Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Cross-listed Bonds and Ratings Conservatism
Yigit Atilgan
Sabanci University
Aloke (Al) Ghosh#
Baruch College
Jieying Zhan...
Cross-listed Bonds and Ratings Conservatism
ABSTRACT
We investigate whether cross-listed bonds are rated more conservative...
1
Cross-listed Bonds and Ratings Conservatism
1 Introduction
While the theory and evidence on conservatism in financial re...
2
favorable ratings than when reporting unfavorable ratings for bonds with higher information
asymmetry, which suggests th...
3
Using a sample of public debt issued in the U.S. by foreign firms between 1990 and
2009, we find evidence that on averag...
4
While our results are consistent with ratings conservatism, they are also consistent with
an alternative explanation tha...
5
conservative, we expect a correction for the conservative bias, i.e., a lower spread for foreign
bonds compared to domes...
6
and accuracy (Cheng and Neamtiu 2009), our paper extends the literature by highlighting a less
frequently examined yet i...
7
In stark contrast to the abundant research on conservatism in financial reporting,
relatively few studies examine whethe...
8
In this paper, we explore an alternative source of rating conservatism, that is, high
information asymmetry.9
The mainta...
9
al. 2007; Edwards et al. 2007). One innovation in our study is that we test our hypothesis by
identifying a group of bon...
10
trading law enforcement. Although cross-listing may improve a foreign firm’s information
environment, Fernandes and Fer...
11
bonds to vary with rating agencies’ reputational cost. While rating agencies have higher
reputation cost from underesti...
12
reflect rating agencies’ private information. When rating agencies have limited public
information, they are likely to ...
13
We also use the bond market to differentiate between rating conservatism and private
information explanation. Prior lit...
14
3.1 Data description
We obtain our sample bonds from the Global New Issues Database of the Securities Data
Corporation ...
15
Table 2 presents the frequency distribution, total public debt offering, and the average
issue size of debt raised by t...
16
where:
Rating = one for firms with the best credit rating (AAA) and the value increase by
one for successively worse ra...
17
expropriation, as defined in La Porta et al. (2006).
Our variable of interest is Non-US. A positive coefficient indicat...
18
(1995) find that there are large variations in the approaches taken by bankruptcy courts from
different countries and t...
19
larger, less levered and more profitable firms, and firms with higher interest coverage. As
expected, we find that rati...
20
upgrades, indicating that cross-listed bonds receive 28% fewer upgrades. We also find a
significantly positive coeffici...
21
investment-grade bonds. Because rating agencies have higher reputation costs from failing to
predict the default of an ...
22
3.5.1 The probability of missing defaults and raising false alarms
As in Cheng and Neamtiu (2009), we define an indicat...
23
investment-grade cross-listed bond compared to a similarly rated U.S. domestic bond.
Concurrently, the significantly po...
24
bond market perceiving the rating bias as an outcome of rating conservatism. In contrast, an
insignificant β1 indicates...
25
benchmark on Non-US after controlling for ratings and issue, issuer and country characteristics.20
We find that for the...
26
covenants contained in the bond contracts. We find that 4.00% of the non-U.S. bonds in our
sample have the rating trigg...
27
4.3 Upgrade and downgrade analysis
In the reported results in Tables 5 to 7, we focus on the three years after issuance...
28
unchanged when we use Offer yield as another measure of the cost of debt.21
Second, we also
examine whether the inclusi...
29
for U.S. domestic bonds. Collectively, the results are consistent with the rating conservatism
explanation and inconsis...
30
REFERENCES
Ahmed, A., B. Billings, R. Morton, & M. Stanford-Harris (2002). The role of accounting conservatism in
mitig...
31
Diether K.B, C.J. Malloy, & A. Scherbina (2002). Differences of opinion and the cross-section of stock
returns. Journal...
32
La Porta, R., F. Lopez-de-Silanes, & A. Shleifer (2006). What works in securities laws. Journal of
Finance 61, 1-32.
La...
33
Table 1
Number, origin and size of cross-listed bonds over time
$ Million $ Million
Year Number of issues Number of cou...
34
Table 2
Cross-listed bonds grouped by country of origin
Country Number of issues Total debt issued Average issue size
A...
35
Table 3
Descriptive statistics for cross-listed bonds and U.S. domestic bonds
Means Medians
Non-U.S. U.S. p-value Non-U...
36
Table 4
Initial rating of cross-listed bonds compared to U.S domestic bonds
Full sample
Independent Variables Coefficie...
37
Table 5
Likelihood of subsequent rating upgrades and downgrades
Upgrade Downgrade
Independent Variables Coefficient (p-...
38
Table 6
Frequency of subsequent upgrades and downgrades
Upgrade Downgrade
Independent Variables Coefficient (p-value) C...
39
Table 7
Time interval before subsequent upgrades and downgrades
Upgrade Downgrade
Independent Variables Coefficient (p-...
40
Table 8
Rating variation by investment and non-investment grade
Investment grade Non-investment grade
Independent Varia...
41
Table 9
Likelihood of missing defaults and raising false alarms
Missed default False alarm
Independent Variables Coeffi...
Ghosh Presentation - European Finance Association Conference, Copenhagen, Denmark
Upcoming SlideShare
Loading in …5
×

Ghosh Presentation - European Finance Association Conference, Copenhagen, Denmark

426 views

Published on

Professor Al Ghosh presented at the European Finance Association Conference, Copenhagen, Denmark, in 2012.

Published in: Economy & Finance
  • Be the first to comment

  • Be the first to like this

Ghosh Presentation - European Finance Association Conference, Copenhagen, Denmark

  1. 1. Cross-listed Bonds and Ratings Conservatism Yigit Atilgan Sabanci University Aloke (Al) Ghosh# Baruch College Jieying Zhang# University of Southern California June 2012 # Corresponding Authors (Aloke.Ghosh@baruch.cuny.edu, Jieying@marshall.usc.edu ) Acknowledgements: Our paper has benefited from the comments of Jay Dahya, Ozgur Demirtas, Mingyi Hung, Bill Rees, Kishore Tandon, Joe Weintrop, and participants at the 6th Accounting Symposium of the Accounting Research Network Netherlands (Leuven, Belgium) and, in particular, the discussant, Carolina Salva. We are particularly grateful to Paquita Davis-Friday for her comments and suggestions while the paper was initially being developed.
  2. 2. Cross-listed Bonds and Ratings Conservatism ABSTRACT We investigate whether cross-listed bonds are rated more conservatively than U.S. domestic bonds. We argue that because of the high information asymmetry of cross-listed bonds, investors are more reliant on ratings, which in turn increases rating agencies’ exposure to reputation losses when foreign issuers default but ratings indicate otherwise. We expect this heightened reputation concerns to motivate rating agencies to be more conservative when rating cross-listed bonds than when rating domestic bonds. Consistent with our expectations, we find that cross-listed bonds have lower ratings at issuance and subsequently are less likely (and take longer) to be upgraded than comparable U.S. domestic bonds. We also find that rating conservatism is more pronounced for investment-grade cross-listed bonds, consistent with the higher reputational concerns for investment-grade bonds. Finally we examine a competing explanation that rating agencies’ private information indicates higher default risk for cross-listed bonds. We find that the lower ratings of cross-listed bonds are more likely to raise false alarms, less likely to miss future defaults, and are also corrected by a lower spread at issuance. Collectively, the evidence suggests that rating conservatism is a more plausible explanation for our finding than the private information explanation. Keywords: cross-listed bonds; credit ratings; rating conservatism; cost of debt JEL classification: M41; G29; G38 Data Availability: The data used in this study are publicly available from the sources identified in the text.
  3. 3. 1 Cross-listed Bonds and Ratings Conservatism 1 Introduction While the theory and evidence on conservatism in financial reporting is extensive (Watts (2003 a&b), the application of the conservatism principle in other areas such as credit ratings has been limited. It is important to study rating conservatism because debtholders have asymmetric payoff function and thus conservatism can be a desirable rating property benefiting debtholders. However, most existing studies on credit ratings focus on the timeliness and accuracy of ratings (e.g., Cheng and Neamtiu 2009). One exception is Beaver, Shakespeare and Soliman (2006), which relies on the debt contracting role of certified credit ratings to explain rating conservatism. In this paper, we propose information asymmetry as another explanation for rating conservatism. Because information asymmetry is likely to be higher for cross-listed bonds than for U.S. bonds, we investigate whether cross-listed bonds are rated more conservatively. The maintained assumption in our study is that rating agencies tend to have an asymmetric loss function, because their reputation costs are higher when an issuer defaults but ratings indicate otherwise than when an issuer has a higher credit quality relative to its ratings (Beaver et al. 2006). Consequently, rating agencies are likely to be conservative, i.e., they would require a higher verification standard to report more favorable ratings than unfavorable ratings. We expect that rating agencies are more conservative when information asymmetry is high for the following reason. When information asymmetry is high, investors are more likely to rely on ratings to assess the default risk of a bond issue for investment and contracting decisions (Ball et al. 2008). Greater reliance on ratings for investment and contracting reasons increases rating agencies’ exposure to reputational costs especially when an issuer defaults but ratings indicate otherwise. Thus rating agencies are more likely to impose a higher standard when reporting
  4. 4. 2 favorable ratings than when reporting unfavorable ratings for bonds with higher information asymmetry, which suggests that ratings are more conservative when information asymmetry is high. Cross-listed bonds provide a unique setting to test our hypothesis that ratings are more conservative when information asymmetry is high. Prior studies suggest that information asymmetry is higher for cross-listed firms than for domestic firms because of several reasons. First, the quality and quantity of public information is lower for cross-listed firms compared to that of U.S. domestic firms (Leuz et al. 2003; Bradshaw et al. 2004; DeFond et al., 2006). For instance, Leuz et al. (2003) find more pervasive earnings management in countries with weak investor protection. DeFond et al. (2006) also show that earnings are less informative in countries with poor accounting quality and weak insider trading law enforcements. Second, although cross-listed bonds are subject to SEC’s jurisdiction, their public reports are still influenced by home regulatory environment and managerial discretion. Consistent with the premise that SEC’s oversight does not entirely overcome the effect of local market, Lang et al. (2006) find that cross-listed reconciled earnings are subject to greater earnings management than U.S. earnings. Third, the private information collection of rating agencies is unlikely to fully offset the higher information asymmetry of cross-listed bond issuers. For example, even though analysts also have incentives to engage in private information collection, we find that analysts forecast dispersion of cross-listed bond issuers is higher than that of domestic issuers. Overall, since cross-listed bonds are likely to have higher information asymmetry, which amplifies rating agency’s asymmetric loss function, we hypothesize that cross-listed bonds are rated more conservatively than U.S. domestic bonds.1 1 It is also possible that the private information collection of rating agencies may mitigate information asymmetry, which works against finding our hypothesized relation. We discuss this possible alternative in detail in Section 2.
