Epsilon Archive for Student Projects

Abstract

This study examines the relationship between generative artificial intelligence exposure and labor market outcomes across Swedish economic sectors. Using AI exposure measures developed by Felten et al. (2021) and converted from the North American Industry Classification System (NAICS) to the Swedish Standard Industrial Classification (SNI), the study identifies sectors with relatively high and low exposure to generative AI technologies. A difference-in-differences framework combined with an event study specification is employed to analyse the effects of AI exposure on hourly wages and employment levels using quarterly data on the sector level.

The results indicate that highly AI-exposed sectors experienced relatively weaker wage developments during the post-treatment period compared to less exposed sectors. The estimated wage effects remain statistically significant across multiple model specifications, although the magnitude decreases following the inclusion of additional control variables. The analysis identifies more limited and less consistent effects on employment levels. The event-study estimates provide partial support for the parallel trends assumption while suggesting that treatment effects evolve gradually over time.