FRANKFURT (Reuters) -A ChatGPT analysis of qualitative commentary in PMI releases can significantly improve GDP forecasting, a paper published by the European Central Bank of Thursday showed.
The ECB has been working with artificial intelligence in recent years, partly to improve its forecasting ability by web-scraping price data and using large language models for data classification.
However, this new study found that even relatively little text associated with PMI releases can provide meaningful guidance on economic developments, improving its ability to assess activity in real time.
“What makes this study unique is its focus on the narrative, tone, and anecdotes reported in PMI news releases,” the working paper argued.
Economists used ChatGPT to generate activity sentiment scores based on the narratives and anecdotes of PMI news releases, then integrated these scores into forecasts of growth in the current quarter, or nowcasts.
“The main compelling result is that the enhancement of the PMI text scores to the two GDP nowcast benchmarks significantly improves the accuracy of GDP nowcasts,” the authors argued.
“Only two pages of text rather than for example millions of newspapers articles can be sufficient to enhance existing hard-to-beat benchmarks,” they added.
(Reporting by Balazs Koranyi; Editing by Hugh Lawson)
