Research On The Influence Of Corporate And Managerial Individual Characteristics On Corporate Violations

Corporate violations of listed companies occur frequently.On the one hand,they will damage the interest of stakeholders and attack investors’confidence.On the other hand,they will reduce the efficiency of resource allocation in the capital market and seriously damage the foundation of market integrity.Therefore,corporate violations have been under the key supervision of government regulators such as the CSRC,and continue to be highly concerned by multiple stakeholders.Effective prevention and control of enterprise violations is of great significance for maintaining the order of the capital market.Existing studies have focused on the factors that affect the company’s internal and external environment on violations.Most of research explored the relations between a single feature at the executive level or the company level and corporate violations.There is little comprehensive analysis or comparative analysis of the influencing factors at all levels on the violations.At the same time,most of research used traditional research methods to theoretically deduce and empirically test the causal relations between characteristics and violations,which provides sufficient theoretical basis for building predictive models through machine learning.With the increasing development of financial technology,new research tools encourage research in the field of corporate governance to explore broader ideas.As a novel research method,machine learning provides a new perspective for constructing models and identifying factors.Therefore,this paper attempts to establish a connection between machine learning method and the research of enterprise violation.Based on Principal-agent Theory,Upper Echelons Theory,and Managerial Power Theory,this paper introduces LightGBM and SHAP to explore the detection of corporate characteristics and managerial individual characteristics on corporate violations in Chinese listed companies from 2008 to 2019.We verify the predictive effect of corporate and managerial individual characteristics on corporate violations,compare the two types of characteristics,and further obtains the importance ranking of features and prediction mode of each feature.The evidence shows that:1)To a certain extent,the machine learning model constructed by corporate characteristics and managerial individual characteristics can predict the violations.2)Corporate characteristics have a greater impact on violations than managerial individual characteristics.Among them,the higher the transparency of listed companies,the greater the ROA,the lower the asset liability ratio,the higher the attention of analysts and the higher the proportion of senior executives,the lower the tendency of being predicted as a violating company.At the level of managerial individual characteristics,the higher the shareholding ratio of executives at the end of the year,the younger the age,and the combination of chairman and CEO,the higher the tendency of being predicted as a violating company.3)The relations between most of predictors and characteristic are non-linear,consistent with the prior literature.This paper uses machine learning to build the models,based on the characteristics of companies and executives,to explore the predictive efficiency and influence of internal governance factors on predicting enterprise violations.It effectively expands the traditional empirical research methods and enriches the evdenices of corporate governance.At the same time,from the perspective of prediction,this paper discusses the violations of listed companies,which is different from the traditional research on the causal relations of influencing factors.It also provides new ideas and empirical reference for relevant stakeholders to predict the violations.