A new study from the Federal Reserve Board reveals that the growth rate for programming-heavy jobs in the US has nearly halved since the launch of ChatGPT in November 2022. Before the AI chatbot's release, these roles were expanding at nearly 5% annually, well above the overall labor market. But since then, the pace has dropped sharply, with sectors like IT services and software development essentially flatlining.
To isolate the AI effect from broader tech industry downturns—such as interest rate hikes, the end of the Covid-era boom, and the crypto crash—researchers built a counterfactual employment curve. Even after adjusting for industry-wide changes, programmer employment is falling by about three percentage points per year, suggesting companies are deliberately reducing their programmer headcount.
The gap between actual and expected employment translates to roughly 500,000 jobs over three years. However, the authors caution that this doesn't necessarily mean 500,000 job losses; many workers likely shifted to adjacent roles, as AI reorganizes tasks across occupations. Wages have not shown a clear decline.
Interestingly, the slowdown is most pronounced in the contract development sector, which employs about 40% of US programmers. The divergence in employment didn't appear until mid-2024, about 1.5 years after ChatGPT's launch, indicating companies took time to assess AI's capabilities before adjusting hiring plans.
The study also highlights a methodological challenge: there's no consensus on which occupations AI impacts most. But programming stands out—over 98% of measurement methods agree it's among the hardest hit.
While AI is a likely driver, the study notes other factors like a 2017 tax law change affecting R&D expensing, though its impact remains unclear. The authors emphasize their work is just a first step in understanding AI's long-term labor market effects.