تاثیر نوسانات بازار سهام، سود اعتباری اوراق قرضه شرکتی

There has been a rapid increase in the number of corporate bonds issued in Australia since the middle of 1998. This increase has stimulated interest in characterising the yield curves and the factors that determine changes in these spreads. The focus of this paper is on measuring any impact of stock market volatility on spreads using two different measures. One measure is based on volatility implied from options prices while the other is derived from a conditional heteroscedastic volatility model of changes in a stock market index. It is found that the former has no significant impact on spreads but the latter is both significant and stable over time. This impact is estimated to be negative implying that an increase in volatility cause a decrease in corporate bond spreads.

It is only since the middle of 1998 that the number of corporate bonds on issue has been sufficiently large to enable an econometric characterisation of credit spreads to be undertaken across a broad range of credit ratings and industries. In a previous line of research [1], fitted yield curves to daily data on spreads to Commonwealth Government Securities for nine (Standard and Poor’s) credit ratings categories (AAA–BBB) and five industry categories. The model constructed was published under the name CBASpectrum, and was estimated separately for each day in an ongoing basis so that the time series of estimated spreads that flow from this model can reasonably be used as data input for this research. Appropriate time series methods can then be used to gain insight into how spreads evolve over time and, in particular, into how they react to market forces external to the corporate bond market. Since the family of estimated yield curves is characterised by 12 parameters, one strategy is to model the evolution of the estimated parameters. Indeed, such an approach implies that the full system could be modelled by a finite number of equations, whereas the direct modelling of spreads, as is done in this paper, requires that an arbitrary finite subset of an infinite number of ratings/duration combinations be considered. In the first instance, the direct modelling approach has been followed as it has far greater intuitive appeal and can reasonably limited to modelling a smaller number of ratings/duration combinations than the number of parameters in the system. While credit spreads can be constructed from actual data, such an approach would suffer from the consequent evolution in duration of the selected bonds over time and any particular changes in the characteristics of these bonds. These problems are removed by implying a credit spread for a generic bond of fixed duration and rating over time. The ‘bonds’ chosen for this study are AAA, AA, A and BBB, each of a constant 5-year duration, and the spreads are measured as deviations from the estimated Commonwealth Government Securities yield curves using CBASpectrum. Two possible external influences on the corporate bond market are considered in this paper: the stock market, as measured by the ASX Large Capitalisation Index, and implied volatility. While 425 daily observations are used in the analysis of bond spreads: 1 July 1998 to 16 March 2001, 800 daily observations are used to measure volatility so as to remove any end-point effects due to the estimation procedure.

The credit spread is the additional yield that is priced into a bond as a compensation for various risks (default, credit migration) and for illiquidity. It has been found in this paper that implied stock market volatility, derived from options prices, has no significant impact on these spreads on a day-to-day basis. On the other hand, an instantaneous measure of market volatility, derived from a GARCH model, has a significant negative impact on these spreads. That is, an increase in market volatility causes a narrowing of spreads to Commonwealth Government Securities. A number of diagnostic procedures were erformed on the model which established the robustness of the model to outliers and structural change over time. However, it was found that an one-sided model that only passed on increases in volatility to spreads provides more realistic forecasts