (4.72) (4.45) Senior 0.15** 0.13** 0.13** 0.12* 0.21* 0.21 (2.28) (2.07) (2.05) (1.89) (1.65) (1.59) Public -0.38*** -0.37*** -0.37*** -0.36*** -0.35*** -0.33*** (4.52) (4.43) (4.42) (4.41) (3.03) (2.90) First Time Issuer 0.41*** 0.40*** 0.40*** 0.41*** 0.25*** 0.25*** (8.52) (8.19) (8.24) (8.31) (3.22) (3.27) Y *** -0.38*** -0.68*** -0.66*** (5.68) (5.64) (6.55) (6.47) (1.28) (0.91) Y *** -0.19*** -0.49*** -0.48*** (2.65) (2.65) (4.62) (4.53) (0.83) (0.46) Y *** 1.61*** 1.31*** 1.35*** 1.28*** 1.41*** (17.17) (17.22) (10.69) (11.09) (2.86) (3.16) N Adj. R-square *significant at 10%; **significant at 5%; ***significant at 1% 26