  5. 5. 3 Using a sample of public debt issued in the U.S. by foreign firms between 1990 and 2009, we find evidence that on average cross-listed bonds have significantly lower ratings at issuance compared to U.S. bonds with similar issuer and issue characteristics. Specifically, in a regression of bond ratings on a cross-listed bond indicator variable and various issue, issuer, country-level control variables, the coefficient on the cross-listed bond indicator variable is 1.25 and statistically significant at less than 1% level. Thus, cross-listed bond ratings are more than one notch lower than comparable U.S. domestic bond ratings. Our results are consistent with the hypothesis that cross-listed bonds are rated more conservatively than domestic bonds. It is possible that the initial conservative rating for cross-listed bonds is only temporary and that the bias reverses as more information becomes available once foreign registrants meet the SEC filing requirements. Therefore, we investigate the likelihood, frequency, and timing of subsequent rating changes. We find that following the initial rating assignment, cross-listed bonds are less likely to be upgraded, are likely to receive fewer upgrades, and take longer to be upgraded. These results suggest that the initial conservative ratings of cross-listed bonds do not reverse in subsequent periods. A key argument in our paper is that ratings are more conservative when reputation costs are higher. While rating agencies have higher reputation cost from underestimating than overestimating default risk (Beaver et al. 2006), the reputation cost of underestimating the default risk is substantially higher for investment-grade bonds than for speculative-grade. Thus, we expect that the investment-grade cross-listed bonds are rated more conservatively compared to non-investment-grade cross-listed bonds. Consistent with our expectation, we find that rating conservatism for cross-listed bonds is more pronounced for investment-grade bonds.
  6. 6. 4 While our results are consistent with ratings conservatism, they are also consistent with an alternative explanation that rating agencies possess private information about the default risk of cross-listed bonds and that lower ratings accurately reflect the additional risk of cross-listed bonds (Ederington et al. 1987; Reiter and Zeibart 1991). We call this the “private information” explanation. 2 While our cross-sectional analysis with respect to investment-grade bond is inconsistent with the “private information” explanation, we conduct two additional tests to further differentiate between the rating conservatism and private information explanation. Our first test directly compares the probability of ratings failing to predict a default, and the probability of ratings raising a false alarm about a future default across cross-listed bonds and U.S. domestic bonds. If the lower rating is due to rating conservatism, cross-listed bond ratings are expected to have a higher probability of raising false alarms (Type I error) and a lower probability of failing to predict future defaults (Type II error). In contrast, if the lower rating of cross-listed bonds accurately reflects higher default risk, we do not expect the probability of raising false alarms and failing to predict future defaults to vary between cross-listed and U.S. domestic bonds. Our results indicate that ratings of cross-listed bonds are more likely to raise false alarms and are less likely to fail to predict a future default than ratings of domestic bonds, providing direct support for the conservatism explanation. Our second test compares the spreads to a benchmark (defined as the yield to maturity less a U.S. treasury yield with similar maturity) on the issuance day between cross-listed bonds and U.S domestic bonds.3 If the bond market perceives ratings of cross-listed bonds as being 2 In a 2002 Moody’s global credit research report entitled “Understanding Moody’s Corporate Bond Ratings and Rating Process”, Moody’s clearly points out that it uses confidential non-public information that issuers provide to Moody’s only for the purpose of assigning ratings (p.5). 3 The underlying assumption of this test is that bond market is efficient, which is a reasonable assumption based on prior literature (Hotchkiss and Ronen 2002; Covitz and Harrison 2003).
  7. 7. 5 conservative, we expect a correction for the conservative bias, i.e., a lower spread for foreign bonds compared to domestic bonds with comparable ratings. On the other hand, if the bond market perceives that rating agencies’ private information captures the additional risk of a foreign issuer, we do not expect the spread to vary between cross-listed bonds and domestic bonds with comparable ratings. We find that the spread for cross-listed bonds is on average 37 basis points lower for cross-listed bonds than for U.S. domestic bonds after controlling for ratings and various issue, issuer, and country characteristics.4 We also find that the price correction is concentrated in investment-grade bonds, which is consistent with rating conservatism being more pronounced among investment-grade bonds. Collectively, our second test also suggests that the lower ratings associated with cross-listed bonds is consistent with rating conservatism explanation instead of private information explanation. Our study makes three contributions. First, our study advances the rating literature by proposing information asymmetry as another explanation for rating conservatism, while Beaver et al. (2006) provide initial evidence that contracting use of ratings explains rating conservatism. Specifically, we provide evidence demonstrating that rating agencies tend to be more conservative when information asymmetry is higher for cross-listed bonds.5 Second, our findings also add to the small yet growing literature on the application of conservatism principle outside financial reporting (Lu and Sapra 2009; Hugon and Muslu 2010). It is important to understand rating conservatism because conservatism can be a desirable rating property benefiting debtholders. While existing rating studies largely focus on rating attributes such as timeliness 4 This result partially explains why foreign bond issuers in the U.S. may not object to rating agencies imposing a higher standard for cross-listed bonds because foreign firms do not appear to bear added borrowing costs. 5 We extend Beaver et al. (2006) in two additional aspects. Beaver et al. (2006) focus on the timeliness of rating downgrades, and we study both initial rating assignments and subsequent rating revisions. Also, by differentiating between conservatism and private information explanations, we rule out one potential alternative explanation for rating conservatism that could also be applicable for the findings in Beaver et al. (2006).
  8. 8. 6 and accuracy (Cheng and Neamtiu 2009), our paper extends the literature by highlighting a less frequently examined yet intuitive rating property. Third, our paper adds to the very limited research on foreign firms issuing debt in the U.S. by documenting the rating properties of cross- listed bonds and their pricing implications.6 The remainder of the paper is organized as follows. Section 2 motivates the paper and develops the hypotheses. Section 3 describes the data and research design and presents the empirical results. Section 4 presents additional analyses and robustness tests. Section 5 concludes the paper. 2 Motivation and hypotheses 2.1 Rating conservatism and information asymmetry The literature on the accounting conservatism suggests five explanations for conservatism in financial reporting ― contracting, litigation, tax, regulation, and information asymmetry. Empirical evidence is largely consistent with these explanations (e.g., Ball et al. 2000; Holthausen and Watts 2001; Ahmed et al. 2002; LaFond and Watts 2008; Zhang 2008; Nikoleav 2010).7 In addition to financial reporting conservatism, researchers find that auditors and analysts are also conservative when they issue audit opinions and earnings forecasts, respectively (e.g., Lu and Sapra 2009; Hugon and Muslu 2010). 6 Additionally, examining cross-listed bonds is important in itself because relatively little attention is paid to understanding the properties of cross-listed bonds, while the evidence regarding cross-listed equity is extensive (Karolyi 2006). Understanding the rating properties of foreign bonds is especially important because foreign firms raise significantly more debt than equity in the U.S. For example, Chaplinsky and Ramchand (2004) report that while foreign firms are allowed to issue either debt or equity, the total amount of capital raised by debt is nearly eight times the amount of raised by equity. 7 While there are various definitions of accounting conservatism, the definition proposed by Basu (1997), i.e., “a higher degree of verification for recognizing good news than bad news in financial statements,” has gained popularity in the past decade.
  9. 9. 7 In stark contrast to the abundant research on conservatism in financial reporting, relatively few studies examine whether rating agencies are conservative when rating debt securities. Studying conservatism in credit ratings is important for several reasons. First, conservatism can be a desirable rating property because the users of ratings, i.e., debtholders, have an asymmetric payoff function. Even though debt contracting is the most frequently offered explanation for conservative accounting (Watts 2003 a&b), and rating agencies have an asymmetric payoff function similar to debtholders, existing rating studies largely focus on rating attributes such as timeliness and accuracy (Cheng and Neamtiu 2009) while overlooking rating conservatism. Second, contrary to the popular perception about a decline in the quality of credit ratings, academic studies tend to find that rating agencies have raised their rating standards over time (Cheng and Neamtiu 2009).8 Studying rating conservatism provides a better understanding of the complex incentive structure of rating agencies and how those incentives affect the properties of credit ratings. To the best of our knowledge, Beaver et al. (2006) are the first to examine conservatism in credit ratings. They attribute rating conservatism to the use of ratings in debt contracts. Specifically, they argue that certified credit rating agencies impose a higher standard when reporting favorable ratings than when reporting unfavorable ratings, i.e., they are conservative, because the use of ratings in debt contracts makes certified rating agencies’ loss function asymmetric. Beaver et al. (2006) provide an important initial step toward understanding rating conservatism. Because ratings issued by certified rating agencies are primarily used for debt contracting and those issued by non-certified rating agencies are mainly used for valuation reasons, Beaver et al. (2006) propose that certified rating agencies issue more conservative ratings than non-certified rating agencies. 8 Blume et al. (1998) also show that rating standards have become more stringent from 1978 through 1995.
  10. 10. 8 In this paper, we explore an alternative source of rating conservatism, that is, high information asymmetry.9 The maintained assumption in our paper is that rating agencies have an asymmetric loss function, because their reputation costs are higher when an issuer defaults but ratings indicate otherwise than when an issuer has a higher credit quality relative to its ratings (Beaver et al., 2006). We argue that high information asymmetry exacerbates rating agencies’ asymmetric loss function for the following reason. When information asymmetry is high, presumably because of insufficient public information, investors are more likely to rely on ratings for investment and contracting decisions. For example, Ball et al. (2008) show that when the quality of accounting information is lower and information asymmetry is higher, the performance pricing provision in the debt contract is more likely to be based on ratings than on accounting information. The greater reliance on ratings for investment and contracting reasons increases rating agencies’ reputation cost when an issuer defaults and ratings indicate otherwise, thereby making rating agencies’ loss function more asymmetric. To minimize reputation costs from failing to predict a default, rating agencies are likely to impose a higher standard when reporting favorable ratings than when reporting unfavorable ratings for bonds with high information asymmetry. Therefore, when information asymmetry is severe, we expect bonds to be rated more conservatively. 2.2 Cross-listed bonds, information asymmetry and conservative ratings Measuring information asymmetry is not straightforward in the context of the debt market. Information asymmetry proxies that are traditionally used in equity markets including the probability of informed trading (PIN) are not applicable for debt markets (Bessembinder et 9 Our analysis in Section 4.4.1 shows that although debt contracts are more likely to contain rating-triggered covenants for cross-listed bonds, the use of rating-triggered covenants is not prevalent for both cross-listed bonds and U.S. domestic bonds, suggesting that the use of ratings in contracting is not the exclusive reason for rating conservatism.
