Scholarlink

<p>We propose a novel sentiment-based index of global financial stress for the period between January 2004 and July 2025. It builds on the dictionary comprising English terms related to financial instability and selected with the aid of more than 200 large language models. The index represents the first principal component from the data series measuring the intensity of search in Google for these terms. It notably spikes with the GFC in September–October 2008, the outbreak of COVID-19 in March 2020, the onset of the Ukrainian conflict in February 2022 and the turmoil in the US banking sector in March 2023. The index is not driven by the extant financial stress or uncertainty measures, exhibiting moderate predictive power for some of them. Furthermore, it produces a detrimental effect on global real economic activity when the latter is in decline. The index Granger causes the worldwide frequency of currency crises, while exhibiting bidirectional linkages with the frequency of banking crises, triple episodes involving banking, currency and debt crises, and the implementation of macroprudential policy measures. Finally, we show that the dictionary underlying our global index can apply to elaborate country-level financial stress indices, using the USA and the UK as an example.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Measuring global financial stress: is there any role for large language models?

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Mikhail Stolbov,
  • Maria Shchepeleva

摘要

We propose a novel sentiment-based index of global financial stress for the period between January 2004 and July 2025. It builds on the dictionary comprising English terms related to financial instability and selected with the aid of more than 200 large language models. The index represents the first principal component from the data series measuring the intensity of search in Google for these terms. It notably spikes with the GFC in September–October 2008, the outbreak of COVID-19 in March 2020, the onset of the Ukrainian conflict in February 2022 and the turmoil in the US banking sector in March 2023. The index is not driven by the extant financial stress or uncertainty measures, exhibiting moderate predictive power for some of them. Furthermore, it produces a detrimental effect on global real economic activity when the latter is in decline. The index Granger causes the worldwide frequency of currency crises, while exhibiting bidirectional linkages with the frequency of banking crises, triple episodes involving banking, currency and debt crises, and the implementation of macroprudential policy measures. Finally, we show that the dictionary underlying our global index can apply to elaborate country-level financial stress indices, using the USA and the UK as an example.