EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Use of AI in market news sentiment analysis and its impact on investment decisions

Diana Kral-Grabka, Krzysztof Podgórski

International Journal of Public Administration, Management and Economic Development · 29.06.2026 · 11 (1) · 1-16

DOI: 10.60026/ijpamed.v11i1.259 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Public Administration, Management and Economic Development, online erschienen am 29.06.2026. Verfasst von Diana Kral-Grabka und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The aim of this study is to analyse the impact of sentiment surrounding extreme geopolitical news on the behaviour of the Polish small-cap index sWIG80 in the first quarter of 2026. This study fills a research gap concerning the reaction of the small-cap segment, dominated by emotionally driven retail investors, by testing the hypothesis of an immediate sell-off in the index in response to extremely negative news shocks. The research process employed a classical event analysis methodology for a narrow, symmetric window [-1; +1]. To assess the sentiment of 1,098 headlines sourced from Google News channels, the Gemini Pro large language model (LLM) was used in a zero-shot architecture, foregoing traditional dictionary-based methods. Subsequently, 13 extremely negative trading days were identified, constituting the lower quintile of the calculated net sentiment index. Market reaction was measured using average (AAR) and cumulative (CAAR) excess returns, verified using Welch’s t-test. The analyses conducted showed that the research hypothesis was not confirmed. The cumulative average abnormal return (CAAR) in the three-day event window was +0.13% with a complete lack of statistical significance (p=0.851), whilst on the day of the shock itself (t=0) a negligible change in AAR of -0.04% was recorded. It has thus been demonstrated that sWIG80 share prices proved highly resilient to superficial geopolitical noise. This study provides unique empirical evidence for the functioning of the semi-strong form of the efficient market hypothesis in the domestic small-cap segment and highlights the clear advantage of contextual inference by LLM models over lexical methods. It is also recommended that future research include an analysis of intraday quotes to examine intraday dynamics and the implementation of domain-specific models, such as Polish RoBERTa.

Zitieren

APA

Kral-Grabka, D., & Podgórski, K. (2026). Use of AI in market news sentiment analysis and its impact on investment decisions. International Journal of Public Administration, Management and Economic Development, 11(1), 1-16. https://doi.org/10.60026/ijpamed.v11i1.259

Vancouver

Kral-Grabka D, Podgórski K. Use of AI in market news sentiment analysis and its impact on investment decisions. International Journal of Public Administration, Management and Economic Development. 2026;11(1):1-16. doi:10.60026/ijpamed.v11i1.259

BibTeX
@article{KralGrabka2026,
  author = {Kral-Grabka, Diana and Podgórski, Krzysztof},
  title = {Use of AI in market news sentiment analysis and its impact on investment decisions},
  journal = {International Journal of Public Administration, Management and Economic Development},
  year = {2026},
  volume = {11},
  number = {1},
  pages = {1--16},
  doi = {10.60026/ijpamed.v11i1.259}
}
Kostenlos starten →