Modeling Sustainable Market Volatility and Sectoral Decoupling Through FinBERT-Based Narrative Analysis

This research explores the structural transformation of contemporary financial mar-kets as they pivot from traditional fundamental indicators toward a "narrative eco-nomics" paradigm, where informational flows and collective sentiment strongly cor-relate with asset price discovery. The primary objective is to evaluate the efficacy of Computational Sentiment Analysis as a real-time early warning mechanism for mar-ket volatility, specifically addressing the critical latency gap inherent in official mac-roeconomic "hard data". Adopting a quantitative methodology rooted in Data Science, the study utilizes the FinBERT deep learning architecture to analyze a 12-month lon-gitudinal dataset (May 2025 – May 2026) of financial news and discourse. Framed strictly as an exploratory, multi-entity case study rather than a sector-wide analysis, the investigation focuses on strategic proxies of the 'Twin Transition' in Eastern Eu-rope: Technology (UiPath), Energy (OMV Petrom/Hidroelectrica), and Banking (Transilvania Bank). Consequently, the findings highlight localized, context-specific dynamics rather than establishing broad, sector-wide behavioral rules. The findings provide exploratory support for a potential "regime shift" in market behavior regard-ing sustainability; ESG narratives transitioned from being perceived as a systemic risk in late 2025 to a primary factor associated with market value by early 2026. The evi-dence indicates patterns consistent with a "digital resilience" effect, wherein technolo-gy assets successfully decoupled from the industrial stagnation of traditional proxies, such as the German Deutscher Aktienindex(DAX). Conversely, the energy sector dis-played diminishing marginal impact of economic narratives regarding geopolitical shocks, with investors increasingly prioritizing long-term transition risks and indus-trial demand over short-term alarmist headlines. The study concludes that unstruc-tured textual data serves as a vital leading indicator for market dynamics, underscor-ing the imperative for integrating advanced Natural Language Processing (NLP) into modern economic forecasting and resilient risk management strategies.