Essays in applied economics using novel data and AI measurement

Lambert, P.ORCID logo (2025). Essays in applied economics using novel data and AI measurement [Doctoral thesis]. London School of Economics and Political Science. https://doi.org/10.21953/lse.00004934
Copy

This thesis examines the transformative role of artificial intelligence in applied economics research and measurement. Chapter I surveys recent advancements in AI-powered economic measurement, positioning these tools as instruments that expand the domain of measurable economic phenomena. The chapter discusses how large language models enable economists to extract signals from vast amounts of unstructured data while addressing methodological challenges unique to AI-based measurement. The subsequent chapters present three original research contributions which deploy AI tools and novel data to address contemporary questions relevant to policy makers. Chapter II analyses the shift toward remote work using an innovative approach to classify over 250 million job vacancy postings across five English-speaking countries. By developing a specialized language-processing model that achieves near-human accuracy, the research tracks remote work patterns at unprecedented granularity, revealing insights that challenge conventional understandings of remote-work suitability. Chapter III introduces the AI-generated Production Network (AIPNET), which maps granular network structures spanning more than 5,000 product categories. Using a novel “build-prune” methodology with generative AI, the analysis reveals shifts in global trade patterns and production localization, showing movement toward more centralized upstream products and providing causal evidence of onshoring responses to supply shocks. Chapter IV investigates bank failures’ effects on firm performance by applying AI to process hundreds of millions of loan documents, developing insights into how financial shocks affect small and medium-sized enterprises during both crisis and non-crisis periods. Together, these chapters demonstrate how AI-powered measurement techniques extend the frontier of applied economics while maintaining econometric rigor, contributing both methodological innovations and substantive findings on pertinent contemporary economic questions.

visibility_off picture_as_pdf

subject
Submitted Version
lock_clock
Restricted to Repository staff only until 2 October 2026


Export as

EndNote BibTeX Reference Manager Refer Atom Dublin Core JSON Multiline CSV
Export