  11. 11. 9 al. 2007; Edwards et al. 2007). One innovation in our study is that we test our hypothesis by identifying a group of bonds that may have higher information asymmetry than another group, i.e., cross-listed bonds vs. U.S. domestic bonds.10 Cross-listed bonds differ from domestic bonds on important dimensions that increase information asymmetry. For instance, market participants have limited public information for foreign firms than for U.S domestic firms. Even though foreign firms issuing public debt are required to register with the Securities and Exchange Commission (SEC) and file a form 20-F which includes reconciliations to U.S. GAAP, the reconciliations provide limited information for U.S. investors relative to the amount of information available for U.S. firms (Bradshaw et al. 2004). While the 20-Fs reconcile the reasons for the differences in financial statements produced under U.S. and home GAAP, there are significant differences in the information contained in the notes to the financial statements.11 Further, reconciliations are less timely because they are required on an annual basis rather than on a quarterly basis. Besides limited financial reporting, information available on debt contracts is also limited in many foreign countries. For example, in Germany and Japan, the details of the debt contracts such as debt covenants are not subject to mandatory disclosure. In addition to limited quantity of information, the quality of information for foreign issuers might also be lower. Leuz et al. (2003) find that earnings management is more pervasive in countries where the legal protection of outside investors is weak. DeFond et al. (2006) show that earnings are less informative in countries with poor accounting quality and weak insider 10 We compare the ratings of cross-listed bonds to those of U.S. domestic bonds to find out the impact of information asymmetry on rating conservatism. While it might be interesting to compare the ratings of foreign bonds cross-listed in the U.S. with those listed in their domestic markets, such comparisons do not provide insights into ratings conservatism. 11 The extent of the reconciliations and required disclosures varies according to whether firms complete Item 17 or 18, in Form 20-F. Item 17 is reserved for limited offerings and requires fewer disclosures than Item 18, which essentially requires the non-U.S. issuer to provide the same disclosures as a U.S. issuer.
  12. 12. 10 trading law enforcement. Although cross-listing may improve a foreign firm’s information environment, Fernandes and Ferreira (2008) show that cross-listing in the U.S. improves the information environment for firms that originate from developed markets but not those from emerging markets. Moreover, although cross-listed bonds are subject to SEC’s jurisdiction, their public reports are still influenced by the regulatory environment and managerial incentives in their home countries. Consistent with the argument that SEC’s oversight does not entirely overcome the effect of local market, Lang et al. (2006) find that reconciled earnings of cross- listed firms are subject to more earnings management (smoothing, target management, lower association with share price, less timely recognition of losses) than earnings of U.S. firms. We consider the possibility that rating agencies collect private information to mitigate the information asymmetry. Rating agencies could communicate with the management of the cross- listed issuers and analyze their risk in detail. However, the private information collection is unlikely to fully offset the information asymmetry of cross-listed bonds. For example, financial analysts also have the incentive and ability to collect private information, yet we find that the analysts forecast dispersion of cross-listed bond issuers is still higher than that of domestic issuers. Later we analyze in detail whether private information collection is an alternative explanation of our findings. Because information asymmetry is higher for cross-listed bonds, investors are more likely to rely on ratings of cross-listed bonds for debt contracting and investment decisions. Greater reliance on ratings increases rating agencies’ exposure to losses when foreign issuers default but ratings indicate otherwise, thereby making rating agencies’ loss function even more asymmetric. Therefore, we hypothesize that the ratings of cross-listed bonds are more conservative than the ratings of U.S. domestic bonds. In addition, we also expect rating conservatism of cross-listed
  13. 13. 11 bonds to vary with rating agencies’ reputational cost. While rating agencies have higher reputation cost from underestimating than overestimating default risk (Beaver et al. 2006), the reputation cost of underestimating the default risk is larger for investment-grade bonds than for speculative-grade. Thus, we also expect that the rating conservatism of cross-listed bonds is more pronounced for investment-grade bonds. Although the link between information asymmetry and rating conservatism appears to be similar to that between information asymmetry and financial reporting conservatism (LaFond and Watts 2008), there are key differences. First, in our setting, the economic entity providing the conservative rating is an information intermediary with unique incentives. Although paid by the issuers, rating agencies are subject to reputation costs if they compromise independence to advance their own interests. Second, the information asymmetry studied in LaFond and Watts (2008) is between shareholders and managers, while in our setting the information asymmetry that prompts the conservative rating is between the borrower (shareholder and managers) and the lender (debtholder). Lastly, conservatism in financial reporting could be costly for other stakeholders such as equityholders who care about timely reporting of both positive and negative news. In contrast, ratings primarily serve debtholders who mainly desire timely reporting of negative news; however, rating conservatism could be costly to rating agencies in terms of lower rating accuracy and timeliness; it also could be costly to firms in the form of higher cost of debt, if the market fixates on ratings. 2.3. Differentiating between rating conservatism and private information explanations A conservative rating bias for cross-listed bonds relative to U.S. domestic bonds is also consistent with that rating agencies have access to private information that indicates higher default risk for cross-listed bonds, and that the lower ratings for cross-listed bonds accurately
  14. 14. 12 reflect rating agencies’ private information. When rating agencies have limited public information, they are likely to acquire more private information to offset their information disadvantage. Since private information collection quantifies unobservable firm- and country- level default risk that cannot be captured in a research setting, a conservative rating bias could potentially reflect the appropriate default risk of the cross-listed issuer.12 We note that higher rating conservatism in investment-grade bonds would be inconsistent with the “private information” explanation, because there are no reasons to presuppose that the private information acquired by rating agencies indicates greater credit risk for investment-grade than for speculative-grade bonds. Nevertheless, we conduct two additional tests to further differentiate between the rating conservatism and private information explanation. 2.3.1 The probability of missing defaults and raising false alarms A difference in ratings between cross-listed and U.S domestic bonds is only indirect evidence of rating conservatism. We provide direct evidence on rating conservatism by analyzing whether ratings agencies are accurate in predicting future default. Accordingly, we directly test (1) the probability of ratings failing to predict defaults accurately, and (2) the probability of ratings raising a false alarm about the probability of a default. If the lower cross- listed bond rating reflects rating conservatism, we expect ratings of the cross-listed bonds to be associated with more frequent false alarms and a lower probability of missing future defaults. In contrast, if lower ratings capture higher intrinsic default risk of cross-listed bonds, we do not expect the probabilities of false alarms and missing defaults to vary systematically between cross-listed and U.S. domestic bonds, since the initial ratings are unbiased. 2.3.2 The bond market correction 12 Since the private information of the rating agencies cannot be captured by observable issuer-, issue-, and country- characteristics, controlling for these characteristics would not eliminate this alternative explanation.
  15. 15. 13 We also use the bond market to differentiate between rating conservatism and private information explanation. Prior literature has established the bond market efficiency. For example, Hotchkiss and Ronen (2002) conclude that the informational efficiency of corporate bonds is similar to that of the underlying stocks. In addition, Covitz and Harrison (2003) find that bond market preempts 75% of the information contained in downgrades. An efficient bond market is expected to be unbiased and correct for a bias if there is any. Therefore, if bond markets perceive ratings to be conservative for cross-listed bonds, the comparative cost of debt is expected to be lower because the inherent quality of cross-listed bonds is superior to comparable U.S. bonds. In contrast, if the bond market perceives ratings as containing private information about the default risk of foreign issuers, the additional default risk is priced by the bond market. In this case, the cost of debt would be the same for foreign and domestic bonds for the same ratings category. Therefore, we expect a price correction under the rating conservatism explanation, but not under the private information explanation. In summary, under the conservatism explanation, we expect (1) a higher probability of raising false alarms and a lower probability of failing to predict defaults for cross-listed bond ratings than for domestic bond ratings, and (2) a bond market correction for the downward rating bias on cross-listed bonds. In contrast, under the private information explanation, we expect (1) no difference in the probability of raising false alarms and the probability of missing defaults between cross-listed and domestic bonds, and (2) no bond market price correction for the downward rating bias on cross-listed bonds.13 3 Data, Research Design, and Empirical Results 13 We do not claim that the rating conservatism and the private information explanations are mutually exclusive. It is possible that a rating bias is an outcome of both explanations. Our tests are designed to tease out the dominating explanation.
  16. 16. 14 3.1 Data description We obtain our sample bonds from the Global New Issues Database of the Securities Data Corporation (SDC) and the Mergent Fixed Investment Securities Database (FISD). We include non-governmental firms (both U.S. and non-U.S.) issuing fixed-rate public debt between 1990 and 2009. We exclude Over-the-Counter (OTC) issues and Rule 144A private placements.14 We delete observations when the spread to benchmark is missing or when offer yield to maturity is negative or coded by the words “Floats,” “Index,” “Market,” “Reset,” “Varies,” or “NA.” We manually match non-U.S. issuers to Datastream to obtain issuer characteristics. For U.S. issuers, we require necessary issuer-level data from Compustat Global. The final sample consists of 2,389 public debt issues by non-U.S. firms (treatment sample) and 11,345 public debt issues by U.S. firms (control sample). Table 1 presents the frequency distribution of cross-listed bonds by year. The number of public debt issues by non-U.S. firms increased from 5 in 1990 to 597 in 2001. However, the last eight years of the sample (2002 to 2009) exhibit a significant decline in the number of issues. The number of cross-listed public debt issues in 2008 was down to 55. This is consistent with the conjecture that the regulations imposed by Sarbanes-Oxley Act in 2002 have discouraged foreign firms from raising capital in the U.S. Similarly, the number of foreign countries raising debt in the U.S. increased from 2 to 21 during the period 1990 to 2001, but declined during the subsequent years of the sample. During the period 1990 to 2009, non-U.S. firms in our sample raised a total of $1.04 trillion in public debt in the U.S and the average size of the debt issued was $437 million. While the number of issues has decreased since 2002, the average size of the issue has increased steadily over the sample period from 1990 to 2009. 14 Rule 144A of the Securities Act of 1933, as amended, allows for the private resale of unregistered securities to “qualified institutional buyers,” which are generally large institutional investors with assets exceeding $100 million.
