Enhancing Decision Quality and Transparency via Machine Learning-Based Goodwill Impairment Estimation in Banks

Abstract

Goodwill impairment assessment remains a judgment-intensive process in banking institutions, where managerial discretion, information asymmetry, and regulatory complexity often challenge the quality of decisions and transparency. While prior studies have widely applied machine learning to financial risk assessment and credit analytics, they have paid limited attention to its role in improving managerial accountability in goodwill impairment decisions. This study aims to address this gap by developing and evaluating a machine-learning–based estimation framework to enhance the quality of decisions and transparency in bank-level goodwill impairment assessments. Using simulation-based analysis on synthetic financial statements, the proposed framework evaluates the performance of impairment estimation using quantitative metrics that capture predictive accuracy, decision consistency, and traceability. The findings demonstrate that ML-assisted estimation can systematically improve decision quality while strengthening transparency and accountability compared to traditional judgment-driven approaches. Beyond technical performance, the results indicate that machine learning can function as a governance-supporting mechanism by enabling more traceable and internally auditable impairment decisions. The study contributes theoretically by operationalizing transparency and accountability as measurable decision outcomes in corporate finance, and practically by offering banks a simulation-based tool for internal evaluation that does not rely on field experiments or sensitive proprietary data. Overall, the research highlights the potential of ML-enabled decision support systems to enhance both the quality and governance of goodwill impairment practices in the banking sector.