Automation, financial frictions, and the long-run labor share

Technological advance has allowed the automation of many tasks historically performed by workers. Recent progress in artificial intelligence raises the prospect that it may ultimately be possible to automate all tasks.In this paper, we take seriously the possibility that all tasks will ultimately be automated, and analyze the consequences of such automation. The starting point for our analysis is that the consequences of automation depend on evolution of the wages and the returns to capital. A central force in our analysis is Baumol and Bowen (1965)’s insight that productivity improvements in the production of a subset of goods and services increase the price of distinct but complementary goods and services.

Specifically, in this paper we analyze a minimal model in which all tasks are asymptotically automated; economic agents are heterogeneous, with “workers” and“capitalists,” endogenously emerging; and in which capital accumulation, labor supply, and factor returns are all determined by standard economic forces. We use this framework to answer a variety of questions, including the following. Do the productivity improvements associated with automation raise or lower labor income?Does automation lead to inequality? If automation leads to a decrease in laborincome, to what extent can workers protect themselves by saving and accumulating capital? Do financial frictions that hamper workers’ ability to accumulate capitalmake the demise of the labor share more or less likely? Does the tendency ofworkers to work less as they grow richer hinder or promote a fall in the labor share?

Our analysis yields a threshold pace of automation that determines whether or not the labor share of the economy shrinks to zero. For automation speeds below this threshold, Baumol’s effect dominates. That is: Even though all tasks are asymptotically automated, the price of non-automated tasks, and hence wages, grows sufficiently fast that the labor share converges to a strictly positive limit.