Date of Award

Summer 8-22-2026

Document Type

Dissertation

Degree Name

Ph.D. in Business

Organizational Unit

Daniels College of Business

First Advisor

Jack Strauss

Second Advisor

Arkadiy Lenchak

Third Advisor

Alex Petkevich

Copyright Statement / License for Reuse

All Rights Reserved
All Rights Reserved.

Keywords

Artificial intelligence discourse, Corporate culture, Earnings calls, FinBERT sentiment analysis, Firm valuation, Natural language processing

Abstract

This dissertation develops and tests a multi-dimensional framework for measuring corporate culture from S&P 500 earnings call transcripts (2010–2025). Extending Li et al. (2021)’s frequency-based approach, we ask whether how firms discuss culture values, not just how often, carries distinct information for firm outcomes.

Paper 1 introduces sentiment as a second dimension of culture measurement. Using FinBERT to classify the tone of culture-tagged sentences across five values (Innovation, Integrity, Quality, Respect, Teamwork), we show that semantic salience (word frequency) and sentiment (positive/negative framing) are weakly correlated (r < 0.22), confirming they measure different constructs. In joint regressions with firm and quarter fixed effects, sentiment survives alongside semantic for return volatility (4/5 values), gross margin (3/5), and revenue growth (4/5). However, a tone benchmark test reveals that culture-specific sentiment traces largely to overall call-level optimism rather than culture-specific tone, establishing a scope condition on the second dimension’s independent contribution.

Paper 2 extends the framework by adding AI Talk as a third dimension, a measure of AI discourse intensity from 38 concept groups. AI Talk independently predicts Tobin’s Q (B̂ = 0.20, p < 0.01), survives all culture and tone controls, and is robust out of sample. Firms in the highest AI Talk group exhibit Tobin’s Q nearly twice that of non-discussing firms, with book-to-market ratios 61% lower, confirming the market prices AI discourse as a growth signal. After the ChatGPT shock (2022Q4), the AI Talk premium halves as discourse becomes ubiquitous, consistent with signal dilution. AI Talk and culture sentiment operate as substitutes: firms emphasizing both gain no incremental valuation benefit.

Together, the two papers establish that corporate culture measurement benefits from moving beyond word counts to incorporate tone, but the economically meaningful signal lies in what firms choose to discuss (AI investment versus culture values) rather than in the sentiment polarity of culture language itself.

Copyright Date

8-2026

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.

Rights Holder

Cody Sullivan

Provenance

Received from ProQuest

File Format

application/pdf

Language

English (eng)

Extent

184 pgs

File Size

2 MB

Download

Available for download on Sunday, September 24, 2028



Share

COinS