Abstract
The digital age has generated vast amounts of data that continue to grow exponentially, and this “big data” revolution is reshaping the financial industry. Complementing text data, a type of big data, textual analysis, and artificial intelligence models have clearly emerged as rather prevalent tools in the empirical study of financial researchers. As the applications of textual analysis continue to expand, it could provide financial researchers with methods to measure relevant economic variables that have traditionally been challenging or even impossible to capture through traditional quantitative data approaches. These new data and methods lead to improvements in research methods and paradigms and enable the exploration of more important research topics.My thesis consists of three chapters that aim to explore significant questions using big data, textual analysis, and artificial intelligence methods. I look forward to contributing theoretically and empirically to academic research in important fields such as culture, innovation, and ESG. I aspire to study the issues that traditional data and methods cannot address through broader information sources and more diverse analytical methods.
The first chapter of my thesis examines how a stronger corporate culture can provide resilience for enterprises exposed to the Sino-US trade war. Using textual analysis tools such as word embedding models and K-Means, we measure the enterprises' exposure to Sino-US trade war risks and the strength of their corporate culture based on analysts' reports and annual reports' text data. It was found that enterprises with high exposure to the Sino-US trade war suffered more negative impacts after the war began, and a strong corporate culture mitigated such negative impacts by enhancing operating performance and relieving financial constraints.
The second chapter explores whether political uncertainty affects enterprises' engagement in rapidly evolving technological innovation. A firm can choose to create patents that contribute incremental advancements over existing technologies with limited impact on technological progress or create patents with technologies in their ascending phase, which often lead to widespread dissemination in subsequent development phases, evolution, and refinement. And it is not necessarily related to indicators such as the number of patents and citations. The number of patent applications and citations cannot fairly measure this difference. Using patent text data, we employed textual analysis methods to measure the positioning of a given patent and firms within technology cycles. We also use textual analysis methods and annual report data to measure the firm’s exposure to political uncertainty. Our findings indicate that enterprises facing higher political uncertainty are less inclined to engage in rapidly evolving technological innovation through increasing financing constraints and executive risk aversion. Additionally, we verified the causal relationship through two exogenous shock events: the Sino-U.S. trade war and the COVID-19 pandemic.
The third chapter investigates whether the chairman and CEO's masculinity femininity culture affect an enterprise's ESG performance. According to Geert Hofstede’s theory, culture acts as a mental program and software of the mind, exerting a subtle and profound influence on people's behaviors and decisions. As a pair of cultural values, masculinity shows features such as assertive, aggressive, achievement-oriented, and confidence, and femininity exhibits features such as nurturance, modesty, caring for others, and focusing on relationships. These differences result in different choices of ESG engagement made by chairpersons and CEOs with different cultural values. To address the limitations of traditional methods such as questionnaires, this paper introduces three artificial intelligence models trained on vast textual data: the word embedding model, ERINE Bot, and ChatGPT to measure masculine-feminine cultural values in various provincial regions in mainland China. We discovered that when the chairman and CEO of a firm come from a masculine region, the enterprise's ESG performance tends to be worse, and vice versa. We verified the causal relationship of this association through the abnormal turnover of the chairman or CEO.
In conclusion, although the study areas are different, the three chapters of my thesis all use big data, textual analysis, and artificial intelligence models to investigate new and significant questions. We contribute to these fields by adding theoretical insights and empirical evidence.
| Date of Award | Nov 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Xiuping Hua (Supervisor), Qingfeng Wang (Supervisor) & Fumin Zhu (Supervisor) |