Date: 19 December 2025, Friday
Time: 13.30 – 14.30
Place: MA-330
“Attention Predicts, Tone Reflects: Mixed-Frequency Evidence on Volatility Drivers in Borsa Istanbul”
by
Alev Atak
METU
Abstract
We examine how retail investor attention and corporate disclosure tone jointly relate to stock market volatility in Türkiye, an emerging market characterized by high retail participation and limited institutional coverage. A monthly Financial Attention Index (FAI) is constructed by applying principal component analysis to Turkish-language Wikipedia pageviews for finance-related terms; disclosure tone is measured using FinBERT-derived sentiment scores from Borsa Istanbul annual reports (2016–2024), weighted by posterior probability and market capitalization. Mixed Data Sampling (MIDAS) regressions integrate daily volatility with monthly attention and annual tone data, preserving information that would be lost under temporal aggregation. Granger causality tests establish that lagged attention predicts volatility, with no evidence of reverse causality, and a difference-in-differences design around the February 2023 earthquake reveals a temporary strengthening of this attention–volatility link during periods of heightened uncertainty. The tone–volatility association, negative in baseline models, becomes statistically insignificant after including year fixed effects—suggesting that disclosure tone primarily reflects prevailing macroeconomic conditions rather than exerting an independent within-year behavioral impact on prices. These findings highlight divergent roles for behavioral signals: Wikipedia-based attention indices capture high-frequency sentiment dynamics suitable for real-time market surveillance, while annual disclosure tone operates as a slower-moving indicator tied to the broader economic cycle. The results support the use of scalable, text-based monitoring tools in retail-heavy emerging markets where traditional attention proxies may be unavailable or costly.
Bio
Alev Atak is Associate Professor of Economics at Middle East Technical University (METU), where she also serves as Associate Director of the Institute of Applied Mathematics (IAM). She received her PhD from Queen Mary, University of London, and held a faculty position in the UK before returning to Türkiye. Her work has appeared in journals including the Journal of Econometrics, Econometric Theory, and Borsa Istanbul Review. Her current research sits at the intersection of machine learning and finance, turning unstructured information into quantifiable signals for understanding market behavior. Using deep learning models originally developed for language understanding, she extracts sentiment, attention, and ESG-related signals from corporate disclosures and online platforms, studying how these textual measures shape market dynamics in emerging economies and cryptocurrency markets.