Income Inequality in the New Age of AI

If you haven’t already, please check out part 1 of this series before reading this article.

How Previous Technologies Have Created Inequality

Less-educated Americans have encountered weak demand for their labor and stagnant or declining wages (Autor & Dorn, 2013). Autor (2014) found that the cause of this phenomenon was skill-biased change. Such technological change disadvantaged low- and medium-skilled workers’ wages (Dauth et al., 2017; Graetz & Michaels 2018). Thus, the polarization of labor significantly contributed to widening gaps in wages. Furthermore, Dauth, Findeisen, Südekum, and Wößner found that technologies raised labor productivity but not wages, thereby contributing to a declining labor income share. A declining labor income share implies an increase in income concentration among owners of capital, amplifying the effect of income inequality (Acemoglu & Restrepo, 2018; Autor, 2022).

Konirek and Stiglitz (2017) identified two channels through which automation gives rise to such income inequality: innovators receiving disproportionate portions of surplus rents and the ability of AI as a general-purpose technology to affect several sectors, modifying wages, demand for labor and capital, and price of products and the products themselves. As such automation consistently has resulted in income and wage inequality (Felten et al., 2019)

How AI Plays into the Discussion

AI may echo the exacerbation of income inequality. While attempting to control for endogeneity, AI innovation increased within-firm inequality as the median worker’s wage grew slower than that of the highest earners at AI-innovating establishments.  (Alderucci et al., 2019). Webb (2020) finds slightly different results applied to the entirety of the labor market, estimating that income inequality will compress in the middle, with the 90:10 ratio decreasing by between 4-9%, and expanding at the top of society, with the 99:90 ratio marginally increasing. Regardless, automation and AI’s displacement effect can potentially increase income inequality and reduce the share of labor in national income, disadvantaging owners of labor (Acemoglu & Restrepo, 2018).

References

Google Document with all references