  17. 17. 15 Table 2 presents the frequency distribution, total public debt offering, and the average issue size of debt raised by the country of origin. Firms from 34 different countries raised public debt in the U.S during the sample period. Canadian firms raised debt most frequently in the U.S. followed by firms from United Kingdom, Germany and Israel. The total amount of debt issued in the U.S. was the largest for Dutch firms, followed by issuers from United Kingdom, Canada, and Germany. Table 3 reports the mean and median comparisons of various variables between cross- listed bonds and U.S. domestic bonds. Since the inferences from the means and medians are similar, we focus our discussion on the results from the mean comparisons. We find that the issue size of the debt offerings is not significantly different between non-U.S. and U.S. firms. Cross-listed bonds tend to have shorter maturities and they include special features such as callability, puttability and sinking funds less often. Cross-listed bonds have lower costs of debt compared to domestic bonds. Finally, non-U.S. firms raising debt in the U.S. are larger, less levered and more profitable. 3.2. Initial rating of cross-listed bonds and rating conservatism To test whether rating agencies are more conservative when rating cross-listed bonds, we estimate the following ordered probit regression, with ratings at issuance as the dependent variable.15 Rating = α + β1Non-US + β2Issue size + β3Maturity + β4 Seniority + β5Callability + β6Puttability + β7Sinking fund + β8Pay-in-kind + β9Default spread + β10Firm size + β11Leverage +β12Profitability + β13Interest coverage + β14Capital expenditures + β15Emerging + β16Civil law + β17Rule of law + β18Creditor rights + β19Judicial efficiency + β20Ex- ante self-dealing + β21Ex-post self-dealing + β22Anti-director rights + β23Public enforcement + β24Disclosure requirements + β25Liability standards + β26Investor protection + Year dummies + Industry dummies + ε (1) 15 We code ratings as 1 for the best rating category. Therefore, a lower value corresponds to a better rating. The complete rating scheme that we use can be found in Appendix B of Cheng and Neamtiu (2009).
  18. 18. 16 where: Rating = one for firms with the best credit rating (AAA) and the value increase by one for successively worse rating categories.16 Non-US = one when debt is issued by a non-U.S. firm and zero otherwise. Issue size = the natural logarithm of the size of the debt issue in millions of dollars. Maturity = the natural logarithm of the number of years to maturity. Seniority = one when the debt is senior and zero otherwise. Callability = one when the bond includes a call provision and zero otherwise. Puttability = one if the bond includes a put provision and zero otherwise. Sinking fund = one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind = one if the bond pays in kind other than cash and zero otherwise. Default spread = the yield difference between AAA- and BAA-rated corporate bonds. Firm size = total assets of the issuer at the end of the fiscal year prior to bond issuance. Leverage = total debt divided by total assets at the end of the fiscal year prior to bond issuance. Profitability = EBITDA divided by total assets in the fiscal year prior to bond issuance. Interest coverage = EBIT divided by interest expense in the fiscal year prior to bond issuance. Capital expenditure = Capital expenditure incurred by the issuer in the fiscal year prior to bond issuance. Emerging = one if the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law = one if the legal origin of the issuing country is the civil law. Rule of law = an index that assesses the extent to which investors have confidence in and abide by the rules of the society, as defined in La Porta et al. (2006). Creditor rights = an index between 0 and 4 that aggregates different creditor rights in case of bankruptcy and reorganization, as defined in La Porta et al. (1998). Judicial efficiency = an index between 0 and 10 that assesses the efficiency and integrity of the legal environment as it affects business and reflects the investors’ assessment of conditions in the country in question, as defined in La Porta et al. (1998). Ex-ante self- dealing & Ex-post self- dealing = indices that range between 0 and 1 and measure the approval requirements for managerial actions and the ease of proving wrongdoing against managers, respectively, as defined in Djankov et al. (2008). Anti-directors rights = an index between 0 and 6 that aggregates different investor rights against directors, as defined in Spamann (2010). Public enforcement = an index between 0 and 1 that aggregates various criminal sanctions against various parties, as defined in La Porta et al. (2006). Disclosure requirements = an index between 0 and 1 that assesses the strength of specific disclosure requirements, as defined in La Porta et al. (2006). Liability standards = an index between 0 and 1 that assesses the procedural difficulty in bringing lawsuits against managers, distributors and accountants, as defined in La Porta et al. (2006). Investor protection = a comprehensive index between 0 and 1 that aggregates various legal dimensions such as liability standards, investor rights and risk of 16 We define the initial rating as the first rating assigned to an issue during the first month after the offering date by Standard and Poor’s, Moody’s or Fitch. If a bond is rated by multiple agencies, we assign the highest of the ratings to the issue. Results do not change when we use the lowest rating or we use the average of all assigned ratings in the regressions.
  19. 19. 17 expropriation, as defined in La Porta et al. (2006). Our variable of interest is Non-US. A positive coefficient indicates that ratings of cross- listed bonds are more conservative than those of U.S. domestic bonds, after controlling for issue and issuer characteristics and country-specific variables. We include additional explanatory variables that are determinants of corporate bond ratings (e.g., Kaplan and Urwitz 1979; Reiter and Ziebart 1992). These control variables fall into three categories: issue characteristics, issuer characteristics, and country-specific variables. Prior studies typically find that issue characteristics are key determinants of bond ratings. For example, Bhojray and Sengupta (2003) posit that larger issues with shorter maturity have superior ratings because larger offerings have more public information and shorter-term bonds have less exposure to interest rate risks. They also conclude that callable bonds have worse ratings because of a prepayment risk and that senior bonds with sinking fund provisions have superior ratings because of a lower default risk. We also include indicator variables for puttable and pay-in-kind bonds. Puttable bonds offer the option of forcing the company to repurchase the bonds before maturity and pay-in-kind bonds allow the issuer the option of paying the bondholder in additional securities rather than cash. We also control for economic conditions at the time of the issue by including the default spread as an additional explanatory variable. We also control for issuer firm characteristics because they are the underlying determinants of the credit risk of a bond issue. We expect larger, less levered and more profitable firms to have superior ratings. We also include the interest coverage ratio which is a key determinant of the liquidity of a firm. Following Miller and Reisel (2011), we also control for capital expenditures to proxy for investment opportunities. Finally, we also control for country characteristics because country-specific default risk affects ratings of foreign issuers (Covrig et al. 2007; Francis et al. 2007). Rajan and Zingales
  20. 20. 18 (1995) find that there are large variations in the approaches taken by bankruptcy courts from different countries and that institutional differences may play a role in determining ratings. Ferri et al. (2001) find a strong linkage between sovereign ratings and firms’ private ratings for developing countries. Purda (2003) finds that, in addition to the influence of firm-specific variables on debt ratings, various country-specific factors predict debt ratings. Perraudin and Taylor (2004) also find that firms domiciled in Japan, Europe and the U.S. pay different yields for particular ratings categories. To control for the impact of country-specific factors on bond ratings, we include indicator variables for the country of origin (emerging economy) and its legal tradition (civil law). We also control for differences in legal environment as in La Porta et al. (1998, 2006), Djankov et al. (2008), and Spamann (2010). In Table 4, we examine whether, controlling for other factors, rating agencies assign differential initial ratings for cross-listed bonds. The p-values reported in the table are associated with robust t-statistics corrected for clustered errors. 17 We find a significantly positive coefficient on Non-U.S. Specifically, the coefficient on Non-U.S. is 1.25 and is significant at less than 1% level. Since higher values of the dependent variable correspond to worse ratings, the result indicates that ratings for non-U.S. firms issuing public debt in the U.S. are more than one notch worse than those for similar U.S. firms. Thus, our results suggest that rating agencies are more conservative when assigning initial ratings to cross-listed bonds than to domestic bonds. Consistent with prior literature, we find that ratings are superior for issues with shorter maturity, higher seniority and puttability feature. We also find superior ratings are assigned to 17 All the regression analysis in this paper report t-statistics corrected for standard errors clustered by firm. We use firm cluster because the large number of U.S. bonds lead to questionable large t-statistics using country cluster. Wooldridge (2003) points out that the clustered standard errors approach is not appropriate when the number of clusters is small relative to the number of observations in each cluster. Thus we use the more conservative approach by clustering at firm level. However, we also conduct sensitivity tests using standard errors clustered by country and find similar results.
  21. 21. 19 larger, less levered and more profitable firms, and firms with higher interest coverage. As expected, we find that ratings are better for foreign bonds from countries with stronger public enforcement, higher liability standards, and stronger investor protection. Other country-specific variables have no significant relationship with ratings. 3.3 Subsequent rating changes of cross-listed bonds To examine whether the conservative rating is only confined to the initial rating assignment, we also investigate various aspects associated with upgrade and downgrade decisions subsequent to the bond issuance. In particular, we replace the dependent variable in Model (1) with: a) an indicator variable for subsequent upgrades or downgrades; b) the number of upgrades or downgrades divided by the total number of assigned ratings; c) the number of days between the initial rating and the first upgrade or downgrade. If the initial conservative ratings are persistent, then we expect the cross-listed bonds: a) to be less likely to receive an upgrade and more likely to receive a downgrade, b) to receive less frequent upgrades and more frequent downgrades, and/or c) to take longer to receive an upgrades and shorter to receive a downgrade. Table 5 reports the probit regression results when the dependent variable is Upgrade (Downgrade), a dummy variable for subsequent rating changes. We find that the coefficient on Non-US is significantly negative when the dependent variable is Upgrade. This result indicates that foreign bonds are less likely to receive a rating upgrade within three years of the offering date. Table 6 reports the OLS regression results when the dependent variable is the relative frequency of subsequent upgrades or downgrades. We find a significantly negative coefficient on Non-US (-0.28, p = 0.00) when the dependent variable represents the frequency of subsequent
  22. 22. 20 upgrades, indicating that cross-listed bonds receive 28% fewer upgrades. We also find a significantly positive coefficient on Non-US (0.17, p = 0.04) when the dependent variable represents the frequency of subsequent downgrades, suggesting that cross-listed bonds receive 17% more frequent downgrades. Finally, Table 7 reports the OLS regression results when the dependent variable is the time interval before the first upgrade or downgrade. We find a significantly positive coefficient on Non-US when the dependent variable is the time interval before the first upgrade, suggesting that it takes longer for a cross-listed bond to receive an upgrade, conditional on the existence of an upgrade. On average, it takes more than half a year longer for a cross-listed bond to receive an upgrade than what it takes a domestic bond to receive an upgrade. We find no significant difference in the time interval before the first downgrade across cross-listed bonds and domestic bonds. Collectively, these results suggest that initial ratings of cross-listed bonds are more conservative and that the conservative rating bias persists subsequent to the issuance. 3.4 Cross-sectional variation in the conservative ratings of cross-listed bonds After establishing the initial lower ratings and subsequent persistence of these lower ratings for cross-listed bonds, we explore the cross-sectional variation of the lower ratings among cross-listed bonds. Specifically, we estimate Model (1) separately for investment- and speculative-grade cross-listed bonds. Table 8 repeats the analysis in Table 4 for investment- and speculative-grade bond subsamples. We observe that while the coefficients on Non-US are positive and significant for both investment-grade and non-investment-grade debt, it is higher in magnitude for investment-grade debt (1.34 vs. 0.55). Therefore, although on average lower ratings for foreign bonds exist in the full sample, the conservative bias is more pronounced for
  23. 23. 21 investment-grade bonds. Because rating agencies have higher reputation costs from failing to predict the default of an investment-grade bond, they are more likely to be cautious when attaching an investment-grade rating to a foreign issuer. Thus, the results from Table 8 are consistent with reputational costs to ratings agencies being substantially higher for investment- grade bonds. While one may consider the conservative rating bias analogous to the “home bias” documented in the accounting and finance literature, we highlight three important differences.18 First, home bias in the equity market refers to disproportionately higher holdings of the domestic shares, but rating bias does not speak to the domestic holdings of foreign bonds. Second, credit quality attestation by rating agencies is a distinct feature of the bond market and rating agencies have their own incentives. While home bias and rating bias may share some common causes such as limited access to information about foreign firms, rating bias has a unique determinant, i.e. rating conservatism arising from rating agencies’ reputation cost. Third, while information asymmetry has been the main explanation for home bias, rating agencies’ private information collection may mitigate the information asymmetry. 3.5 Differentiating between rating conservatism and private information explanation If the cross-listed bonds have different ratings compared to domestic bonds, it is consistent with both the rating conservatism explanation and the private information explanation. Under the private information explanation, the rating differences between comparable cross- listed and U.S. bonds could indicate rating agencies’ assessment of the intrinsic default risk of cross-listed bonds. We distinguish between the two competing explanations using the following two tests. 18 A “home bias” refers to investors holding less than the optimal amount of foreign equities and requiring a higher rate of return for foreign firms compared to domestic firms (French and Poterba (1991); Cooper and Kaplanis (1994); Armstrong and Riddick (2000); and Bradshaw et al. (2004)).
