Back to search

University of South Florida

Essays on Machine Learning and Corporate Disclosure

Abstract

dc:description

This dissertation contains two essays that shed new light on the dynamic of information disclosure on the capital market. For both two essays I leverage computational linguistics and 20 million turns of dialogues from earnings conference calls from 2006-2022. In the first essay, I introduce a novel measure that quantifies the disparity in narrative focus between managers’ disclosures and analysts’ questions during these calls, denoted as Topic Attention Divergence (TAD). A higher level of TAD indicates a higher level of firm-investor asymmetry and lower information quality. My results confirm that higher TAD in earnings conference calls inversely (positively) predict firms’ future stock liquidity (cost of equity capital). Further, the predictive power of TAD is more pronounced in firms with higher information processing costs characterized by smaller firm size, lower levels of analyst coverage, and lower institutional ownership. A long-short portfolio sorted by TAD earns an annual 8.18% risk-adjusted return and cannot be explained by existing factor models. In my second essay, similarly, I introduce TAD2, which quantifies the divergence in narrative focus between the firm and its industry average. I argue this measure serves as a signal of the uncertainty of the firms’ information environment. Firms with higher TAD2 are inversely related to future stock liquidity and positively related to investors’ disagreement, and the cost of equity capital suggesting that investors require a higher equity premium to compensate for the uncertain information environment of a firm. Furthermore, the predictability of TAD2 on future returns and liquidity is more pronounced in firms with lower institutional ownership and lower industry competition. Finally, I find that analysts systematically overestimate future earnings when firms pay greater attention to unique information topics in earnings calls (i.e., higher TAD2). Overall, my results support the bounded rationality hypothesis (Hong and Stein, 1999).

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy (Ph.D.)
Level thesis:degree_level
Dissertation
Discipline thesis:degree_discipline
Finance
Grantor
University of South Florida
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Xiao, Zicheng (Leo)
Contributors dc:contributor
  • Daniel Bradley, Ph.D.
  • Jared Williams, Ph.D.
  • Christos Pantzalis, Ph.D.
  • Ninon Sutton, Ph.D.
  • Erwin Danneels, Ph.D.

Subjects

dc:subject × 8

Identifiers

dc:identifier.*
Repository record dc:identifier
https://digitalcommons.usf.edu/etd/10583

Chain of custody

source
Harvested from
University of South Florida
Base URL
digitalcommons.usf.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Xiao, Zicheng (Leo). Essays on Machine Learning and Corporate Disclosure. Dissertation thesis, University of South Florida, 2024. https://digitalcommons.usf.edu/etd/10583