If AI helps produce more at lower cost and fewer people have sufficient or stable income, who buys everything companies can now produce more efficiently?
The discussion about technology and employment is not only about jobs. It is about aggregate demand. An economy can improve productivity, automate processes, and reduce costs. But if at the same time it reduces the wage mass or concentrates income in a small slice of the population, it also limits the purchasing capacity that sustains sales, investment, and expansion.
Daron Acemoglu, an MIT economist, has been warning for years that a large share of modern automation displaced workers faster than it created equivalent replacements. The effect was not only lower employment in certain segments, but also substitution toward occupations with lower income or less stability.
The result was a more polarized structure: high-skill jobs grow and low-wage jobs also grow, while the weight of middle jobs declines. From a macroeconomic point of view, that matters because middle-income groups usually have a higher propensity to consume than top-income groups.
That is where a classic tension appears: companies need efficiency to compete, but they also need demand to sell.
If productivity rises and a relevant share of the benefits does not translate into more broadly distributed income, the economy gains productive capacity but may lose absorptive capacity. More gets produced, but not necessarily more gets sold at the same pace.
It is also worth avoiding linear explanations. Not everything was driven by automation.
In several analyses of the United States, trade liberalization, offshoring, and certain deregulation processes explain a large part of the industrial and wage deterioration of recent decades. The so-called China Shock is often cited as an example of competitive pressure that affected employment, wages, and regional economies.
While people were arguing about robots, many of the key decisions were being signed off in Excel 2007.
That discussion returns today with artificial intelligence. It can raise productivity, lower operating costs, and accelerate innovation. But it can also deepen a familiar problem: a growing capacity to produce without an equivalent expansion of demand.
The same tool can complement workers, raise incomes, and expand markets, or replace employment and concentrate rents. The final impact depends less on the algorithm than on how it is implemented and how the productivity gains are distributed.
Because without enough consumers, even the most efficient company ends up selling less than it expected. And when less is sold, less is invested and growth also slows.
I pulled this from here:
- MIT News — Study links automation to growing wage inequality
- AEA — Automation and New Tasks: How Technology Displaces and Reinstates Labor
- China Shock working paper
- Import Competition, Unemployment, and the Decline of Local Labor Market Adjustment
- Chinese Imports and Innovation, IT and Productivity
- World Bank document on jobs, trade, and technology