  24. 24. 22 3.5.1 The probability of missing defaults and raising false alarms As in Cheng and Neamtiu (2009), we define an indicator variable for missed defaults and an indicator variable for false alarms. For a sample of issuers with a default, Missed default equals one if an issue defaults within one year from the rating date but the rating indicates a low default risk (investment-grade), and zero otherwise. False alarm equals one if an issuer does not default within one year from the rating date but the rating indicates a high default risk (non- investment-grade), and zero otherwise. In the regression using Missed default as a dependent variable, a negative coefficient on Non-US indicates that rating agencies are less likely to miss predicting the default of cross-listed bonds that are investment-grade than similar U.S. domestic bonds. In the regression using False alarm as a dependent variable, a positive coefficient on Non-US indicates that rating agencies are more likely to provide false warnings about the default risk of cross-listed bonds that are non- investment-grade than similar U.S. domestic bonds. Taken together, a negative coefficient on Non-US for Missed default regression and a positive coefficient on Non-US for the False alarm regression are consistent with rating conservatism, i.e., rating agencies trade off more false alarms for less missed defaults to reduce the costs associated with missing defaults for cross- listed bonds. Under the private information explanation, there are no reasons for the coefficients of Missed default and False alarm to vary between cross-listed and U.S. domestic bonds. Table 9 presents the results from directly comparing the probabilities of missing defaults and raising false alarms between cross-listed bonds and U.S. domestic bonds. In particular, we regress the dummy variables for missing defaults and false alarms on the Non-US dummy along with other control variables. The significantly negative coefficient on Non-US in the first column of Table 9 suggests that rating agencies are less likely to miss predicting the default of an
  25. 25. 23 investment-grade cross-listed bond compared to a similarly rated U.S. domestic bond. Concurrently, the significantly positive coefficient on Non-US in the second column of Table 9 implies a higher probability of false alarms and indicates that non-investment-grade cross-listed bonds are more likely to give false alarms than non-investment-grade U.S. domestic bonds. These results indicate that rating agencies are willing to incur a higher cost by raising false alarms to reduce the cost of missing a default, thereby providing direct evidence that rating agencies are conservative in rating cross-listed bonds to reduce the cost associated with missing defaults. 3.5.2 The bond market correction We use the cost of debt on the issuance date as a second test to differentiate between the conservative rating bias and private information explanations. We use Spread to benchmark to proxy for the cost of debt.19 The spread to benchmark is defined as the yield to maturity on the offer date (Offer yield) minus the yield of a U.S. Treasury security issued on the same date with comparable maturity. We test for differences in the cost of debt between non-U.S. and U.S. firms using the following regression: Cost of debt = α + β1Non-US + β2Rating + β3Issue size + β4Maturity + β5 Seniority + β6Callability + β7Puttability + β8Sinking fund + β9Pay-in-kind + β10Default spread + β11Firm size + β12Leverage + β13Profitability + β14Interest coverage + β15Capital expenditures + β16Emerging + β17Civil law + β18Rule of law + β19Creditor rights + β20Judicial efficiency + β21Ex-ante self-dealing + β22Ex-post self-dealing + β23Anti-director rights + β24Public enforcement + β25Disclosure requirements + β26Liability standards + β27Investor protection + Year dummies + Industry dummies + ε (2) Our interest is on β1, the coefficient of Non-US, which captures the bond market’s “correction” for a possible rating bias. A negative β1 indicates that, compared to U.S. firms with similar ratings and issue characteristics, non-U.S. firms have a lower cost, consistent with the 19 Spread to benchmark is winsorized at the 1st and 99th percentiles to control for outliers. Our results are qualitatively similar when we do not winsorize the variables.
  26. 26. 24 bond market perceiving the rating bias as an outcome of rating conservatism. In contrast, an insignificant β1 indicates that, compared to U.S. firms with similar ratings and issue characteristics, non-U.S. firms have similar cost of debt which is consistent with the private information hypothesis. Because the control variables in the cost of debt regressions have been commonly used in prior literature (e.g., Kidwell et al. (1984); Miller and Puthenpurackal (2002)), we provide a limited discussion of these variables. We expect the cost of debt to be lower for bonds with superior credit ratings (Ratings). Larger issues (Issue size) have lower cost of debt because these issues tend to generate more public information. Bonds with longer maturity (Maturity) are expected to have higher cost of debt because of greater interest rate risk. Bonds with Callability (Puttability) feature give the issuer (bondholder) the option to force the bondholder (issuer) into prepayment (repurchase) before the maturity date, which results in higher (lower) cost of debt. Senior bonds (Seniority) are less risky than subordinated bonds and the presence of Sinking fund provisions can reduce the default risk of an issue by requiring orderly payment of the principal over the bond’s life which lowers the cost of debt. Bondholders might be more averse towards Pay-in-kind bonds, increasing the cost of debt. We include the Default spread to control for the economic conditions at the time of the issue. In addition, we expect larger, less levered and more profitable firms, and firms with high interest coverage to have lower cost of debt. We again control for capital expenditures to proxy for investment opportunities. Finally, as in the rating regressions, we control for a large set of country-specific factors. Table 10 presents the results from comparing the initial costs of debt for cross-listed and U.S. bonds in the same rating group. Specifically, we run OLS regressions of Spread to
  27. 27. 25 benchmark on Non-US after controlling for ratings and issue, issuer and country characteristics.20 We find that for the full sample, the coefficient on Non-US is significantly negative, consistent with the expectation that the bond market corrects for the rating bias by assigning a lower spread for the cross-listed bonds in the same rating group as the domestic bonds. After partitioning the sample into investment and speculative-grade bonds, we find that on average, spreads of investment-grade, cross-listed bonds are 42 basis points lower than the spreads of U.S. bonds with similar ratings and issue characteristics, while we find no such results for speculative-grade bonds. The correction for rating bias only in investment-grade debt is consistent with rating bias being more pronounced among investment-grade bonds (Table 8). Overall, the results from the cost of debt analyses are again consistent with the rating conservatism and inconsistent with the private information explanation. The results on most of the control variables from Table 10 are consistent with prior studies. The coefficients on Rating and Maturity are significantly positive. Consistent with our expectations, Callability (Puttability) has a significantly positive (negative) coefficient. Bonds that pay-in-kind have higher costs of debt. Larger, less levered and more profitable firms and firms with higher interest coverage have lower costs of debt. The adjusted-R2 is 65% for the full- sample regression. 4 Additional analyses and robustness tests 4.1 Rating-based covenant We argue that investors of cross-listed bonds are more dependent on ratings for investment and contracting reasons. To substantiate this claim, we study the rating-based 20 We repeat the regressions with Offer yield as the dependent variable. Results are very similar and available from the authors upon request.
  28. 28. 26 covenants contained in the bond contracts. We find that 4.00% of the non-U.S. bonds in our sample have the rating trigger covenant whereas only 0.98% of the domestic bonds contain similar covenants. In an untabulated probit regression of the rating trigger covenant dummy on the cross-listed dummy and control variables, the cross-listed dummy has a coefficient of 0.74, which is highly significant, indicating that the cross-listed bonds are more likely to include rating trigger covenants than U.S. domestic bonds. This analysis lends support to our argument that ratings of cross-listed bonds are more likely to be used in debt contracting. 4.2 Analyst forecast dispersion In order to present additional evidence on the higher information asymmetry for cross- listed firms, we investigate the analyst forecasts dispersion because it has been used as a measure of information asymmetry (Diether et al. 2002). We merge our full sample with the I/B/E/S forecasts both for the nearest quarterly EPS forecast and the nearest annual EPS forecast prior to the bond issuance. Since the standard deviation of the forecasts is scale dependent, we standardize the measure by dividing the standard deviation by the absolute value of the mean forecast. Then we test the equality of means for the dispersion measure between foreign and U.S. firms on 7,330 U.S. and 753 non-U.S. observations for the nearest annual EPS forecast. The means of the dispersion measure are 0.0708 and 0.1532 for U.S. vs. cross-listed firms, respectively, and the difference is significantly different from zero. Similarly, for the nearest quarterly EPS forecast, the means of the dispersion measure are 0.1121 and 0.1452 based on a slightly smaller sample (7,030 U.S. and 455 non-U.S. observations) and the difference is again significantly different from zero. These results provide corroborating evidence to our argument that cross-listed firms have higher information asymmetry than U.S. firms even after private information collection of analysts.
