The impact of rating announcements on stock returns: A nonlinear assessment

The question of whether credit rating announcements by credit rating agencies can influence the short-run performance of firms has been long debated. Although evidence points to the occurrence of abnormal returns after the announcement, most event studies do not take into account the effect of potential non-linearities between the event and stock returns. This study proposes a novel approach to assess the impact of credit announcements, based on machine learning. In the first step a statistical test is performed to evaluate whether the mean returns before and after the announcement are different, then in the second step we predict whether an actual variation of mean returns takes place. Based on an international dataset of credit events on returns from DAX, FTSE100, and NIKKEI225 indices, our results show that (i) machine learning has a slight edge over standard classification models, (ii) significant rating changes are seemingly more complex to predict.