BEJ

This paper examines the relationship between technological change and labor market outcomes using cross-country and cross-industry data spanning 25 years (1995–2020). Specifically, it investigates the impact of automation, proxied by total factor productivity (TFP) growth, on two key labor market indicators: aggregate employment and aggregate labor share of value added. The theoretical framework of this paper derives from Autor and Salomons (2018) and delineates four channels—comprising one direct effect and three indirect effects—by which automation influences labor market outcomes. This paper extends A&S’ analysis by 13 years and makes methodological changes by revising the lag structure of TFP growth to account for longer, more variable innovation-to-productivity effects and incorporating previously omitted controls to capture the final demand effects of automation. Therefore, the primary goal of this paper is to reassess Autor and Salomons (2018)’s analysis by employing methodologies better suited to the new data. The theoretical framework in this paper is based on the core idea that, while certain technological innovations may displace labor, countervailing responses within the economy can mitigate downward shifts in aggregate labor demand, making it crucial to estimate both the direct and indirect effects of automation. Consistent with Autor and Salomons (2018), this paper finds that for both employment and labor share of value added, there is a negative direct effect of automation on labor outcomes in the industry where the innovation occurs. For employment, this negative direct effect is offset by countervailing forces elsewhere in the economy — including upstream and downstream linkages, final demand effects, and compositional changes — yielding a net positive effect of automation. In contrast, the net effect on labor share of value added is negative, reflecting the broader trend of declining labor shares observed in developed countries over recent decades. However, our quantitative analysis also reveals a significant discrepancy compared to A&S’ findings: the size of the net positive impact of automation on aggregate employment is notably diminished. This disparity suggests that recent technological advancements, such as advanced artificial intelligence and robotics introduced in the last decade, may differ fundamentally in their labor market effects from older technologies. Consequently, these newer technologies may exert a less favorable influence on labor outcomes, underscoring the need for nuanced understanding and strategic adaptation to technological change in contemporary economies.

JEL classification: J21, J24