|
Article Excerpt
UNSOLICITED BANK CREDIT ratings assigned to banks by nationally recognized statistical rating organizations (NRSROs), such as Standard & Poor's Ratings Services (S&P's) and Moody's Investor Service (Moody's), are controversial. Credit ratings that are initiated and paid for by issuers are called "solicited ratings," and credit ratings that are not paid for by the issuing firm are called "unsolicited ratings." The U.S. Department of Justice (DOJ) has recommended that the U.S. Securities and Exchange Commission (SEC) require rating agencies to disclose when credit ratings are unsolicited (Gasparino 1996a). The DOJ (1998) stated that "'unsolicited' ratings may not be as accurate as ratings by retained agencies.... When unsolicited ratings are not based on the same type of information as solicited ones, the ratings agency runs the risk that its rating is not accurate." Based on the survey of Baker and Mansi (2002), there are concerns that unsolicited ratings are less accurate than ratings that are paid for in the traditional manner because the rating agency does not have access to confidential information in the traditional ratings process.
Rating agencies encounter potential conflicts of interest because they serve both issuers and investors (Baker and Mansi 2002). Investors are the main users of credit ratings, but fees paid by the issuers are the principal source of income of the agencies. For example, about 90% of Moody's and Fitch's revenues come from issuer fees (SEC 2003). Michael Oxley (2004), chairman of the House Committee on Financial Services, said in hearings about "The Ratings Game" that "officials from Northern Trust Corporation have stated that the major rating agencies have requested payment for unsolicited ratings and strong-armed the company to pay the fees in return for a good rating. Northern Trust is not the only company to register a complaint about these practices." The following year, Oxley (2005) also stated in hearings about "Reforming Credit Rating Agencies" that "given the inherent conflicts and evidence that unsolicited ratings tend to be lower, this practice begs for reform, if not outright prohibitions."
In 2005, Eliot Spitzer, New York State attorney general, subpoenaed Moody's for documents related to the company's unsolicited credit ratings and other credit-rating practices (Klein 2005, Stempel 2005). James Kaitz, president and chief executive officer of the Association for Financial Professionals, said that issuers often feel compelled to participate in the rating process and pay for the unsolicited rating. He asked the SEC to explore the potential for abuse in unsolicited ratings (McTague 2005). On the other side of the rating process, Kathleen Corbet, during her tenure as president of S&P's, defended unsolicited ratings on the grounds that they benefit the market and said that the company issued these ratings only if there was meaningful market interest and adequate public disclosure by the issuer (McTague 2005). Moody's (1999) considers the assignment of unsolicited ratings to be the market's best defense against rating shopping. Rate shopping occurs when issuers shop among various agencies for the highest ratings and to suppress lower conclusions.
Against this background, the main research issues that we examine are whether the credit ratings of unsolicited banks would be higher if they were solicited, and alternatively, whether the credit ratings of solicited banks would be lower if they were unsolicited. These issues are complicated as they must be answered by taking into account: (i) the differences in the financial characteristics of the two groups (the clientele effect), (ii) the potential self-selection bias whereby better firms may self-select to be rated and poor-quality firms may not request to be rated, and (iii) the differences in the importance of the same factors in determining the ratings of the banks between the two groups. As the next section shows, previous studies in the literature have not addressed these hypothetical questions.
We attempt to take all of the above factors into account by adopting an endogenous regime-switching model and using a relatively simple testing procedure. The two regimes are "solicited banks" and "unsolicited banks." The selection mechanism is endogenous and depends on observable firm characteristics and market performance. Specifically, we use the estimated coefficients in each regime equation of the endogenous regime-switching model. We use the estimated model to obtain the expected credit ratings of the bank in an alternative regime. The expected credit rating of the bank is calculated using the estimated coefficients of the other regime, thus utilizing the bank's own characteristics. These are counterfactual measures that cannot be observed directly. Then, we evaluate several treatment effects in hypothetical situations.
The important point is that we evaluate the hypothetical ratings after controlling for different bank characteristics and endogenous selection bias. Our approach is different from the standard Heckman's selection bias model as we use heterogeneous parameters of the rating determinants in the selection equation. We attempt to control for selection bias based on observable firm characteristics. We cannot account for the unobservable factors if firms with something to hide may select not to be rated. Further discussion of the potential limitation of our analysis is presented in Section 3.4. However, our methodology also allows us to decompose the observed difference in ratings into two different treatment effects using selected counterfactual measures and, in particular, permits us to divide the usual treatment effects into two different sources that are more meaningful in our analysis. We decompose the sources of rating difference into two components: (i) the clientele effects caused by differences in financial profile, or financial characteristics, holding the solicitation status fixed, and (ii) the treatment effects caused by a change in solicitation status, holding the financial characteristics fixed. The latter effects constitute the focus of the main questions of this study. Our results show that the observed differences between the solicited and unsolicited ratings can be explained by both financial profile and solicitation status. In some cases, the effect of solicitation status on bank ratings is larger than the effect caused by differences in financial profile. This is a new contribution to the literature.
The remainder of the paper is organized as follows. Section 1 provides background information on credit ratings and the relevant literature. Section 2 explains the research design and methodology. Section 3 discusses the results of our tests, and Section 4 presents our conclusions.
1. BACKGROUND AND RESEARCH ISSUES
Large firms that issue publicly held debt believe that credit ratings from NRSROs are indispensable for managing interest costs and attracting investors. The SEC recognizes five NRSROs: A.M. Best Company, Inc., Dominion Bond Rating Service Ltd., Fitch Inc. (Fitch), Moody's, and S&P's in 2005 (SEC 2005). These firms are also recognized internationally. Moody's began to assign unsolicited credit ratings in 1909 (Moody's Investors Service 1999). However, Stempel (2005) reported that Moody's discontinued assigning unsolicited credit ratings in 2000.
