Economist Jon Hartley recently highlighted a new working paper on generative AI and the labor market in a thread on X: Jon Hartley thread. The paper, available on SSRN, is co-authored by Jolevski, Melo, and Moore.

Summarizing the findings, Hartley writes: "Generative AI adoption is widespread, but substantial aggregate labor-market disruption is not yet visible. Workers nevertheless perceive substantial displacement risk, especially when firsthand use reveals that AI can perform key tasks for their job."

The study adds to a growing body of research examining how generative AI tools are being adopted across the workforce and whether that adoption is translating into measurable changes in employment. That literature includes earlier experimental work by Erik Brynjolfsson, Danielle Li, and Lindsey Raymond on customer-support agents — which found productivity gains concentrated among less-experienced workers — as well as surveys such as the Census Bureau's Business Trends and Outlook Survey, which have tracked rapid diffusion of AI use across firms. While large language models and similar tools have spread rapidly among workers since their introduction, aggregate employment data have so far shown limited evidence of widespread displacement, according to the paper's findings as characterized in Hartley's thread.

The research also points to a notable gap between perception and observed outcomes: even where aggregate disruption is not yet measurable, individual workers report meaningful concern about being displaced, and that concern appears strongest among those who have used AI directly and seen it perform tasks central to their own jobs. That perception-versus-outcome gap matters for policymakers, because worker fears about displacement can influence training decisions, job switching, and wage negotiations even before aggregate statistics register any change — and because earlier episodes of technological change, such as computerization, have historically been followed by lagged labor-market adjustments that are difficult to detect in real time with coarse aggregate data.

For further context, Marginal Revolution's Alex Tabarrok published a related post the previous day on AI and employment, drawing on what firms themselves report about AI's effects on hiring and staffing. Together, the paper and the accompanying commentary reflect ongoing debate among economists about whether the labor-market effects of generative AI have simply not yet materialized at scale, or whether existing data are not yet granular enough to capture them. A natural question for subsequent research is whether task-level and occupation-specific datasets will begin to show effects that national aggregates have so far obscured.

Source: Marginal Revolution