Integrating News-Derived Events and Structured Variables for Causal Discovery: Code and Data

This repository contains the code and data accompanying the study Integrating News-Derived Events and Structured Variables for Causal Discovery: A Case Study on UK Inflation, by Martina Cesca, Fernando Delbianco, Fernando Thome, and Ana Maguitman The repository supports the analysis of relationships between economic events extracted from financial news and structured economic indicators. The study focuses on UK inflation and covers the period from January 2016 to June 2022. Financial news articles from The Guardian are processed using a large language model to identify and classify economic events. The resulting event mentions are aggregated into monthly frequency measures and integrated with quantitative economic variables to construct multivariate time series and causal graphs. The repository contains: Monthly frequencies of economic events extracted from financial news articles. URLs for the The Guardian articles included in the news corpus. Structured economic variables relevant to UK inflation. Jupyter notebooks for time-series analysis, exploratory analysis, and causal discovery. The economic variables were collected from multiple official and market-based sources, including the Office for National Statistics (ONS), the Bank of England, Yahoo Finance, Investing.com, S&P Global/CIPS, and OECD/GfK. The causal analysis applies multiple causal-discovery methods to the news-derived event variables, the structured economic variables, and their integrated representation. The repository is intended to support reproducibility, facilitate further research on causal analysis from heterogeneous data sources, and enable the extension of the proposed methodology to other domains and datasets. The repository includes article URLs and derived data; users should consult the terms of use of the original data providers when accessing or redistributing source materials.