Skip to main content

PRINZ Seminar

Chatting about innovation

Prashant Garg (Imperial College London)


Tuesday 30 September 2025 11:00 - 12:00

This event is both online and in person

SAL 2.04, 2nd Floor Conference Room, Sir Arthur Lewis Building, LSE, 32 Lincoln's Inn Fields, London WC2A 3PH

About this event

Patents are the workhorse indicator of innovation, but they miss large parts of inventive activity, especially in lower-income economies and green technologies. Innovation surveys broaden coverage but are costly and sparse outside rich countries. In general, we lack a scalable way to observe \emph{what} firms bring to market across countries and industries. We show that a large language model (LLM) can “survey” public firm narratives at scale to recover comparable indicators of product and process innovation. We classify these introductions as distinct (novel), environmentally oriented (green), and by their labor margin: whether they augment or save labor. Applying the pipeline to 4.0 million firms in 48 countries, we benchmark LLM-measured innovation against official patent counts and find strong co-movement. We document five facts. First, the share of firms with distinct (relationally novel) introductions exceeds the share with any patent with an even larger measurement gap for green innovation. Second, composition differs: patenters cluster in codifiable fields in manufacturing and utilities, while non-patenters drive environmental and service introductions. Third, a higher green share aligns closely with frontier novelty. Fourth, many economies automate internally while selling tools that augment users, whereas manufacturing-heavy exporters like China, Japan, and South Korea save labor on both fronts; patenting firms are also more automation-tilted than non-patenters. Fifth, firms that innovate in green and novel ways usually do so by augmenting processes, not products. The approach offers a scalable methodology for studying business behavior. More broadly, it shows how LLMs can make large-scale business surveys feasible.


Participants are expected to adhere to the Events Code of Conduct.


This event will take place in SAL 2.04, 2nd Floor Conference Room, Sir Arthur Lewis Building, LSE, 32 Lincoln's Inn Fields, London WC2A 3PH.

The building is labelled SAL on the map. Enter the building via Lincoln's Inn Fields.

Campus
LSE Campus Map