Automation Experiments and Inequality

TL;DR

This paper formalizes how automation impacts inequality, showing that the effects depend on task skill correlations and technology capabilities, with inequality potentially decreasing then increasing as technology advances.

Contribution

It provides a theoretical framework for understanding automation's inequality effects, emphasizing the roles of task skill correlation and technology capability.

Findings

Inequality effects depend on task skill correlation and technology capability.

The sign of inequality impact can be non-monotonic as technology improves.

Diversity of automation technologies influences inequality evolution.

Abstract

An increasingly large number of experiments study the labor productivity effects of automation technologies such as generative algorithms. A popular question in these experiments relates to inequality: does the technology increase output more for high- or low-skill workers? The answer is often used to anticipate the distributional effects of the technology as it continues to improve. In this paper, we formalize the theoretical content of this empirical test, focusing on automation experiments as commonly designed. Worker-level output depends on a task-level production function, and workers are heterogeneous in their task-level skills. Workers perform a task themselves, or they delegate it to the automation technology. The inequality effect of improved automation depends on the interaction of two factors: ($i$) the correlation in task-level skills across workers, and ($ii$) workers' skills relative to the technology's capability. Importantly, the sign of the inequality effect is often non-monotonic -- as technologies improve, inequality may decrease then increase, or vice versa. Finally, we use data and theory to highlight cases when skills are likely to be positively or negatively correlated. The model generally suggests that the diversity of automation technologies will play an important role in the evolution of inequality.