Motivation for Dividends of Listed Companies in China:Evidence Based on Machine Learning
Corporate dividends are not only a direct reflection of shareholders'return on investment but also serve to bolster investor confidence.It is an important means to"build an investor-oriented capital market".However,the"dividend puzzle"remains a fundamental challenge in both theory and practice,as existing theories fail to fully explain the true motivations behind dividend payouts.This paper employs a variety of machine learning methods to analyze the mechanisms of corporate dividends under the influence of multiple factors,and evaluates the predictive effect of different types of company characteristics on dividend behavior,aiming to more accurately interpret the dividend behavior of China's enterprises.The results are as follows.Firstly,the ensemble learning methods represented by the gradient boosting regression tree and the random forest are outstanding in predicting and explaining corporate dividend behavior,which significantly exceed conventional statistical models such as multiple linear regression.Secondly,among the characteristics of multiple companies,the life cycle characteristics and corporate tax rate characteristics have a better prediction effect on the dividend practice of China's capital market,followed by the first and second types of agency problems,while the financing demand and investor sentiment characteristics have a weaker prediction effect on the dividend behavior of enterprises.Thirdly,among the characteristic variables of various dividend drivers,the ratio of retained earnings to assets,the characteristics of corporate tax rate,the ratio of other receivables to assets,and the level of dividend payment in the previous period contribute the most to the prediction of dividend level.Heterogeneity analysis shows that the main factors affecting corporate dividends are significantly different before and after the promulgation of dividend policies such as semi-mandatory dividend policy,dividend tax reform,and the Cash Dividend Guidelines.There are significant differences between enterprises with different ownership types and different free cash flow conditions.At the same time,whether to adopt a dividend catering strategy and investor dividend sentiment will also affect corporate cash dividend behavior.The results show that the regulatory authorities can pay more attention to the important characteristics that affect corporate dividend payment behavior,to achieve the purpose of improving corporate cash dividend level.This paper helps investors to identify enterprises that meet their dividend expectations,thus achieving reasonable planning of holding funds and sharing the growth dividend of enterprises earlier.This paper provides new evidence for"strengthening the supervision of cash dividends and enhancing investor returns",and offers important insights for deeply understanding the dividend practices and activating China's modern capital market.
NETL
NSTL
万方数据
