Abstract:
According to CNBC, as wage growth for American workers slows, economists are beginning to focus on another impact of AI on the job market. Compared with directly replacing jobs, AI may weaken the wage growth and bargaining power of some workers earlier.
The latest U.S. non-farm payrolls data were better than expected, but wage growth lagged behind inflation. The Employment Cost Index released by the U.S. Bureau of Labor Statistics shows that as of June this year, real wages and salaries in the United States fell by 0.4% year-on-year.
Workers' share of economic output is also declining. In the second quarter of 2026, the labor income of the U.S. non-agricultural business sector fell to 52.8% of output, the lowest level since the statistics began in 1947. Some researchers believe that decades of automation have driven this change, and that AI could further accelerate the process.
Existing data are not enough to prove that AI has caused a slowdown in U.S. wage growth. The labor market was extremely tight during the epidemic, and wages rose rapidly. The current growth rate is closer to the pre-epidemic level. At the same time, high-wage industries such as technology and professional services are laying off workers, and lower-wage industries such as hotels, restaurants, and medical care have become the mainstay of new employment. This change in the employment structure will also drive down average wages.
However, more and more studies are beginning to try to find the impact of AI on wages from actual data.
A recent study released by Apollo Global Management chief economist Thorsten Slok and researcher Sania Edlich shows that after 2023, the real wage growth of occupations with high AI exposure will be 6.7 percentage points slower than that of occupations with lower AI exposure. The study found no statistically significant impact of AI on employment numbers.
Two researchers proposed based on this that some companies may obtain the productivity gains brought by AI by suppressing wage growth and have not yet implemented large-scale employee reductions.
This study had a small sample size. Of the approximately 800 occupations counted by the U.S. Bureau of Labor Statistics, the study could only use data from 321 occupations, and only 11 occupations met the high AI exposure standards. The researchers also called the conclusion "early evidence."
Ben Zippeler, senior economist at the Economic Policy Institute, believes that AI may indeed reduce corporate demand for some jobs and put downward pressure on wages, but the Apollo sample size is not enough to draw convincing conclusions.
He used the software development industry as an example. If AI reduces the cost of developing software, the money saved by companies may also flow to other departments, increasing the demand for other workers. In this case, the salary performance of positions with high AI exposure will appear worse, and part of the difference comes from increased income in other positions.
The adjustments brought about by previous over-recruitment in the technology industry cannot be ignored. After the epidemic, the recruitment demand for programmers and related positions has significantly cooled down, which may simultaneously drag down employment and wage growth, and is difficult to completely distinguish from the impact of AI.
Daron Acemoglu, an economics professor at MIT, also believes that there is currently no convincing evidence that AI has had a significant impact on wages in a certain industry or a specific group. Many occupations do not yet have widespread use of specialized AI tools, and some estimates of job displacement may be exaggerated.
However, he believes there is increasing evidence that AI is affecting entry-level jobs. The U.S. job market has high liquidity and the social security system is relatively weak, so the impact of AI on workers may ultimately be more reflected in lower wages than in the number of jobs.
David Otto, chairman of the MIT Economics Department, reminded that simply judging whether an occupation is "exposed to AI" cannot predict how its wages and employment will change. Whether technology takes over professional aspects of work or lowers the threshold for employment may bring completely different results.
Otto and MIT researcher Neil Thompson compared accounting clerks to inventory clerks. Over the past 40 years, employment of accounting clerks has decreased by 32%, but wages have increased by 39%. During the same period, employment of inventory clerks increased by 175%, while wages fell by 13%.
Both professions have been affected by computer technology, but accounting clerk jobs have become more specialized, and those who remain earn more. Inventory management positions are open to more workers, employment opportunities have increased, and wages have decreased.
An analysis released by the Federal Reserve Bank of Dallas in February also found no direct link between overall wages and exposure to AI, but the study showed that occupations with lower work experience requirements may already be facing more pronounced wage pressures.
This type of position relies less on implicit experience and professional judgment, and it is easier for AI to replace both newcomers and senior employees. The traditional white-collar training model may also be affected by this. In the past, companies would let fresh graduates start with basic jobs that could be standardized and gradually accumulate experience on the job. After AI takes over these tasks, companies may decide that the cost of training new employees is no longer cost-effective.
In the long term, companies will still need new senior employees. If newcomers can't get onto the career ladder, there won't be enough skilled workers in the future. Therefore, the popularity of AI may force companies to redesign entry-level positions to allow young employees to gain experience in new ways of working.
Acimoglu believes that AI can also help workers master new skills and take on new tasks, but the current main investment direction in the technology industry is still towards automation and replacing labor. If this direction continues, AI’s impact on wages and employment will increase as its application scope expands.
Comments