This dissertation consists of two essays. The first essay examines the interaction effect of human capital investment in firms with dual-class... Show moreThis dissertation consists of two essays. The first essay examines the interaction effect of human capital investment in firms with dual-class shares (DCS) structure. In this study, I find that although more input in human capital, measured by employee welfare index (EWI), can enhance the valuation of single-class (SCS) firms, human capital investment in DCS firms is not valued by the market, but even hurts firm value. This result is consistent with the prediction of agency theory. The management entrenchment effect in DCS firms causes valuation discount when managers can transfer private benefit through investing in humans. To get a robust result, I use propensity score matched data of SCS and DCS firms and get the same conclusion. Overall, my paper provides the evidence that human capital investment plays a different role in firm value under different circumstances, especially under different ownership structures.The second essay examines the relationship between director network centrality and firm credit risk. By using a comprehensive data including both rated firms and unrated firms, I discover that director network is positively associated with firm’s probability of default. This positive effect is more robust in firms without agency credit ratings. I further examine that when firm’s cash flow increases, firm’s default risk increases with director network. But when investment increases and firm’s debt finance increases, the default risk decreases with director network. These combined results imply that director network leads to more agency problems when firms have plenty of cash flow but benefit firms when the cash flow goes into investment or when directors utilize their network to get more debt finance for firms. Also, I find that director network helps loss firms other than profitable firms and decreases firm default risk during the financial crisis. Show less
