Enhancing GDP nowcasts with ChatGPT: A novel application of PMI news releases

Abstract

This study tasks ChatGPT to classify activity sentiment scores from PMI news releases and assesses their incremental value for nowcasting euro area GDP. We find that incorporating these ChatGPT-derived scores significantly enhances euro area GDP nowcasts relative to both ECB/Eurosystem Staff projections and Eurostat’s first GDP estimate, even when controlling for the traditional PMI diffusion index. These improvements persist across various models, GDP growth quantiles, and excluding extreme events like the pandemic. On average, out-of-sample forecast accuracy improved by about 20% apart from the two most recent years. Our findings highlight the potential policy value of integrating targeted qualitative data into economic forecasting.