Are acquirer stock price reactions to M&A announcements in any way predictable? A machine-learning analysis

We examine whether acquirer stock price reactions to M&A deal announcements can be
forecasted based on ex ante acquirer, target, deal, and macroeconomic characteristics. We
employ machine learning methodologies with out-of-sample testing and standard crossvalidation procedures to assess the forecasting accuracy of various parametric and nonparametric models. While overall predictability is low, nonparametric models exhibit some ability
to forecast acquirer stock price reactions to M&A announcements, whereas parametric models do not. Feature importance analyses reveal that a handful of predictors, including
acquirer size and (relative) deal size, contribute most to the predictions. Our findings have
practical implications for corporate managers and various corporate stakeholders