  29. 29. 27 4.3 Upgrade and downgrade analysis In the reported results in Tables 5 to 7, we focus on the three years after issuance to identify upgrades and downgrades. However, our results remain the same if we change the horizons to one or five years to identify rating changes. 4.4 Missing default and false alarm analysis We use investment- and non-investment-grade dichotomy to define the dummy variables for Missed default and False alarm analyses in Table 9. However, our results are robust to other cutoff points such as CCC+ and CC. 4.5 Robust standard error clustered by country In our main analysis we use standard errors clustered by firm because large number of observations from U.S. renders questionable statistics using country cluster. Nonetheless, we also conduct our analyses using country-cluster and all our results remain unchanged and even strengthened in some cases. Thus, we conclude that firm-cluster is a more conservative approach. 4.5 The bond market analysis We conduct a variety of sensitivity analyses to assess the robustness of our cost of debt results. First, even though we report t-statistics from pooled regressions that are corrected for clustered errors, the test statistics might be inflated if the residuals are cross-sectionally correlated. Therefore, similar to the Fama and MacBeth (1973) procedure, we estimate our regressions by year and the tests of significance are based on the distribution of the annual coefficient estimates. We find that our results are qualitatively similar using the Fama-MacBeth approach. For investment-grade debt, the time-series average of the coefficient of Non-US in the spread to benchmark regression is -36.48 and highly statistically significant. The result remains
  30. 30. 28 unchanged when we use Offer yield as another measure of the cost of debt.21 Second, we also examine whether the inclusion or exclusion of firms from the neighboring countries of the U.S. (Canadian and Mexican) affects our results. We find that our results are qualitatively similar and even stronger when we exclude firms from Mexico and Canada. Finally, when we exclude bonds with special features (callable, puttable, pay-in-kind and sinking funds), our results remain unchanged. If we drop all bonds with special features from the investment-grade bond sample, the coefficient of Non-US becomes -55 and is still highly significant. 5 Conclusion We find that cross-listed bonds have more conservative ratings than comparable U.S. domestic bonds. In particular, cross-listed bonds are associated with lower ratings at issuance compared to similar U.S. domestic bonds, they are also less likely to be upgraded, are associated with fewer upgrades, and take longer to receive an upgrade subsequent to the issuance. In addition, we show that the lower rating is concentrated among investment-grade cross-listed bonds, consistent with rating agencies bearing a higher reputation cost when they fail to predict the default of an investment-grade bond. To differentiate between the rating conservatism and private information explanations, we provide direct evidence that compared to similar U.S. bonds, cross-listed bond ratings are more likely to be associated with false alarms and are less likely to be associated with a missed default. Moreover, we conclude that the bond market corrects for the rating bias in cross-listed bonds because we find that the issuance yield spread is lower for cross-listed bonds than similar 21 Based on some of the recent evidence that the change in default premium affects the cost of capital rather than the level of the default premium, we replace Default with ∆Default in equation (1), where ∆Default is defined as the default premium for the current year less the number from the prior year. We find that our result is robust to this modification.
  31. 31. 29 for U.S. domestic bonds. Collectively, the results are consistent with the rating conservatism explanation and inconsistent with the private information explanation. Our paper extends the rating conservatism literature (Beaver et al. 2006) by identifying information asymmetry as another explanation for rating conservatism. We provide evidence that rating agencies are more conservative when information asymmetry is stronger. Our study complements other academic studies on rating properties which tend to focus on accuracy and timeliness (Cheng and Neamtiu 2009) by documenting conservatism in ratings as an additional rating property. Our paper also contributes to a small yet growing literature on the application of conservatism principle outside financial reporting (Lu and Sapra 2009; Hugon and Muslu 2010). Lastly, our paper adds to the very limited research on foreign firms issuing debt in the U.S. by documenting the rating properties of cross-listed bonds and their pricing implications.
  32. 32. 30 REFERENCES Ahmed, A., B. Billings, R. Morton, & M. Stanford-Harris (2002). The role of accounting conservatism in mitigating bondholder-shareholder conflicts over dividend policy and reducing debt costs. The Accounting Review 77, 867-890. Armstrong, S.V., & L. Riddick (2000). Evidence that differences in bankruptcy law among countries affect firm returns. Working Paper, University of Washington. Ball, R., S.P. Kothari, & A. Robin (2000). The effect of international institutional factors on properties of accounting earnings. Journal of Accounting and Economics 29, 1-51. Ball, R., R. Bushman, & F. Vasvari (2008). The debt-contracting value of accounting information and loan syndicate structure, Journal of Accounting Research 46, 247-287. Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics 24, 3-37. Beaver, W.H., C. Shakespeare, & M.T., Soliman (2006). Differential properties in the ratings of certified vs. non-certified bond rating agencies. Journal of Accounting and Economics 42, 303-334. Bessembinder, H., W. Maxwell, & K. Venkataraman (2007). Market transparency, liquidity externalities, and institutional trading costs in corporate bonds. Journal of Financial Economics 82, 251-288. Bhojraj, S., & P. Sengupta (2003). Effect of corporate governance on bond ratings and yields: the role of institutional investors and outside directors. Journal of Business 76, 455-475. Blume, M., F. Lin, & A.C. MacKinlay (1998). The declining credit quality of U.S. corporate debt: myth or reality? Journal of Finance 53, 1389–1413. Bradshaw M.T., B.J. Bushee, & G.S. Miller (2004). Accounting choice, home bias and U.S. investment in non-U.S. firms. Journal of Accounting Research 42, 795-841. Chaplinsky, S., & L. Ramchand (2004). The impact of SEC rule 144A on corporate debt issuance by international firms. Journal of Business 77, 1073-1097. Cheng, M., & M. Neamtiu (2009). An empirical analysis of changes in credit rating properties: timeliness, accuracy and volatility. Journal of Accounting and Economics 47, 108-130. Cooper, I., & E. Kaplanis (1994). Home bias in equity portfolios, inflation herding and international capital market equilibrium. Review of Financial Studies 7, 45-60. Covitz, D.M., & P. Harrison (2003). Testing conflicts of interest at bond ratings agencies with market anticipation: evidence that reputation incentives dominate. Working paper, Federal Reserve Board. Covrig, V., M. DeFond, M., & M. Hung (2007). Home bias, foreign mutual fund holdings, and voluntary adoption of international accounting standards. Journal of Accounting Research 45, 41-70. DeFond M., M. Hung, & R. Trezevant (2006). Investor protection and the information content of annual earnings announcements: international evidence. Journal of Accounting and Economics 43, 37-67.
  33. 33. 31 Diether K.B, C.J. Malloy, & A. Scherbina (2002). Differences of opinion and the cross-section of stock returns. Journal of Finance 67, 2113-2141. Djankov, S., R. La Porta, F. Lopez-de-Silanes, & A. Shleifer (2008). The law and economics of self- dealing. Journal of Financial Economics 88, 430-465. Ederington, L., J. Yawed, & B. Roberts (1987). The information content of bond ratings. Journal of Financial Research 10, 211-226. Edwards, A.K., H. Lawrence, & S.P. Michael (2007). Corporate bond market transparency and transactions costs. Journal of Finance 62, 1421-1452. Fama, E.F., & J.D. MacBeth (1973). Risk, return and equilibrium: empirical tests. The Journal of Political Economy 81, 607-636. Fama, E.F., & K.R. French (1997). Industry costs of equity. Journal of Financial Economics 43, 153-193. Fernandes, N., & M.A. Ferreira (2008). Does international cross-listing improve the information environment? Journal of Financial Economics 88, 216-244. Ferri, G., L. Liu, & G. Majnoni (2001). The role of the rating agency assessments in less developed blue countries: the impact of the proposed Basil guidelines. Journal of Banking and Finance 25, 115-148. Francis, B.B., I. Hasan, & M. Waisman (2007). Does geography matter to bondholders? Working Paper, Federal Reserve Bank of Atlanta. French, K.R., & J. Poterba (1991). Investor diversification and international equity markets. American Economic Review 81, 222-226. Holthausen, R., & R. Watts (2001). The relevance of the value relevance literature for financial accounting standard setting. Journal of Accounting and Economics 31, 3-76. Hotchkiss, E., & R. Ronen, (2002). The informational efficiency of the corporate bond market: an intraday analysis. Review of Financial Studies 15, 1325-1354. Hugon, A., & V. Muslu (2010). Market demand for conservative analysts. Journal of Accounting and Economics 50, 42-57. Kaplan, R.S., & G. Urwitz (1979). Statistical models of bond ratings: a methodological inquiry. Journal of Business 52, 231-261. Karolyi, G. A. (2006). The world of cross-listings and cross-listings of the world: challenging conventional wisdom. Review of Finance 10, 99-152. Kidwell, D., M.W. Marr, & G.R. Thompson (1984). SEC Rule 415: the ultimate competitive bid. Journal of Financial and Quantitative Analysis 19, 183-195. La Porta, R. F. Lopez-de-Silanes, A. Shleifer, & R.W. Vishny (1998). Law and finance. Journal of Political Economy 106, 1113-1155.
  34. 34. 32 La Porta, R., F. Lopez-de-Silanes, & A. Shleifer (2006). What works in securities laws. Journal of Finance 61, 1-32. LaFond, R., & R. Watts (2008). The information role of conservatism. The Accounting Review 83, 447- 478. Lang, M., H.K. Lins, & D. Miller (2003). ADRs, analysts, and accuracy: does cross-listing in the U.S. improve a firm’s information environment and increase market value. Journal of Accounting Research 41, 317-343. Leuz, C., D. Nanda, & P. Wysocki (2003). Investor protection and earnings management: an international comparison. Journal of Financial Economics 69, 505-527. Lu, T., & H. Sapra (2009). Auditor conservatism and investment efficiency. The Accounting Review 84, 1933-1958. Miller, D.P., & J.J. Puthenpurackal (2002). The costs, wealth effects, and determinants of international capital raising: evidence from public cross-listed bonds. Journal of Financial Intermediation 11, 455- 485. Miller, D.P., & N. Reisel (2011). Do country-level investor protection impact security-level contract design? Evidence from foreign bond covenants. Review of Financial Studies 25. Nikolaev, V.V. (2010). Debt covenants and accounting conservatism. Journal of Accounting Research 48, 51-89. Perraudin, W., & A. Taylor (2004). On the consistency of ratings and bond market yields. Journal of Banking and Finance 28, 2769-2788. Purda, L.D. (2003). Consistency of global credit ratings: an analysis of firm versus country-specific factors. Working Paper, Queen’s University. Rajan, R., & L.G. Zingales (1995). What do we know about capital structure? Some evidence from international data. Journal of Finance 50, 1421-1460. Reiter, S., & D. Zeibart (1991). Bond yields, ratings, and financial information: evidence from public utility issuers. Financial Review 26, 44-74. Spamann, H. (2010). The antidirector rights index revisited. Review of Financial Studies 23, 467-486. Watts, R. (2003a). Conservatism in accounting, part I: explanations and implications. Accounting Horizons 17, 207-221. Watts, R. (2003b). Conservatism in accounting, part II: evidence and research opportunities. Accounting Horizons 17, 287-301. Wooldridge, J.M. (2003). Cluster-sample methods in applied econometrics. American Economic Review 93, 133-138. Zhang, J. (2008). The benefits of accounting conservatism to lenders and borrowers. Journal of Accounting and Economics 45, 27-54.
