Cross-Document Emotion Consistency (CDEC): A Feature Family Framework for Financial Disclosure Risk Screening

Financial disclosure occurs through multiple channels with fundamentally different legal constraints. Mandatory SEC filings undergo extensive legal review under Sarbanes–Oxley Section 302, while the earnings call Q&A segments represent comparatively spontaneous communication protected by safe harbor provisions. This structural difference creates a natural experiment for detecting management of information asymmetry through emotional consistency analysis. This paper presents Cross-Document Emotion Consistency (CDEC), a framework measuring emotional alignment between Management’s Discussion and Analysis (MD&A) sections and earnings call Q&A using domain-adapted 28-dimensional emotion classification. The framework jointly analyzes MD&A and Q&A for cross-channel emotion alignment, linking 8-K filings as external event validation. Neither CDEC nor financial fundamentals alone achieve meaningful risk separation; their interaction does. The integrated framework achieves approximately twice the risk separation of the strongest 3-class sentiment baseline tested (27.4 percentage points vs. 14.6 pp for RoBERTa-based consistency and 8.7 pp for FinBERT), indicating that 28-dimensional emotion granularity captures disclosure risk not detected by standard sentiment classification. The Financial Health Indicator (FHI) achieves an out-of-sample Area Under the Curve (AUC) = 0.671, distinguishing firms likely to beat Standardized Unexpected Earnings (SUE) expectations (62.8% beat rate among large-cap sector-representative firms) from those likely to miss (35.4%). The framework serves as a risk-screening tool for due diligence, internal audit, and regulatory oversight among large-cap firms, identifying firms warranting scrutiny rather than generating trading signals.