运筹与管理2024，Vol.33Issue(12)：224-231,8.DOI:10.12005/orms.2024.0409
高管治理视角下的企业绩效预测研究
Enterprise Performance Prediction from Perspective of Executive Governance:Empirical Evidence from Machine Learning
摘要
Abstract
The shareholder governance and board of directors in corporate governance emphasize that the manage-ment's self-interest behavior is constrained from the perspective of supervision,while the executive governance reflects the endogenous driving force of the management's due diligence from the perspective of"autonomy".This paper focuses on the study of enterprise performance under executive governance from the perspective of predictive modeling.Most extant studies have focused on causal inference based on explanatory models and are limited to the impact of one-dimensional factors due to potential multicollinearity problems.However,the endog-enous issue is still difficult to be properly addressed.In order to investigate the predictive ability of multi-dimen-sional executive governance factors on corporate performance(including financial performance and market performance),this paper involves pooling a cross-sectional sample of non-financial listed Chinese A-share companies in the Shanghai and Shenzhen exchanges from 2011 to 2020,and conducts an empirical study by establishing a supervised learning prediction model-eXtreme Gradient Boosting tree(XGBoost).The purpose is to answer the following three research questions:(1)Can the characteristics of executive governance improve the prediction of corporate performance?(2)Which executive governance factors have more obvious prediction effect on enterprise performance?(3)How do these important executive governance factors affect corporate performance forecast? Firstly,the XGBoost algorithm is used to develop predictive corporate performance models based on factors related to the firm's underlying and executive governance.Different from the explanatory modelling in previous works,this study examines the predictive accuracy of adding executive governance variables to firm-specific underlying variables,in order to answer the question,"can executive governance help predict corporate perform-ance?".Since this paper measures corporate performance continuously,two commonly-used error measure indicators are used to evaluate forecasting outcomes:mean squared error(MSE)and goodness of fit(R2).In addition,the Ordinary Least Squares(OLS)regression model is used as a benchmark for evaluating the XGBoost algorithm.The empirical results indicate that the addition of executive governance characteristics has significantly enhanced the prediction of corporate performance,and the XGBoost algorithm outperforms the OLS regression model for both in-sample and out-of-sample observations. Then,based on the prediction results of the XGBoost model,we calculate the relative importance of the input features to identify the variables that contribute significantly to the prediction results,namely key features.In addition,we also describe the prediction effect of key variables under different values by drawing partial dependence diagram to reveal its prediction mechanism.This part aims to answer"which executive governance characteristics have more obvious effects on the prediction of corporate performance?"and"how do key executive governance characteristics affect the corporate performance forecast?".It is found that management capability and executive compensation incentives contribute more to corporate financial performance,whereas executive compensation and equity incentives have a relatively strong prediction contribution to corporate market perform-ance,which is basically consistent with existing explanatory studies.Additionally,the partial dependence dia-grams reveal that there is an obvious nonlinear relationship between the above senior management governance fac-tors and corporate performance,which is mainly characterized by periodic changes. Lastly,a series of empirical tests are conducted,including adjusting the sliding window length,replacing corporate financial and market performance measures,and replacing XGBoost with alternative machine learning algorithms,to validate the robustness of our conclusions.The main findings are generally supported by the results,suggesting that predictive modelling offers a new perspective on theories of reliability testing. In summary,this paper introduces a machine learning approach into the existing literature on the relation-ship between executive governance and corporate performance,which improves the limitations of traditional analysis on the one hand and provides methodological reference and inspiration for future research on the other hand,and enriches the literature on the motivation of corporate performance research and the economic conse-quences of executive governance.The research findings provide practical significance for developing managerial abilities and designing incentive mechanisms in corporate governance.关键词
高管治理/企业绩效/机器学习/XGBoost/预测Key words
executive governance/corporate performance/machine learning/XGBoost/prediction分类
管理科学引用本文复制引用
牛红丽,徐坤亮,熊梦圆..高管治理视角下的企业绩效预测研究[J].运筹与管理,2024,33(12):224-231,8.基金项目
教育部人文社会科学研究规划基金项目(23YJAZH102) （23YJAZH102）
