Abstract
We examined the effects of sovereign risk on bond duration in European and Latin American sovereign bond markets over the period 1996 to 2011. We compared the sovereign risk-adjusted duration with the Macaulay duration for both investment- and speculative-grade US dollar-denominated sovereign bonds. We found that the sovereign risk-adjusted duration is significantly shorter than its Macaulay counterpart for all ratings, and the ‘shortening’ effect is stronger for lower rated bonds, which generally intensified during the recent financial crisis. Results are robust when credit default swap (CDS) prices are used as a proxy for changes in sovereign risk. This study provides evidence for advocating the importance of adjusting the bond duration for sovereign risk. More important, this study provides a practical methodology for estimating a sovereign risk-adjusted duration measure for managing international bond portfolios.
Notes
1 Our data analysis shows that 5% or more of the observations are outliers or high leverage data, which are observations that have extreme values for the explanatory variables.
2 To avoid possible confounding effects of the call risk and the interactive effect of the foreign exchange risk with the interest rate risk on the duration measure, we include only noncallable bonds denominated in US dollars in our analysis.
3 This is the entire sample period. Some bonds do not have data to span the entire period.
4 Bloomberg provides bond ratings assigned by Moody’s, Standard and Poor’s and Bloomberg Composite. For our analyses, we adopt the conservative approach in grouping the bonds by the lowest rating the bond receives.
5 The daily US Treasury spot rates are obtained from the Federal Reserve System.
6 In the case that there are no two consecutive daily spreads, we calculate the change in yield spread as the spread today minus the spread in the most recent trading day. However, if the change exceeds the 99th percentile or below the 1st percentile of the changes in yield spreads of bonds issued by the same country, we replace the change in yield spread with the 99th percentile or 1st percentile.
7 We tested different time lags and found that the lagged 1-day return was most significant. Hence, we only included the lagged 1-day return in our analysis. Considering the time lags between the European countries and the US, we adjust the effect of time zone when handling the European stock return data.
8 We also estimated Equation 1 without including the contemporaneous and lagged 1-day change in CDS prices. Estimates of β1 do not have significant changes. The results show that changes in CDS prices capture the effects of other factors beyond the sovereign risk on the changes in sovereign yield spreads.