Content area
Full text
1 Introduction
Not only is the literature on the determinants of credit spread changes not extensive but most of the published studies examine the determinants for individual bonds. Except for the study by [18] Pedrosa and Roll (1998), which examines the systematic determinants of credit spreads for credit portfolios, these studies provide little guidance for the risk management and performance evaluation of portfolios or buckets based on credit spread targeting or that cater to specific credit spread clienteles such as high-yield funds. Thus, the primary purpose of this paper is to investigate the explanatory power of credit spread changes and their determinants for portfolios using regression models whose determinants include a new determinant (portfolio diversification) and are measured using expectations (and realizations) for some previously identified determinants, such as gross domestic product (GDP).
[3] Avramov et al. (2007) calculate the yield spreads on individual bonds by subtracting the treasury (risk-free) yield for the same maturity from the bond's yield, where the term structure of risk-free yields are obtained by linear interpolation using the benchmark treasury yields for maturities of three, five, seven, ten, and 30 years. In contrast, based on the arguments in [11] Elton et al. (2001)[1] , our credit spreads are obtained as the difference between the term structures of zero spot rates on corporates and governments (where the later are derived from all treasury bonds in the database) and not on the yields to maturity for individual bonds. Furthermore, we only eliminate bonds with maturities less than one year (as in [11] Elton et al. , 2001) unlike [3] Avramov et al. (2007) who eliminate bonds with maturities of less than four years.
This paper makes four contributions to the literature. The first contribution is the use of a new variable (the undiversified risk of bond portfolios) that is shown empirically to be a significant determinant of changes in credit spreads, as is expected theoretically. The second contribution is the use of ex ante estimates instead of ex post realizations for some of the potential determinants of credit spread changes. As expected, this approach improves the explanatory power of the chosen determinants since term structures reflect future expectations, while the subsequently reviewed literature uses realizations for these variables. The third contribution...