Earnings Management Behavior and Financial Reporting Credibility in Emerging Economies

Abstract

This research introduces a novel, cross-disciplinary framework for analyzing earnings management behavior in emerging economies by integrating computational linguistics, network theory, and institutional sociology. Moving beyond traditional accrualbased models, we propose that earnings management is not merely an accounting
phenomenon but a complex socio-technical signal embedded within the broader digital information ecosystem of a firm. Our methodology employs a hybrid technique of
transformer-based natural language processing (NLP) on narrative sections of annual
reports, coupled with a dynamic network analysis of a firm’s digital footprint—including
press releases, social media sentiment, and regulatory filing metadata. We conceptualize financial reporting credibility as a function of the coherence and resilience of this
information network, rather than a binary outcome of audit opinion. Applying this
framework to a unique longitudinal dataset from 500 firms across Brazil, India, and
Vietnam from 2015 to 2023, we uncover previously unobserved patterns of ”signaling
cascades,” where subtle linguistic manipulations in management discussion are amplified through correlated digital channels to shape market perception, often independent
of underlying accrual adjustments. Our results demonstrate that in institutionally weak
environments, sophisticated earnings management migrates from the balance sheet to
the narrative and digital periphery, creating a credibility gap that traditional audit and
compliance mechanisms are ill-equipped to detect. The findings challenge the primacy
of quantitative financial statement analysis in emerging markets and advocate for a
holistic, systems-oriented approach to assessing reporting quality that accounts for the
multi-channel, narrative-driven nature of modern corporate communication.