中国企业对海外上市企业并购研究 ——基于并购成功概率模型的分析-research on chinese enterprises mergers and acquisitions of overseas listed companies —— analysis based on the probability model of mergers and acquisitions.docxVIP

万方数据 万方数据 摘要 摘 要 2008 年金融危机对世界各国的经济都造成了强烈的冲击，这就给中国企业实 行“走出去”战略，进行海外并购提供了重要的契机。因此，在中国企业进行跨 国并购的事件中，如何合理、准确地运作一次并购套利的金融行为，是一个复 杂的问题，而企业的并购成功概率则是制定并购套利的重要参考依据。 为此本论文首先分析了 10 种影响并购结果的风险因素，然后运用 K 均值 聚类、Adaboost 分类、随机森林分类三种统计学数据分析方法，结合 200 个中 国企业并购海外上市企业的案例，构建中国企业海外并购成功概率模型，对并 购成功的概率进行预测和分析，筛选出影响企业并购成功概率的关键风险因 素。 根据本文的分析，基于 Adaboost 分类的并购成功概率模型表现最为良好， 不仅模型在训练集和测试集中的表现并无显著的差异，可靠性较强，能够很好 的预测企业的并购结果；另外，对于合并失败的样本，基于 Adaboost 分类的并 购成功概率模型在识别能力上大幅优于其他两种模型，能够较好的控制投资中 的风险。同时基于 Adaboost 分类的并购成功概率模型筛选出的影响企业并购成 功概率的 3 个关键风险因素是：目标方所在行业、并购企业所在大洲、目标方 管理层态度，这三个因素对并购结果的影响大幅高于其他因素。 当然，本文由于时间所限和作者能力所及，不足之处和有待进一步研究的 问题也有很多。比如由于中国企业并购海外上市企业还处在初级阶段，现存的 资料很少，建立的并购成功概率模型必然还需要随着实践的深化不断修正；收 集的样本数据较少，数据收集依赖于企业披露的信息，而且是间接的收集数据 等问题。 关键词：跨国并购；成功概率；风险因素 I Abstract Abstract Abstract The 2008 world economic crisis had a big impact on many countries on the world, which led to Chinese companies adopting a strategy of go out into the world, in order to capitalize on the opportunities that arose in the realm of international mergers and acquisitions. How to correctly capitalize on this situation, through merger arbitrage opportunities, is a very complicated question. At the heart of this issue is the probability of success or failure for any given cross border acquisition by a Chinese company. This paper analyzes 10 types of risk factors related to the success probability of Chinese cross border acquisitions. Afterwards, we uitilize K mean clustering, Adaboost clustering, and random forest classification on 200 sample Chinese cross border acquisition cases in order to build a success probability model. From this analysis we are able to identify the most important risk factors for Chinese cross border acquisition success or failure. According to the analysis performed in this paper, the Adaboost clustering acquisition success rate model is the most accurate, not only does the model show very little deviance in results between cases, but also it is able to accurately predict the probability of a Chinese cros