Other rating agencies continue to assign unsolicited ratings because investors want them. In addition, unsolicited ratings can be used as a strategy for entering new markets, or when there is a new asset class (McTague 2005). For example, when S&P's entered the Japanese market in recent years, it assigned 176 unsolicited long-term ratings (i.e., 63% of the 278 ratings) to Japanese issuers (S&P's 2003a). S&P's does not use the terms "solicited" and "unsolicited" in its monthly ratings publications. Rather, it labels unsolicited ratings with "pi" subscripts (e.g., [AA.sub.pi]) that stand for "public information" as opposed to other ratings. According to S&P's, ratings with "pi" subscripts are based solely on the analysis of an issuer's public information, that is, the issuer's published financial information and additional information in the public domain. They do not reflect in-depth meetings with an issuer's management. Thus, these ratings are based on less comprehensive information than ratings without "pi" subscripts (S&P's 2000b).
Some firms do not want solicited ratings by NRSROs, while others do want to be rated. Consider China, which has regulatory systems and accounting standards different from those in the United States. Chinese bond issuers that intend to offer their bonds only in China are not required to obtain ratings from NRSROs. Hence, they use local rating agencies that are recognized by their own regulatory bodies. However, those issuers that intend to raise funds in international capital markets, or that intend to cross-list their shares on foreign stock exchanges, want ratings from NRSROs to gain international acceptance.
Equally important, not all international issuers want unsolicited credit ratings by NRSROs, especially those issuers that have been assigned unfavorable ratings (Adams 1996, Japan Economic Newswire (JEN) 1998). Harington (1997) said that some banks consider Moody's practice of assigning unsolicited ratings as "tantamount to blackmail." This implies that banks would receive more favorable ratings should they cooperate and pay for them. Other issuers have complained that unsolicited ratings mislead investors because they believe that the ratings were assigned without the input of the bond issuer (Gasparino 1996b, Williams 2005). This results in an issue of information friction. Recently, the German Insurance Association expressed considerable concern about the appropriateness and transparency of Fitch's methodology in issuing unsolicited ratings (Miller 2005).
The determinants of credit ratings have been researched extensively. Altman et al. (1981) reviewed the early evolution and application of statistical techniques to bond ratings and other financial analyses. Moon and Stotsky (1993), Cantor and Packer (1997), and Pottier and Sommer (1999) found that rating scales, rating determinants, or weights attached to rating determinants differ across rating agencies after accounting for self-selection bias. Moon and Stotsky demonstrated that there is a significant difference between Moody's and S&P's in rating determinants. Their results indicate that split ratings reveal differences in both the degree of importance that is assigned to the specific determinants of the ratings and in the way that the bonds are classified.
Cantor and Packer (1997) used the long-term ratings of U.S. corporations that were assigned by Moody's, S&P's, Duff & Phelps Credit Rating Co. (Duff & Phelps), and Fitch to show that there were differences in the rating scales among these agencies. (1) They provided limited evidence of self-selection bias. After examining a sample of property-liability insurer financial strength ratings that had been assigned by Moody's, S&P's, and A.M. Best (Best), Pottier and Sommer (1999) found that rating determinants and their weights differed across the three agencies. Whereas Moon and Stotsky (1993) detected self-selection bias in Moody's ratings but not in S&P's ratings, Pottier and Sommer identified selection bias only in Best's insurer ratings but not either Moody's or S&P's ratings.
On the other hand, Poon (2003) used S&P's long-term ratings of 265 corporations in different industries in 15 countries from 1998 to 2000 and found that unsolicited credit ratings were lower than solicited ratings. Profitability and sovereign credit risk were the two major factors used to determine long-term corporate ratings. Byoun and Shin (2003) used the unsolicited and solicited ratings of non-U.S. corporations between 1996 and 2002 to study the effects of solicited and unsolicited ratings on firm value. For unsolicited ratings, they found significant negative market reactions to downgrade announcements and positive market reactions to upgrade announcements. In contrast, for solicited ratings, they found only significant positive market reaction to upgrade announcements.
Butler and Rodgers (2003) examined a sample of 360 bond ratings issued by nonfinancial companies during 1997 and they found little evidence that there were observable conflicts of interest between bond rating agencies and issuing firms. Poon and Firth (2005) used a sample of 1,060 ratings of major banks from 82 countries to analyze shadow ratings (2) that are based largely on public information to shed light on the controversy surrounding unsolicited ratings. Their results indicate that shadow ratings are lower than nonshadow ratings.
1.1 Research Issues
In general, the aforementioned studies reveal that unsolicited credit ratings are lower than solicited credit ratings. However, this is not necessarily evidence of the significance of solicitation, the blackmail hypothesis, or information friction. It is possible that low-quality banks may choose not to solicit ratings or pay for ratings. Accordingly, there is a selection bias issue. Equally important, the financial characteristics of solicited firms and those of unsolicited firms differ. Collectively, these factors could explain the difference in the average ratings, bur they do not necessarily support the existence of prejudice or frictional information. Therefore, it is important to control for selection bias and different observable firm characteristics when evaluating the net effect of solicitation.
The previous factors are important because rating agencies may evaluate financial characteristics differently when the ratings of banks are unsolicited. Hence, we examine the extent to which factors regarding rating procedures contribute to the difference in the credit ratings issued by the credit rating agencies, while at the same time controlling for the effects of different financial profiles. The main objective of our study is to examine whether solicitations matter in bank credit ratings. Thus, we begin by asking whether the credit ratings of unsolicited banks would be higher if...
|
