Abstract:
A reverse event study approach is used to investigate how the South African sovereign bond yield curve react to headline news. Abnormal return dates in the zero-coupon yields are identified using GARCH models on the daily return series and news items that are classified into categories using supervised machine learning. A regression model is fitted to determine the link between the abnormal daily returns and news categories. The results indicate that for abnormal increases in returns, indicating an increase in yield (negative news) the entire yield curve was impacted by political news and the medium term (5-year) was also impacted by international news. For abnormal decreases in returns, indicating a decrease in yields (positive news) political news had the greatest impact on the long end (15-and 20-year) of the yield curve, and economic news had the greatest impact on the medium term (10-year).
