RDP 2004-02: The Impact of Rating Changes in Australian Financial Markets 4. Data

Adam Creighton, Luke Gower and Anthony Richards

March 2004

We analyse the impact of credit rating changes on Australian issuers by Moody's and Standard and Poor's, the two largest agencies in the Australian market.[8] The sample covers changes in credit ratings between January 1990 and July 2003, with the sample limited to those ratings announcements that were not accompanied by other value-relevant announcements.[9] These ratings are all ‘solicited’, being requested by the issuer, rather than being assigned by the agencies based only on public information. We treat simultaneous rating changes by both agencies as one event.[10] To increase the size of our sample, our ratings events include both announcements of actual changes in ratings and announcements that an agency is actively considering a near-term rating change.[11] We refer to the latter as ‘watch’ events, and treat them as either upgrades or downgrades based on the direction of the likely change indicated by the agency.[12] We refer to actual changes as either ‘anticipated’ if they are preceded by a watch, and ‘unanticipated’ if the rating was changed with no prior warning.

The sample of events for our analysis of bond spreads is substantially constrained by data problems. One problem is that many listed companies do not issue bonds. More importantly, since corporate bonds are traded much less actively than equities and are not traded in a centralised exchange, it is difficult to get good daily data for bond prices and spreads. Indeed, the investment banks that calculate corporate bond indices are essentially the only source for daily bond data, and these banks tend to focus only on a subset of bonds which meet certain criteria regarding size and liquidity. Accordingly, for the event study using bond yields, our sample of rating changes is limited to those companies for which we could obtain daily data for at least one bond from either UBS Australia or Merrill Lynch, two providers of corporate bond indices. In principle, it would have been possible to get price data for a wider sample of bonds from Bloomberg, but many of these prices appear to be matrix-based, rather than actual indicative price quotes as are used in the bond indices.

Our sample of events for which we have both rating changes and yield spread data includes 33 announcements consisting of 21 downgrades and 12 upgrades. The bonds in question are all issued in Australia and denominated in Australian dollars. Most have maturities between two and three years, and spreads are calculated relative to a Commonwealth Government security of similar maturity. Our upgraded bonds have a median initial rating of A−/A3, and the day before the rating announcement have a median spread of 67 basis points. Downgraded bonds have a slightly higher median initial rating of A/A2, but a higher median initial spread of 87 basis points.

In the case of equities, the number of events is much larger because shares are more actively traded and price information is readily available.[13] The sample includes 141 rating announcements for which we have accompanying equity returns and which are not ruled out because they occur at the same time as a corporate announcement. Of these 141 events, 33 are rating watches and 108 are announcements of actual rating changes, 23 of which are more than one ‘notch’ in magnitude.[14] Of the 108 rating changes, 48 were preceded by a watch and so are ‘anticipated’ according to our definition, and the other 93 events (including watches) are therefore ‘unanticipated’.[15] Downgrades (including negative watches) account for 95 of the events and upgrades (including positive watches) account for 46 events. The 141 events involve 62 different firms from 15 industry groups, and their industry concentration is broadly representative of the composition of the All Ordinaries Index. Reflecting the recent growth in the corporate bond market and the wider use of bond ratings, our sample is dominated by more recent events, and the second half of the sample period contains about three-quarters of all the events. Finally, rating changes by S&P account for 100 of the equity events, broadly reflecting that agency's greater presence in the Australian market. Appendix A gives more detail for both bond and equity events.

Footnotes

All data on ratings are taken from Bloomberg. In the case of bonds we look at the specific rating that applies to the particular bond for which we have data. In the case of equities, we look at a general issuer rating, either the long-term local currency rating for S&P, or the issuer rating or senior unsubordinated debt rating for Moody's. [8]

We use Bloomberg to check for other news about companies, which includes announcements concerning earnings, mergers and divestments. [9]

Cases of multiple announcements that occur within 10 days of each other are treated as only a single event. [10]

In cases where companies are removed from credit watch without a subsequent rating change, the removal of the watch was not considered an event in our sample. These cases are relatively rare and we were more uncomfortable about classifying them as being equivalent to the other upgrades or downgrades. [11]

Information from the rating agencies suggests that ‘watch’ announcements are followed more than two-thirds of the time by actual rating changes within a few months. Accordingly, they are a much stronger expression of an agency's views than an ‘outlook’ which is simply an indication of the possible direction of movement in the rating over a much longer horizon. [12]

Equity returns are based on the indices for total returns (prices plus dividends) constructed by Thomson Financial. [13]

Broad ratings categories are typically divided into three subcategories or ‘notches’ (the exception is AAA/Aaa). [14]

The reason why we have fewer (33) rating watches than rating changes that were preceded by a watch (48) is that many announcements that the rating is on watch occur at the same time as other corporate announcements and are excluded for that reason. [15]