This thesis introduces statistical matching properly to the research domain of takeover likelihood modelling. Matching is a data pre-processing method aimed at improving causal inferences. Matching was employed to investigate two interrelated areas within the domain, where prior literature has reported varying and inconsistent findings: (1) the determinants of the target firm’s takeover likelihood and (2) predictive capability of the takeover models. To investigate the determinants of takeover likelihood, several logit regression models were developed using a training sample of 23 096 firm-year observations on publicly listed US firms between 1999-2013. Predictive capability for the models was measured in an out-of-sample test covering the period between 2014-2018. The findings from the explanatory analysis showed that inefficient management, firm undervaluation, smaller firm size, available free cash flow, lower sales growth and higher leverage increase takeover likelihood, while share purchase activity decreases it. The predictive power was considered low with the most accurate model reporting precision of 1,73% and accuracy of 66,81%. Models using matching consistently reported superior explanatory power compared to the benchmark of no matching. On the contrary, matching had a neutral impact on predictive power. Inconsistency in the explanatory and predictive results of matching suggests a separation between explanatory and predictive analysis of takeover likelihood in terms of methodology. Matching is recommended for understanding the constituents of the target firm’s takeover likelihood, but alternative methodology might be superior for predicting future targets.
|Educations||MSc in Finance and Strategic Management, (Graduate Programme) Final Thesis|
|Number of pages||121|