Developing a Contextual Corporate Integrity Measure Using Large Language Models

Wirtschaftsinformatik & Management · Springer Nature · 2026-11-11 · DOI: 10.1007/s10796-026-10822-3

Аннотация

This study develops a new Contextual Corporate Integrity Score using large language models (LLMs) to measure firms’ integrity culture more accurately. We analyze the unscripted question-and-answer sections of 183,043 quarterly earnings call transcripts from 2008 to 2024, where executives respond spontaneously to analysts. Unlike traditional dictionary-based methods that rely on keyword counts, our approach interprets language in context and identifies genuine integrity-related discourse. The score captures four dimensions of integrity: Accountability (responsibility and governance), Compliance (adherence to laws and rules), Moral Integrity (fairness and ethical conduct), and Honesty (transparency, trust, and truthful communication). These components are combined into an overall integrity measure. We validate the score in two ways. First, the model shows high agreement with human judgments, indicating strong reliability. We also find that over 70% of text segments containing potential integrity-related keywords (suggested in earlier studies) do not actually discuss integrity when context is considered. Second, the score is meaningfully associated with important firm outcomes. Firms with higher integrity scores tend to have more independent boards and greater female board representation, are more likely to disclose and correct accounting issues transparently, have lower discretionary accruals, pay higher dividends, and enjoy greater access to trade credit. Overall, the findings show that contextual AI-based measures provide a more informative view of corporate culture than traditional keyword methods. The measure offers practical value for investors, boards, regulators, and researchers seeking to monitor corporate integrity and understand its effects on governance, reporting quality, and financial decisions.

Читать у издателя →

Похожие статьи