Despite apparent rapid technological advancement, US GDP growth has slowed considerably, from an annual rate of 2.51% (1948-1968) to just 1.12% (2008-2022). This paradox raises fundamental questions about the relationship between technological innovation and economic growth. Additionally, TFP gaps remain a significant driver of GDP per capita differences across countries, making it crucial to understand how technology diffusion affects cross-country productivity disparities.

A methodological fusion of cutting-edge large language models (LLMs) with time series methods to uncover the hidden drivers of productivity growth, this study creates a uniquely granular measure of technological exposure across technology fields by leveraging LLMs to extract semantic relationships between patent texts and work tasks. This novel integration of natural language processing with a VAR framework allows us to trace the complex pathways through which technological innovations propagate through the US economy and ultimately diffuse across borders, offering insights into the mechanisms of productivity growth both domestically and internationally.

By examining how innovations diffuse across borders and contribute to global TFP differences, the study could inform policies aimed at accelerating beneficial technology transfer between countries. This work may help policymakers design targeted approaches to technological development that promote both growth and more equitable outcomes, potentially narrowing productivity gaps that underlie global economic disparities.