  35. 35. 33 Table 1 Number, origin and size of cross-listed bonds over time $ Million $ Million Year Number of issues Number of countries Total debt issued Average issue size 1990 5 2 2,000 400 1991 19 4 5,350 282 1992 28 7 6,975 249 1993 40 9 11,320 283 1994 35 11 7,809 223 1995 56 12 14,246 254 1996 72 15 14,508 201 1997 86 16 19,117 222 1998 96 14 23,544 245 1999 195 19 46,824 240 2000 272 20 225,220 828 2001 597 21 245,568 411 2002 368 22 122,410 333 2003 177 16 53,600 303 2004 67 13 22,590 337 2005 45 10 24,225 538 2006 69 10 51,356 744 2007 87 14 71,422 821 2008 55 7 54,130 984 2009 20 7 21,700 1,085 Total 2,389 1,043,912 437 Table 1 presents the number of issues and the number of countries that issuing firms originate from each year and the annual total and average size of fixed-rate non-governmental public debt issues in the U.S. by non-U.S. firms between 1990 and 2009. The total and average issue sizes are in terms of millions of dollars. The data are obtained from the SDC Global New Issues Database and Mergent Fixed Investment Securities Database. We exclude observations if the spread to benchmark is missing, or the yield to maturity is negative, or the yield to maturity is coded by non-numeric characters, such as “Floats”, “Index”, “Market”, “Reset”, “Varies” or “NA.”. We also exclude observations if country-specific or firm-specific factors are missing for an issuer.
  36. 36. 34 Table 2 Cross-listed bonds grouped by country of origin Country Number of issues Total debt issued Average issue size Argentina 20 3,800 190 Australia 87 38,866 447 Austria 22 12,768 580 Belgium 32 3,990 125 Brazil 15 11,530 769 Canada 663 186,415 281 Chile 14 3,160 226 Colombia 1 400 400 Denmark 23 4,260 185 Finland 9 3,175 353 France 161 95,875 595 Germany 218 132,782 609 Hong Kong 9 2,725 303 India 1 300 300 Indonesia 6 1,050 175 Ireland-Rep 9 2,434 270 Israel 209 2,718 13 Italy 24 18,166 757 Japan 7 8,500 1,214 Malaysia 2 630 315 Mexico 14 4,713 337 Netherlands 167 225,979 1,353 New Zealand 4 800 200 Norway 25 8,255 330 Philippines 12 2,200 183 Portugal 14 2,040 146 Singapore 3 1,375 458 South Africa 1 1,000 1,000 South Korea 22 5,961 271 Spain 33 17,498 530 Sweden 91 17,885 197 Switzerland 34 7,344 216 United Kingdom 433 214,918 496 Venezuela 4 400 100 Table 2 presents the number of issues and the total and average size of fixed-rate non-governmental public debt issues in the U.S. for each foreign country that issuing firms originate from between 1990 and 2009. The total and average issue sizes are in terms of millions of dollars. The data are obtained from the SDC Global New Issues Database and Mergent Fixed Investment Securities Database. We exclude observations if the spread to benchmark is missing, or the yield to maturity is negative, or the yield to maturity is coded by non-numeric characters, such as “Floats”, “Index”, “Market”, “Reset”, “Varies” or “NA.” We also exclude observations if country-specific or firm-specific factors are missing for an issuer.
  37. 37. 35 Table 3 Descriptive statistics for cross-listed bonds and U.S. domestic bonds Means Medians Non-U.S. U.S. p-value Non-U.S. U.S. p-value Issue characteristics Issue size 5.118 5.155 0.22 5.298 5.371 0.26 Maturity 1.448 2.189 0.00 1.619 2.304 0.00 Seniority 0.953 0.938 0.00 1.000 1.000 0.00 Callability 0.207 0.259 0.00 0.000 0.000 0.00 Puttability 0.032 0.084 0.00 0.000 0.000 0.00 Sinking fund 0.002 0.007 0.00 0.000 0.000 0.01 Pay-in kind 0.000 0.000 0.03 0.000 0.000 0.03 Default spread 0.922 0.990 0.00 0.840 0.880 0.00 Rating 5.375 6.744 0.00 5.000 6.000 0.00 Spread to benchmark 86.185 154.378 0.00 71.000 113.000 0.00 Offer yield 5.708 6.529 0.00 5.930 6.381 0.00 Issuer characteristics Firm size 0.156 0.0001 0.00 0.058 0.0001 0.00 Leverage 0.267 0.294 0.00 0.240 0.262 0.00 Profitability 0.051 0.024 0.00 0.022 0.020 0.00 Interest coverage 8.947 10.352 0.66 3.560 3.413 0.84 Capital expenditures 0.037 0.036 0.36 0.004 0.021 0.00 Country-specific variables Emerging 0.141 0.000 0.00 0.000 0.000 0.00 Civil law 0.405 0.000 0.00 0.000 0.000 0.00 Rule of law 1.710 1.920 0.00 1.970 1.920 0.00 Creditor rights 2.163 1.000 0.00 2.000 1.000 0.00 Judicial efficiency 9.247 10.000 0.00 9.250 10.000 0.00 Ex-ante self-dealing 0.422 0.330 0.00 0.330 0.330 0.05 Ex-post self-dealing 0.747 0.980 0.00 0.900 0.980 0.00 Anti-director rights 4.259 2.000 0.00 4.000 2.000 0.00 Public enforcement 0.629 0.000 0.00 1.000 0.000 0.00 Disclosure requirements 0.696 1.000 0.00 0.667 1.000 0.00 Liability standards 0.629 1.000 0.00 0.660 1.000 0.00 Investor protection 0.597 1.000 0.00 0.594 1.000 0.00 Observations 2,389 11,345 2,389 11,345 Table 3 presents the descriptive statistics for cross-listed bonds and U.S. domestic bonds. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Rating equals one for firms that have the best credit rating (AAA) and increases by one for successively lower rating categories. Spread to benchmark is the difference between the offer yield and the yield of a U.S. Treasury security issued on the same date with comparable maturity. Offer yield is the yield to maturity on the offer date. Firm size is equal to total assets. dollars. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor rights and Judicial efficiency as defined in La Porta et al. (1998), Ex-ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.
  38. 38. 36 Table 4 Initial rating of cross-listed bonds compared to U.S domestic bonds Full sample Independent Variables Coefficient (p-value) Non-US 1.25 (0.00) Issue characteristics Issue size 0.04 (0.22) Maturity 0.05 (0.08) Seniority -0.81 (0.00) Callability 0.05 (0.75) Puttability -0.87 (0.00) Sinking fund -0.11 (0.71) Pay-in kind 0.39 (0.14) Default spread -0.01 (0.92) Issuer characteristics Firm size -1.70 (0.00) Leverage 1.13 (0.00) Profitability -4.78 (0.00) Interest coverage -0.01 (0.05) Capital expenditures -0.57 (0.57) Country-specific variables Emerging 0.25 (0.49) Civil law -0.30 (0.43) Rule of law -0.46 (0.17) Creditor rights 0.05 (0.60) Judicial efficiency 0.00 (0.97) Ex-ante self-dealing -0.65 (0.10) Ex-post self-dealing 0.02 (0.96) Anti-director rights -0.17 (0.12) Public enforcement -0.53 (0.01) Disclosure requirements -0.93 (0.24) Liability standards -1.11 (0.03) Investor protection -2.50 (0.01) Industry dummies Yes Yes Year dummies Yes Yes Pseudo R-squared 9.38% Observations 13,734 Table 4 presents the results from cross-sectional regressions with Rating as the dependent variable. The reported results are based on ordered probit estimation using robust standard errors clustered by issuer. Rating equals one for firms that have the best credit rating (AAA) and increases by one for successively worse rating categories. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Firm size is equal to total assets. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor rights and Judicial efficiency as defined in La Porta et al. (1998), Ex-ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.
  39. 39. 37 Table 5 Likelihood of subsequent rating upgrades and downgrades Upgrade Downgrade Independent Variables Coefficient (p-value) Coefficient (p-value) Non-US -0.72 (0.02) 0.14 (0.63) Issue characteristics Issue size -0.02 (0.51) -0.07 (0.00) Maturity -0.05 (0.14) -0.01 (0.74) Seniority 0.37 (0.00) 0.08 (0.19) Callability 0.05 (0.41) 0.00 (0.97) Puttability 0.29 (0.09) -0.03 (0.87) Sinking fund -0.50 (0.04) Default spread 0.09 (0.44) -0.35 (0.00) Issuer characteristics Firm size -1.24 (0.05) -1.37 (0.01) Leverage -0.33 (0.02) 0.28 (0.01) Profitability -1.21 (0.33) -2.24 (0.01) Interest coverage 0.02 (0.00) 0.00 (0.11) Capital expenditures -0.04 (0.94) 0.47 (0.28) Country-specific variables Emerging -1.27 (0.08) 0.43 (0.38) Civil law 1.17 (0.12) 0.33 (0.50) Rule of law 2.24 (0.00) 0.36 (0.49) Creditor rights 0.29 (0.14) -0.03 (0.81) Judicial efficiency 0.56 (0.03) -0.04 (0.83) Ex-ante self-dealing 0.84 (0.35) -0.14 (0.84) Ex-post self-dealing 2.50 (0.02) 0.73 (0.37) Anti-director rights 0.55 (0.06) -0.21 (0.15) Public enforcement -0.12 (0.79) 0.14 (0.66) Disclosure requirements -2.19 (0.09) -0.36 (0.71) Liability standards 0.42 (0.66) -0.14 (0.84) Investor protection 2.32 (0.13) -0.26 (0.81) Industry dummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Pseudo R-squared 9.15% 7.94% Observations 6,114 6,114 Table 5 presents the results from cross-sectional regressions with indicator variables for upgrades and downgrades as the dependent variables. In the UPGRADE (DOWNGRADE) regression, the dependent variable takes the value of one if the issue’s rating has been upgraded (downgraded) by Standard and Poor’s at the end of the three years after the offering compared to the initial rating. The reported results are based on probit estimation using robust standard errors clustered by issuer. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Firm size is equal to total assets. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor rights and Judicial efficiency as defined in La Porta et al. (1998), Ex-ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.
