Evaluating the Influence of Audit Committee Independence on Financial Statement Credibility

Posted: Apr 18, 2023

• Lara Stone
• Ezekiel Carter
• Maisie Flores

Abstract

This research introduces a novel computational framework for assessing the relationship between audit committee independence and financial statement credibility using machine learning and natural language processing techniques. Traditional approaches in accounting research have relied primarily on statistical regression models and manual coding of corporate governance attributes. Our methodology represents a significant departure by employing transformer-based language models to analyze the semantic content of financial disclosures, coupled with graph neural networks to model the complex relational dynamics between audit committee members and corporate leadership. We developed a proprietary dataset of 1,247 publicly traded companies over a five-year period, incorporating not only conventional financial metrics but also extracted linguistic features from earnings calls, board meeting minutes, and corporate governance documents. The findings reveal that conventional binary measures of independence fail to capture the nuanced ways in which social and professional networks influence committee effectiveness. Our model identifies three distinct patterns of relational independence that correlate with financial statement reliability, with the most significant predictor being the density of pre-existing professional connections between committee members and executive management. This research contributes to both accounting practice and computational social science by demonstrating how advanced AI techniques can uncover previously obscured relationships in corporate governance structures. The framework developed herein offers regulators and investors a more sophisticated tool for evaluating governance quality beyond traditional check-box approaches.