Novel AI-driven methods to measure economic activity

Peter Lambert, Thiemo Fetzer, Bennet Feld, Prashant Garg

This paper leverages generative AI to build a network structure over 5,000 product nodes, where directed edges represent input-output relationships in production. We lay out a two-step build-prune approach using an ensemble of prompt-tuned generative...

Peter Lambert, Stephen Hansen, Nick Bloom, Steven Davis, Raffaella Sadun, Bledi Taska

We examine more than 250 million job vacancy postings across five English-speaking countries to measure the pandemic-driven shift to remote work. Our state-of-the-art NLP framework achieves 99% accuracy in flagging postings that advertise remote work...

Peter Lambert, Yannick Schindler

We deploy Big Data and LLM tools to digitise 36 million loan records, building a novel dataset on the credit relationships of 1.8 million US firms. Using 179 bank failures from 1990 to 2023, we find that firms banking with a subsequently failed insti...

Yannick Schindler, Peter Lambert

This paper charts technological progress embodied in capital equipment. We digitise archival administrative filings from 1998 to 2024 and extract 50 million capital equipment transactions from five large US states. We deploy an agentic AI measurement...

Yannick Schindler, Andrew Scott

We quantify the macroeconomic consequences of chronic illness in the UK, combining health data with macroeconomic indicators to estimate the effects of disease burden on employment, output, and government finances.

Peter Lambert, Aniket Baksy, Daniel Chandler

We study the adoption and impact of CNC machines and industrial robots in UK manufacturing using novel granular data on automation equipment.