Generative AI Could Undercut Labor Income, Affecting the Entire Economy

Generative AI — the type that generates text and images — offers benefits and troubles. Time-saving to have software that can do near-creative tasks that once took people. Dangerous in its proclivity to invent facts.

Now, a new paper from economists suggests a new issue. “Unlike previous technologies, AI may undermine labor’s share of national income, and technological innovation could, for the first time, permanently reduce the importance of labor in the economy, even if full employment is maintained,” Lukasz Drozd, economic advisor and economist at the Federal Reserve Bank of Philadelphia, and Marina Tavares, economist at the International Monetary Fund, wrote.

That could translate into a negative impact on the economy and people’s ability to pay for goods and services, indirectly reducing demand for many forms of commercial real estate.

There have been multiple industrial revolutions in the past due to general purpose technologies, or GPTs, as the authors said. A GPT has four traits: a generic technology, wide adoption across the economy, multiple distinct applications, and the “spillover” generation of new innovations. “According to one analysis, as of the mid-2000s there had been at least 24 GPTs, ranging from language and the wheel to the steam engine and the computer,” they noted.

Many people counter predictions of job elimination by technology by saying it hasn’t happened before. But generative AI has some significant differences from previous GPTs. “The concerning aspect of AI, as we see it, is that it is a major GPT with the potential to broadly and persistently tilt the incoming flow of new capital-productivity-augmenting innovations toward those that automate tasks, rather than augment the productivity of capital in previously automated tasks,” the researchers argued.

In the past, changes in technology added to labor’s share of national income. Products became cheaper, leaving workers with more money in constant purchasing power. “Simply put, to purchase goods, labor must effectively pay for its own input into production and for capital (including profits),” they wrote. “Since capital costs less, labor can purchase more goods for the income it earns, and so its share rises at the expense of capital.”

What prevented wide displacement of labor is that previous implementations of technology weren’t easily exchanged between industries and tasks. Generative AI is different, the economists note. Although the technology doesn’t think or comprehend in a human way, it could mimic the ability closely enough to enhance machine productivity in areas that hadn’t been automated.

Even if there was growth in income for workers, the declining share would increase income and wealth inequality. One example they offered was the fixed supply of land. Those who want to purchase could use their average ability to bid higher to lock out people with less. And “even if labor income continues to grow, housing may become unaffordable for those who supply labor. It is perhaps no coincidence that labor unrest coincided with a persistent but ultimately transient decline in labor’s share of income in the 19th century.”

More generally, consumer spending constitutes 68% of GDP. Displaced workers can’t spend as much, and wealthier people can’t consume enough to make up the difference. Providers of goods and services would take a hit to revenues and profits. That would mean less need for commercial real estate.