  40. 40. 38 Table 6 Frequency of subsequent upgrades and downgrades Upgrade Downgrade Independent Variables Coefficient (p-value) Coefficient (p-value) Non-US -0.28 (0.00) 0.17 (0.04) Issue characteristics Issue size 0.01 (0.00) -0.02 (0.00) Maturity 0.01 (0.00) -0.06 (0.00) Seniority -0.05 (0.00) 0.03 (0.00) Callability -0.02 (0.00) 0.09 (0.00) Puttability 0.03 (0.00) 0.03 (0.19) Sinking fund 0.06 (0.00) -0.23 (0.00) Pay-in-kind 0.02 (0.10) -0.06 (0.00) Default spread 0.11 (0.44) -0.35 (0.00) Issuer characteristics Firm size 0.03 (0.68) -0.36 (0.00) Leverage 0.00 (0.73) 0.03 (0.06) Profitability 0.45 (0.00) -0.45 (0.00) Interest coverage 0.00 (0.39) 0.00 (0.22) Capital expenditures -0.02 (0.69) -0.22 (0.00) Country-specific variables Emerging -0.01 (0.85) 0.14 (0.14) Civil law 0.11 (0.08) 0.02 (0.86) Rule of law 0.27 (0.00) -0.23 (0.03) Creditor rights -0.04 (0.06) -0.01 (0.78) Judicial efficiency 0.10 (0.00) -0.04 (0.40) Ex-ante self-dealing 0.20 (0.01) 0.04 (0.74) Ex-post self-dealing 0.24 (0.04) -0.23 (0.17) Anti-director rights 0.14 (0.00) -0.07 (0.07) Public enforcement 0.09 (0.00) -0.06 (0.26) Disclosure requirements -0.07 (0.62) 0.37 (0.06) Liability standards 0.41 (0.00) -0.17 (0.22) Investor protection -0.46 (0.00) -0.10 (0.66) Industry dummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes R-squared 15.32% 23.47% Table 6 presents the results from cross-sectional regressions with the relative frequencies of upgrades and downgrades as the dependent variables. In the UPGRADE (DOWNGRADE) regression, the dependent variable is equal to the number of upgrades (downgrades) divided by the total number of ratings assigned to an issue by Standard & Poor’s during the three years after the offering. The reported results are based on OLS estimations using robust standard errors clustered by issuer. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Firm size is equal to total assets. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor rights and Judicial efficiency as defined in La Porta et al. (1998), Ex-ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.
  41. 41. 39 Table 7 Time interval before subsequent upgrades and downgrades Upgrade Downgrade Independent Variables Coefficient (p-value) Coefficient (p-value) Non-US 253.51 (0.03) 53.44 (0.29) Issue characteristics Issue size 53.85 (0.07) -87.71 (0.00) Maturity 17.15 (0.67) 112.76 (0.00) Seniority 74.15 (0.26) 256.08 (0.00) Callability 39.80 (0.54) -201.11 (0.00) Puttability 686.90 (0.00) 243.02 (0.08) Sinking fund -373.97 (0.01) 418.70 (0.00) Default spread 35.67 (0.68) 111.36 (0.12) Issuer characteristics Firm size -754.02 (0.02) 1,099.96 (0.09) Leverage 553.51 (0.00) -193.80 (0.03) Profitability 2,085.56 (0.06) -81.51 (0.89) Interest coverage 0.28 (0.84) 0.04 (0.00) Capital expenditures -1,857.16 (0.00) -229.42 (0.54) Country-specific variables Emerging 92.84 (0.75) -774.79 (0.03) Civil law -1,310.03 (0.03) -93.46 (0.81) Rule of law 1,039.35 (0.01) 120.48 (0.63) Creditor rights -20.05 (0.92) -13.07 (0.93) Judicial efficiency -179.63 (0.03) -51.15 (0.56) Ex-ante self-dealing 510.43 (0.41) 455.22 (0.51) Ex-post self-dealing 535.57 (0.56) 880.65 (0.13) Anti-director rights 189.64 (0.23) -80.32 (0.26) Public enforcement -690.85 (0.11) 416.87 (0.01) Disclosure requirements -975.38 (0.20) -1,510.29 (0.00) Liability standards -780.03 (0.42) -300.58 (0.54) Investor protection -750.43 (0.65) 423.13 (0.71) Industry dummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes R-squared 29.51% 24.65% Table 7 presents the results from cross-sectional regressions with the days until upgrades and downgrades as the dependent variables. In the UPGRADE (DOWNGRADE) regression, the dependent variable is equal to the number of days between the initial rating assigned to an issue by Standard and Poor’s and the first subsequent upgrade (downgrade) conditional on the existence of a rating change. The reported results are based on OLS estimations using robust standard errors clustered by issuer. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Firm size is equal to total assets. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor rights and Judicial efficiency as defined in La Porta et al. (1998), Ex-ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.
  42. 42. 40 Table 8 Rating variation by investment and non-investment grade Investment grade Non-investment grade Independent Variables Coefficient (p-value) Coefficient (p-value) Non-US 1.34 (0.00) 0.55 (0.00) Issue characteristics Issue size 0.03 (0.44) -0.26 (0.00) Maturity 0.12 (0.00) -0.41 (0.00) Seniority -0.10 (0.68) -0.63 (0.00) Callability 0.42 (0.01) 0.96 (0.00) Puttability 0.41 (0.00) -0.03 (0.42) Sinking fund 0.19 (0.51) -0.13 (0.01) Pay-in kind 0.80 (0.00) Default spread 0.16 (0.14) -0.41 (0.00) Issuer characteristics Firm size -1.76 (0.00) -2.22 (0.00) Leverage 0.52 (0.21) 0.68 (0.00) Profitability -3.09 (0.01) -2.41 (0.00) Interest coverage -0.01 (0.03) -0.02 (0.00) Capital expenditures -1.18 (0.26) 1.02 (0.00) Country-specific variables Emerging 0.05 (0.89) -1.33 (0.39) Civil law -0.14 (0.72) 1.01 (0.16) Rule of law -0.50 (0.15) -0.85 (0.00) Creditor rights 0.11 (0.22) -0.48 (0.00) Judicial efficiency -0.01 (0.91) 0.67 (0.42) Ex-ante self-dealing -0.44 (0.26) 7.66 (0.00) Ex-post self-dealing 0.11 (0.83) 2.39 (0.54) Anti-director rights -0.17 (0.08) -2.24 (0.00) Public enforcement -0.63 (0.00) -0.38 (0.00) Disclosure requirements -0.45 (0.59) -1.21 (0.00) Liability standards -1.26 (0.01) 2.60 (0.24) Investor protection 2.47 (0.01) -2.82 (0.00) Industry dummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Pseudo R-squared 7.18% 17.34% Observations 12,484 1,250 Table 8 presents the results from cross-sectional regressions with Rating as the dependent variable. The regressions are separately run for investment and speculative grade issues. The reported results are based on ordered probit estimation using robust standard errors clustered by issuer. Rating equals one for firms that have the best credit rating (AAA) and increases by one for successively lower rating categories. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Firm size is equal to total assets. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor right and Judicial efficiency as defined in La Porta et al. (1998), Ex- ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.
  43. 43. 41 Table 9 Likelihood of missing defaults and raising false alarms Missed default False alarm Independent Variables Coefficient (p-value) Coefficient (p-value) Non-US -0.54 (0.03) 0.55 (0.01) Issue characteristics Issue size 0.33 (0.80) -0.05 (0.00) Maturity 0.07 (0.90) -0.05 (0.00) Seniority -0.52 (0.71) -0.43 (0.00) Callability -0.93 (0.32) 0.11 (0.00) Puttability -0.47 (0.01) 0.25 (0.00) Sinking fund -0.48 (0.00) -0.23 (0.00) Default spread -2.55 (0.00) 0.14 (0.00) Issuer characteristics Firm size -17.89 (0.72) -0.35 (0.00) Leverage 10.77 (0.04) 1.83 (0.00) Profitability 12.09 (0.89) -6.79 (0.00) Interest coverage 0.09 (0.85) 0.00 (0.81) Capital expenditures -29.29 (0.43) 0.64 (0.00) Country-specific variables Emerging 0.06 (0.81) Civil law -0.77 (0.01) Rule of law -0.45 (0.07) Creditor rights 0.16 (0.06) Judicial efficiency -0.03 (0.74) Ex-ante self-dealing -1.22 (0.00) Ex-post self-dealing -0.62 (0.12) Anti-director rights 0.02 (0.81) Public enforcement 0.22 (0.11) Disclosure requirements -0.98 (0.03) Liability standards -1.15 (0.00) Investor protection 1.12 (0.03) Industry dummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Table 9 presents the results from cross-sectional regressions of indicator variables for missed defaults and false alarms as the dependent variables. Missed default is an indicator variable that takes the value of one for missed defaults and zero otherwise. Specifically, for a sample of issuers that experience an event of default within one year from the rating date, this variable takes the value of one if a debt issue is investment-grade and zero otherwise. False alarm is an indicator variable that takes the value of one for false warnings and zero otherwise. Specifically, for a sample of issuers that do not experience an event of default within one year from the rating date, this variable takes the value of one if a debt issue is non-investment grade and zero otherwise. In Missed default regressions, the legal variables are dropped from the specification due to multicollinearity. Non-US equals one when debt is issued by a non-US firm and zero otherwise. Issue size is the natural logarithm of the size of the debt issue in millions of dollars. Maturity is the natural logarithm of the number of years to maturity. Seniority equals one when the debt is senior and zero otherwise. Callability equals one when the bond includes a call provision and zero otherwise. Puttability equals one when the bond includes a put provision and zero otherwise. Sinking fund equals one when the bond includes a sinking fund feature and zero otherwise. Pay-in-kind equals one when the bond pays in kind and zero otherwise. Default spread is the yield difference between AAA- and BAA-rated corporate bonds. Firm size is equal to total assets. Leverage is total debt to total assets. Interest coverage is EBIT to interest expenses. Profitability is EBITDA to total assets. Capital expenditures is capital expenditures to total assets. Emerging equals one when the issuing country is defined as being part of an emerging market as defined by Morgan Stanley Capital International. Civil law equals one when the legal origin of the issuing country is the civil law. We also include variables capturing country-specific legal characteristics. These variables are Creditor rights and Judicial efficiency as defined in La Porta et al. (1998), Ex-ante self-dealing and Ex-post self-dealing as defined in Djankov et al. (2008), Anti-directors rights as defined in Spamann (2010) and Rule of law, Public enforcement, Disclosure requirements, Liability standards and Investor protection as defined in La Porta et al. (2006). We also include industry dummies using Fama and French (1997) industry definitions and year dummies.

×