RDP 2004-02: The Impact of Rating Changes in Australian Financial Markets Appendix B: A Cross-sectional Analysis of Bond Spreads

The results presented in Section 5 estimate the short-run impact of rating changes on bond spreads. To compare these results with the expected long-run ‘equilibrium’ impact on spreads, we estimate a simple cross-sectional relationship between spreads, credit ratings, and modified duration in the Australian corporate bond market. Our sample comprises the 466 bonds that appeared in the Merrill Lynch Australian corporate bond index at the end of each of the years from 1999 to 2003, with the deletion of a small number of obvious outliers. Because we are using data for different years, we also include year dummies to account for the variation in average spreads over time. The results appear in Table B1.

For simplicity, the spread in basis points is used as the dependent variable, but the results are similar for more complex (and arguably more appropriate) transformations of the spread. Alphabetic ratings are converted to a cardinal scale from 0 (AAA/Aaa) to 9 (BBB−/Baa3). The OLS regression includes dummy variables to control for differences between spreads in particular years, and also squared terms to account for non-linear relationships among spreads and the main independent variables. The equation has an adjusted R² of 0.52.

The ratings coefficients (in particular the squared term) imply that bonds with poorer ratings have higher spreads, and the coefficients for duration show that spreads narrow as bonds near their maturity. The median rating change in our sample of bonds from Section 3 is between A and A−, which translates into a 9 bps change using the analysis above. Additionally, the passage of 100 days (or about 0.3 years of duration), as occurs during our estimation window, is associated with a fall in spreads of around 4 bps.