Research
Working papers and selected research projects.
Working papers
Abstract. Major AI announcements now rank among the key information events for equity markets, yet little is known about their firm-level transmission on the announcement day itself. This paper asks whether different AI-news types affect different subsets of AI-exposed firms, whether information transmission is centralized around Magnificent Seven (M7) stocks, and whether the main source node depends on the event type. Using high-frequency intraday data for M7 stocks and smaller AI-exposed firms, we estimate event-day networks with a DAG-based Directed R2 connectedness framework. The evidence shows that AI news does not follow a single market-wide transmission channel: the main source node changes across event types. M7 firms emerge as event-day hubs in compute and adoption-related events, whereas disruption-related information can shift the main source node to a non-M7 firm. The results indicate that AI-related information is transmitted through event-specific equity-market channels, creating spillover vulnerabilities when information transmission is organized around a few dominant source nodes.
Abstract. The output gap is a central but unobservable concept in macroeconomic analysis, and its estimation remains especially challenging for small open European economies exposed to common financial shocks. This paper examines whether European and global financial factors improve country-level output-gap estimates for 26 European economies. Using a Borio-style state-space model, observed GDP is decomposed into latent potential output and a cyclical component, while the cyclical equation is augmented with one common factor at a time. The candidate factors include the Eurostoxx50 return, financial stress, Brent oil price changes, exchange-rate changes, euro-area unemployment, short-term interest rate changes and a euro-area share price index. The evidence shows that the Eurostoxx50 return is the strongest common factor according to likelihood-based criteria, improving the baseline model in most countries. Financial stress and oil prices also contain useful information, but their relevance is more selective and country-specific. However, Phillips-curve validation, real-time revision measures and visual inspection of potential-output paths give a more mixed picture. The results indicate that common European financial market information is relevant for output-gap estimation, but likelihood-based improvements alone are not sufficient to establish economic reliability.
Abstract. Small and medium-sized enterprises are central to the Hungarian economy, but their aggregate dynamics are difficult to measure because SME-related data are more limited and fragmented than standard macroeconomic indicators. This paper constructs a composite SME index for Hungary using annual indicators on firm activity, employment, personnel costs, turnover, value added, exports, investment and R&D. The indicators are deflated, transformed into annual growth rates, and aggregated using alternative weighting approaches based on inverse volatility, principal component analysis and factor analysis. The paper compares these methods in order to assess how different weighting choices shape the measurement of Hungarian SME dynamics.