Empirical Corporate Finance and Deep Learning

This PhD thesis comprises four empirical essays on financial stability, corporate
board diversity, and executive communication. The first essay finds that greater
heterogeneity among U.S. bank holding companies—across business models,
funding sources, and balance sheets—reduces systemic risk and enhances
resilience to financial shocks. The second study examines the long-term effects of
board gender quotas and public childcare in Europe, revealing that board diversity is associated with non-negative firm performance, particularly in countries with supportive childcare policies. The third evaluates deep learning models for extracting vocal cues from CEO speech, finding that spontaneous responses, analyzed using Wave2Vec 2.0, best predict stock returns. Building on this, the fourth essay integrates vocal, textual, and financial data to forecast excess returns, uncovering that the CEO's vocal tone also captures broader macroeconomic sentiment.