Corporate governance and top management fraud in construction companies: a China study

Abstract

Corruption in construction companies often leads to injuries and deaths. To explore the antecedents of corrupt practices in construction companies, most previous studies have emphasized the effects of operational managers while neglecting to examine the impact of top managers. As decision-makers in a firm, top managers can determine the establishment and dissolution of project teams and may force project managers to save costs excessively by cutting corners or remain silent when their subordinates engage in some unlawful practices. This dissertation investigated the role of top manager characteristics, internal governance mechanisms, and the other organizational contexts in top management fraud in construction companies at the individual and corporate levels. Though a great number of factors have been identified as contributing to the possibility of engaging in fraudulent behaviors, minimal research has focused on ranking the importance of these factors and using them to predict corporate fraud in the construction industry. Thus, this dissertation also examined the most influential factors among organizational features and constructed a prediction model. A combination of statistical methods and machine learning tools was used in this dissertation. Hierarchical linear modeling was applied to explore the drivers of an individual executive's occupational fraud due to the multilevel, specifically individual and corporate level, structure of the data. Then a hierarchical logit regression model with fixed effects was adopted in investigating the determinants of corporate fraud. Random forest (RF), a machine learning tool, was introduced for ranking the importance of factors associated with corporate fraud. This tool was also used to construct a corporate fraud prediction model. Using a multi-year sample of construction firms in China, this dissertation draws several important conclusions. First, regarding an individual executive's fraudulent behaviors, executives near retirement are associated with a lower likelihood of occupational fraud, and this likelihood is further reduced if his/her firm has a less independent board or a higher percentage of shares held by the state. Second, corporate fraud is positively affected by top management team (TMT) compensation. Aspiration-performance discrepancies have an inverted V-shaped relationship with the probability of illegal activities. The positive relationship between TMT compensation and corporate fraud is strengthened by aspiration-performance discrepancies. Third, based on the variable importance analysis of RF, the 11 most important variables associated with an increased risk of corporate illegal activities were obtained. All 11 variables relate to corporate governance, rather than financial performance. Last, RF is recommended for detecting corporate fraud in the construction industry. This dissertation facilitates our understanding of corruption in construction companies and contributes to academic theories in the fields of organization theory, strategic management, and business ethics. The used machine learning tools provide alternative ways for researchers to investigate and evaluate construction companies. The results are likely to be of interest to decision-makers including top managers, boards of directors, shareholders, investors, and relevant regulators.