RDP 2004-02: The Impact of Rating Changes in Australian Financial Markets 2. Previous Literature

The existing empirical literature on rating agencies addresses a number of questions. For example, there is evidence that ratings are highly correlated with subsequent default rates (see Standard and Poor's (2001)), although there appears to be little evidence as to whether ratings are better predictors of default rates than other information or variables. There are also studies as to whether agencies' ratings contain information that is different from the ratings of equity analysts (see Ederington and Goh (1998)). In the case of banks, there are studies about the relative information content of agencies' ratings and supervisory ratings (e.g., Berger, Davies and Flannery (2000)). And, given the debate over the potential role of rating agencies in the new Basel Capital Accord and concern that tying capital requirements to credit ratings may be destabilising, recent studies have also examined whether ratings are procyclical (e.g., Cantor, Mahoney and Mann (2003); Amato and Furfine (2003)). Finally, there is a substantial literature looking at the market pricing of debt and equity around the time of announcements of changes in ratings. This paper focuses on this last topic, using data from the Australian bond and equity markets.

The early literature on the behaviour of debt and equity prices around rating announcements invariably used data from the US markets, although there are now also a number of studies using European data. Of those studies that look at bond returns or yields, a substantial number have failed to find significant effects from ratings, perhaps due to data problems (e.g., Gropp and Richards (2001)). Nonetheless, some studies have found evidence that bond prices or spreads do change in the expected direction around rating announcements (e.g., Hand, Holthausen and Leftwich (1992); Cantor and Packer (1996); Hite and Warga (1997); Kliger and Sarig (2000)). Although there is some variation in results across studies, there is evidence that downgrades are associated with larger market movements than upgrades, that lower-rated bonds respond more than higher-rated ones, and that the market movements around rating announcements are substantially smaller than the movements seen in the weeks or months prior to the announcements.

Each of these findings accords with expectations. If firms have incentives to release positive information about their prospects but to downplay negative information, then markets will look to third-parties such as rating agencies for objective analysis, and downgrades will be more newsworthy. And given the historical mapping between ratings and default probabilities, whereby default probabilities increase more sharply towards the lower end of the rating spectrum, it is to be expected that market prices respond more to rating changes in the latter range. Finally, if markets are reasonably efficient and rating agencies have little access to non-public information and change their ratings relatively infrequently, then it is not surprising that much of the price adjustment around rating changes occurs prior to the announcement of the change.

Due to data availability problems with bonds (which we discuss in Section 4), a larger number of studies of announcement effects have actually used equity returns rather than bond market data. The assumption implicit in most of these studies is that information that is good (bad) news for bondholders will also be good (bad) news for equityholders, so equity prices should respond in the same way as bond prices, rising following upgrades and falling following downgrades. However, as is discussed by Goh and Ederington (1993), this assumption may not always be correct. In particular, it is possible to think of circumstances where a change in a bond rating may not reflect any view about the overall profitability of a company (where rating changes have similar implications for both bondholders and equityholders), but instead may reflect decisions that are being taken by management that benefit one class of claimants at the expense of the other. While such differential impacts on the different claimants are theoretically possible, it seems unlikely that such cases constitute a particularly large subset of all rating changes, and indeed most studies proceed on the assumption that if rating announcements are relevant for equity pricing, downgrades (upgrades) will be associated with negative (positive) returns.

Indeed, the empirical literature suggests that equity prices generally respond in the ‘expected’ direction around the time of rating changes (e.g., Holthausen and Leftwich (1986); Hand et al (1992); Schweitzer, Szewcyzk and Varma (1992); Goh and Ederington (1993); Billett, Garfinkel and O'Neal (1998)). However, while many studies find significantly negative returns around downgrades, some studies find that returns around upgrades are statistically insignificant. In addition, the magnitude of announcement effects is often quite small, especially when compared with the movements in stock prices that occur prior to the rating announcements. For example, Holthausen and Leftwich (1986) find cumulative average abnormal returns in the US equity market of around −20 per cent in the 300 trading days prior to downgrades, but an announcement effect of only about −1 per cent.

To sum up, most previous evidence from foreign markets suggests that rating changes have relatively little impact on market prices of both bonds and equities and that rating decisions tend to lag earlier movements in market prices. The finding that announcement effects are relatively small suggests that market participants generally perceive that there is only limited new information in the decisions of agencies. Furthermore, the evidence that announcement effects are typically far smaller than pre-announcement market movements implies that much of the information that prompts rating changes is already reflected in market prices prior to the announcement of the change. Indeed, supporting evidence for this proposition can be found in the work of Ederington and Goh (1998) who show that most bond downgrades are preceded by declines in actual corporate earnings and in stock analysts' forecasts of earnings.

The only earlier Australian evidence on the impact of rating announcements on financial market prices appears to be the study by Matolcsy and Lianto (1995) who examine the impact of changes in ratings of Australian Ratings (acquired by S&P in 1990) over 1982–1991. The study excludes rating changes if they are accompanied by corporate announcements such as mergers, but includes rating changes that coincide with earnings announcements. The analysis of weekly stock market returns around rating announcements shows that returns in either 11-week or 25-week windows are significantly negative around downgrades and significantly positive around upgrades. However, after controlling for the information in earnings announcements in these windows, the authors suggest that returns are significantly different to zero only for ratings downgrades.

This earlier study is complementary to ours in that there is almost no overlap in the sample periods of the two studies. However, our study makes three important innovations relative to the earlier one. In particular, it examines movements in bond yields as well as equity prices; it uses daily rather than weekly data and therefore focuses on the precise day of each announcement; and, due to the growth of the Australian corporate bond market, it has a substantially larger sample size (even after excluding all events that were accompanied by value-relevant announcements).