The literature has identified various motives for mergers. Studies point out to differential merger announcement returns depending on the type of merging firms. In general, acquirers do worse than targets and horizontal mergers do best- vertical mergers do better in non-competitive environments. Our study tries, for the first time, to correlate announced merger motives with outcomes.  We use textual analysis, supervised machine learning techniques, to dissect public merger announcement. This advanced methodology enables us to develop a new text-classification framework and by doing so, we will also contribute to the research community. Using this innovative framework, researchers will be able to perform quantitative research based on textual analysis without having to deal with most of the complicated and time-demanding technical issues. This contribution will lead more researchers to use automatic textual analysis as one of their main research tools. After classifying the announced motives in different ways, we examine the predictive validity of these motives using short term announcement returns as well as looking at the post-merger performance and analyzing the sources of the gains.

The preliminary results, focusing on motive classification, seem very encouraging. The analysis of the sample was done in two stages: the first stage was done manually (without any programming) and was very complicated and time-demanding, and the second stage used the files from the first stage using an innovative and facilitating text-classification framework. The first stage identified the fragments share value, synergy, complementary, market share, cost reduction, and taxes. The second stage, using the new framework managed to get over 70% F1 accuracy in predicting the top motives (share value and market share) and got reasonable results in predicting a few more motives (synergy and complementary). In the research we will improve these results and aim to achieve over 85% F1 accuracy for most motives.

 

Key words: M&As announcements, Motive, Performance, Machine-learning, Textual analysis