This dissertation analyzes three drivers of firms’ long-term performance within the dynamic landscape of the contemporary business environment: stakeholder management, resource management and artificial intelligence (AI) tools management.
The debate concerning whether corporate social responsibility (CSR) activities are value- enhancing or value-destroying is very active in both academic and business words. In the first chapter we contribute to this academic debate by analyzing the relationship between corporate social performance (CSP) and corporate financial performance (CFP). We use the Covid-19 pandemic as an exogenous shock and use both stock returns and risk measures to examine this relationship across a sample of companies in the travel and leisure (T&L) industry, which were heavily impacted during this period. Consistent with stakeholder salience theory, our results indicate that CSP in salient CSR activities is associated with higher CFP. We also find that companies following a stakeholder salience approach outperform their peers adopting either a stakeholder approach (high-CSP for all CSR activities) or a shareholder-only approach (low- CSP for all CSR activities) and constitute the only group of companies that did not experience a significant decline in stock returns during the Covid-19 market downturn. Finally, we find that the CSP-CFP relationship has an inverted U-shape suggesting the existence of an optimal level of CSP that maximizes CFP.
This work has both theoretical and practical contributions. managers need to define effective and context-dependent CSR activities by focusing on salient CSR activities and defining the optimal CSP level. In can also be informative to regulators, NGOs and other stakeholders to understand the stakeholder-firm relationship. More specifically, it can help deduce when
pressures on managers are necessary – i.e., when it is not (or no longer) financially beneficial for firms to voluntarily address these issues – and deploy resources and measures accordingly. In the second chapter, we empirically examine the relationship between asymmetric cost adjustments (i.e., costs stickiness) and capital structure and profitability. We develop a novel measure of cost stickiness and address endogeneity issues using instrumental variables. We first find that high-sticky-cost firms have lower financial leverage, shorter debt maturity, and higher cash holdings. Our findings imply that cost stickiness increases the risk of default, reducing the optimal leverage. They also suggest that cost stickiness increases financial constraints, leading managers to favor internal financing to pay for operational excess capacity and to sustain investments when sales are low. Moreover, we find that cost stickiness has a positive effect on profitability. Finally, we compare the effects of cost stickiness and the related concept of operating leverage. We observe that, while they have similar effects on capital structure,
operating leverage has an overall negative effect on profitability unlike cost stickiness.
While extensive research has been conducted to document cost stickiness and the factors under which it is amplified, research on the effects of cost stickiness have been very limited. Therefore, by introducing an accounting topic (cost stickiness) as an important determinant of financial leverage, cash holdings, and debt maturity, we add to both the accounting literature by looking further at the financial consequences of asymmetric cost behavior and the corporate finance literature examining the firms’ operating policies affecting capital structure.
In the third chapter, we examine the algorithm appreciation phenomenon and how gender and knowledge influence the level of trust humans place in AI. AI is increasingly utilized to provide real-time assistance and recommendations across a wide range of tasks, especially since the emergence of AI Chatbots such as ChatGPT. However, it is unclear how users perceive the
trustworthiness of these tools, more so given the publicized “hallucinations” that they may experience. We conduct a randomized field experiment to analyze how subject characteristics affect trust in AI versus human peers. We randomly assign students to two experimental groups receiving advice labeled to come from an AI system (treatment group) or labeled as coming from human peers (control group). Our results are in line with recent laboratory experiments documenting algorithm appreciation. However, we find that algorithm appreciation varies with subject knowledge and gender. Specifically, both male and high-knowledge subjects place considerably less weight on AI advice. Our results remain consistent even over an extended out-of-sample period and after providing subjects with performance information.
This highlights the need to tailor AI tools to subject characteristics to significantly enhance their effectiveness and ultimately also the adoption rates. A personalized approach to AI can enhance engagement and mitigate potential adoption barriers. Further research in technology management is needed to explore various factors influencing AI trust. For example, future experimental projects, in collaboration with current co-authors explore various related questions such as: how to build and restore trust in AI; identify the main subject characteristics influencing trust in AI; understand the drivers of gender differences in AI and peer trust; and exploring task characteristics affecting trust in AI.
From a managerial perspective, understanding the circumstances in which AI assistance is beneficial and determining effective control mechanisms is crucial. While literature explores the benefits and challenges of AI, the impact of controlling AI assistance on task performance remains relatively unexplored. One of our ongoing research projects focuses on investigating the effects of restraining and controlling AI assistance on task performance, considering potential trade-offs and implications for human-AI dynamics.