Impact of bond index revisions

Abstract

This appears to be the first investigation of the impact of bond index additions and deletions on the returns of bonds and stocks of the underlying issuers using various unconditional and conditional return-generating models. The effect of additions and deletions is symmetric for each asset class, and robust across various return-generating models. While bond returns are positively (negatively) affected by bond index inclusions (exclusions), stock returns are unaffected by these bond index revisions. These results suggest that, although bond index additions and deletions materially affect bond values when measured at market, equity investors do not perceive any material change in financial risk from such changes.

Acknowledgements

Financial support from the Concordia University Research Chair in Finance, IFM2, SSHRC, SSHRC-CREF and SSQRC-CIRPÉE are gratefully acknowledged. The usual disclaimer applies.

Notes

¹ The database does not provide information about the specific indices involved in the index revisions. It simply provides a flag indicating that a bond has been added (deleted) to (from) a Lehman Brothers bond index.

² The number of stocks is lower than the total number of bonds in the samples of bond index additions and deletions because many firms have more than one bond that is subject to an index revision. For each stock, we find the corresponding bonds in both bond samples to study the effect of the revision on stock returns around the announcement date and to assign the appropriate values for the dummy variables in Equation 3.

³ The effective date on which a bond is included or excluded from the index is at most a month after the announcement date.

⁴ For example, if we are studying an event period of [−1], then t ₀ = T = −1. Similarly, if we are studying an event period of [−1, 1], then t ₀ = −1 and T = 1.

⁵ Factors in other two-factor bond models include: instantaneous yields on maturity and consol bonds, which represent the short and long ends of the yield curve, respectively (Brennan and Schwartz, 1982); real and nominal factors (Pennacchi, 1991); and a short-term factor and the stochastic volatility of an interest-rate factor (Longstaff and Schwartz, 1992). Other multifactor return models for bonds include an Arbitrage Pricing Theory (APT) (Gultekin and Rogalski, 1985; Elton et al., 1988); a relative APT model with four or six fundamental economic variables (Elton et al., 1995); and up to six bond indices (Blake et al., 1993).

⁶ The monthly dividend yields are computed using CRSP data following the approach of Campbell and Shiller (1988) and Bansal et al. (2005). They are the differences between the New York Stock Exchange/Amex and Nasdaq value-weighted returns with and without dividends. Data to compute the slope of the term structure is obtained from the Federal Reserve Board using the 3-month and 10-year constant maturity rates.