Research Seminar "Large Language Models and the Labour Market: Spatial Evidence from Job Ads" (Eszter Baranyai (Central Bank of Hungary))

Large Language Models and the Labour Market: Spatial Evidence from Job Ads

Eszter Baranyai (Central Bank of Hungary)

Wednesday, 19 March 2025 at 14.00-15.00  Zoom

Little is known about the spatial variation of jobs' exposure to large language models (LLMs) both within and across countries despite LLMs' spectacular rise in popularity in recent years and the repercussions of technological leaps on regional productivity and employment trends. We webscrape detailed task descriptions from all job ads listed on the largest online job portal in Hungary and apply a mapping approach. Extrapolating to the county's labour market, we estimate average exposure across jobs to LLMs in Hungary to be around 8% - somewhat lower than in the US. Mapping different occupational classifications, Hungary has a higher share of occupations associated with physical labour - and thus with lower LLM potential - and a lower share of LLM-exposed office occupations. Very rarely does job-level exposure exceed 30% in our Hungarian sample, highlighting complementary attributes. Of the factors studied, industry appears most closely related to LLM exposure. Exposure is higher in more urban areas with a higher population of young adults. Results draw attention to areas - spatial and industry - where productivity benefits could occur and where education and employment policies could help manage the effects of technological transition on employment. Co-authors: Marcell Granát (Central Bank of Hungary), Mór Szepesi (Yale University, Central Bank of Hungary)

Co-authors:  Marcell Granát (Central Bank of Hungary), Mór Szepesi (Yale University, Central Bank of Hungary)