Generative AI, perceived job displacement, and policy preferences: Experimental evidence from China

The rapid advancement of generative artificial intelligence (AI) has raised concerns about its potential to displace white-collar employment, yet little is known about how affected workers change their policy preferences in response to such risks. This paper presents experimental evidence from a field survey of 758 white-collar workers in Guangzhou, China. To prime individuals to consider labor market risks posed by generative AI, participants were randomly assigned to receive information about the large language model exposure rates of white-collar occupations. Treated individuals became significantly more likely to believe that their own jobs and those of other white-collar workers are at risk of replacement by AI. Treatment also led to a measurable shift in policy preferences: support declined for government subsidies to the AI industry, while support increased for progressive social policies including expanded unemployment insurance, universal basic income, and publicly funded retraining. Notably, the treatment did not increase support for education reforms emphasizing social or STEM skills, but it did increase the share of respondents who prefer that their children pursue public-sector careers, primarily for reasons of job stability. These patterns suggest a shift toward nonmarket risk-mitigation strategies rather than proactive upskilling or interest in private-sector opportunities.