Date of Award

2026

Document Type

Dissertation

Degree Name

PhD in Accountancy

Department

Department of Accountancy

First Advisor

Gopal V. Krishnan

Second Advisor

Ari Yezegel

Third Advisor

Rong Yang

Abstract

Corporate disclosures serve as a critical channel for firms to communicate their financial performance, strategic initiatives, and firm-specific risks. In recent years, rapid advances in natural language processing (NLP) have provided powerful new tools for extracting insights from the growing volume and complexity of corporate disclosures. This dissertation consists of three studies that explore how advanced textual analysis can reveal deeper insights into corporate disclosures, managerial behavior, and auditor practices.

The first paper (co-authored) develops a topic-based approach to measuring the substantive content of risk factor disclosures. Using an embedding-based topic modeling framework, we identify 30 interpretable risk topics and organize them into 11 broader categories. We then construct firm-year measures of risk disclosure breadth and specific risk categories, such as legal risk and financial reporting risk. The results show that these measures are associated with firm characteristics linked to underlying risk exposure, reporting complexity, and external monitoring, suggesting that embedding-based topic viii modeling provides a scalable way to convert risk disclosures into structured empirical measures.

The second paper (sole-authored) examines how firms differentiate their MD&A disclosures from their peers, and whether such differentiation reflects informative communication or impression management. I distinguish between two dimensions of disclosure uniqueness: content uniqueness, which captures the discussion of new or rare topics, and style uniqueness, which captures the linguistic deviation within commonly discussed topics. I find that content uniqueness is associated with more favorable market reactions, improved analyst forecast accuracy, and better liquidity, whereas style uniqueness is associated with weaker or muted market responses and greater information frictions. These findings suggest that investors respond differently to what managers disclose than to how they rhetorically frame it, and that disclosure differentiation can either reduce or exacerbate processing costs depending on its form.

The third paper (co-authored) examines whether SEC comment-letter scrutiny is associated with subsequent changes in auditor CAM disclosures. Using firm-year data from 2019 to 2024, we find that recent comment-letter receipt is associated with more CAMs, a greater likelihood of issuing new CAMs, and greater document-level revision in CAM disclosures. We also find that office-level exposure to comment letters received by other clients is associated with textual CAM revision, even when it is not associated with CAM counts or new CAM issuance. Overall, the evidence suggests that CAM disclosures are responsive to regulatory scrutiny and that direct client effects and audit-office spillovers operate through different channels.

Download

Share

COinS