A new approach to economic research.
The forces that truly drive outcomes have always lived beyond our data.
Generative AI transforms unstructured data into measurable signals — opening doors that traditional methods couldn't unlock.
Text and image embeddings create numerical proxies for latent factors like ability, motivation, and quality — variables that bias results when ignored.
LLMs convert unstructured text (reviews, reports, news) into structured databases with consistent categories — trackable over time like any economic indicator.
Knowledge graphs reveal relationships invisible in traditional data — which skills connect careers, which occupations share common pathways.
Used resume text embeddings as proxy for unobserved worker ability in wage models — eliminated omitted variable bias.
Converted 16,000 airline reviews into a database of 36 service issue categories — now trackable quarter over quarter.
Built a knowledge graph linking 10,000 occupations through 84,000 skill relationships — revealed 24% more transition paths than job titles alone.
These methods are actively applied at ECES in real projects — delivering insights for the Egyptian government and international partners including the World Bank, USAID, and GIZ.