When presidents make the news, do crude oil markets listen? An LLM-driven analysis

This study investigates the impact of presidential statements by the leaders of the United States (US), China, and Russia on crude oil markets. Using a dataset of 20,998 full-text news articles and headlines from major international newspapers, we analyze the period from March 2018 to June 2022. We employ open-source Large Language Models (LLMs) to score each presidential statement on a scale from −1 (significant price decrease) to 1 (significant price increase). From these scores, we construct a series of geopolitical indices, referred to as Geopolitical Oil Presidential Risk (GOPR) indices. Among the tested models, Llama 3.1 405B proves most effective at capturing geopolitical sentiment relevant to crude oil markets, particularly when this sentiment is measured using full news articles rather than headlines. Findings from autoregressive models with exogenous variables (ARX) suggest that presidential statements carry statistically significant explanatory power for daily crude oil returns. However, the effects vary across leaders and markets. For example, statements by Trump significantly influence returns in the Brent, West Texas Intermediate (WTI), and Shanghai crude oil futures (SC) markets, whereas statements attributed to Xi Jinping only affect the SC market. Conversely, statements by Biden and Putin show no measurable impact on any crude oil market.